Conference program

The conference program consists of five keynotes and 56 regular presentations. The program starts at 4:45 UTC and ends at 23:30 UTC. All times in the program are in Coordinated Universal Time or UTC. The time in the UK (GMT) is currently equal to UTC, the Central European Time (CET) is UTC+1, and in Boston, MA, the Eastern Daylight Saving Time (EDT) is UTC-4.

Inner speech is accompanied by a temporally-precise and content-specific corollary discharge

Bradley Jack (UNSW Sydney)
Co-authors: Mike E. Le Pelley and Nathan Han (UNSW Sydney), Anthony W. F. Harris (University of Sydney and Westmead Institute for Medical Research), Kevin M. Spencer (Harvard Medical School), Thomas J. Whitford (UNSW Sydney)
Click here for abstract
When we move our articulator organs to produce overt speech, the brain generates a corollary discharge that acts to suppress the neural and perceptual responses to our speech sounds. Recent research suggests that inner speech – the silent production of words in one’s mind – is also accompanied by a corollary discharge. Here, we show that this corollary discharge contains information about the temporal and physical properties of inner speech. In two experiments, participants produced an inner phoneme at a precisely-defined moment in time. An audible phoneme was presented 300 ms before, concurrently with, or 300 ms after participants produced the inner phoneme. We found that producing the inner phoneme attenuated the N1 component of the event-related potential – an index of auditory cortex processing – but only when the inner and audible phonemes occurred concurrently and matched on content. If the audible phoneme was presented before or after the production of the inner phoneme, or if the inner phoneme did not match the content of the audible phoneme, there was no attenuation of the N1. These results suggest that inner speech, similar to overt speech, is accompanied by a temporally-precise and content-specific corollary discharge; this finding is consistent with the notion of a functional equivalence between the neural processes that underlie the production of inner and overt speech, and provides empirical support for the influential, but hitherto untested, hypothesis that inner speech is a special form of overt speech. Finally, these results provide a foundation for investigating abnormalities in inner speech, such as auditory-verbal hallucinations in schizophrenia, and for the development of brain-computer interfaces capable of deciphering and utilizing inner speech for people who are unable to produce overt speech.

Noradrenergic modulation of cognitive information processing dynamics

Mac Shine (The University of Sydney)
Co-authors: Mike Li (Brain and Mind Centre and Center for Complex Systems, The University of Sydney), Dennis Hernaus (Maryland Psychiatric Research Center, University of Maryland School of Medicine), Eli Muller and Gabriel Wainstein (Brain and Mind Centre, The University of Sydney), Joseph Lizier (Center for Complex Systems, The University of Sydney)
Click here for abstract
Cognition involves the dynamic adaptation of information processing resources as a function of task demands. To date, the neural mechanisms responsible for mediating this process remain poorly understood. In this study, we integrated cognitive neuroscience with information theory, network topology and neuropharmacology to advance our understanding of the fundamental computational processes that give rise to cognition in the human brain. In our first experiment, we translated dynamic whole-brain blood oxygen level dependent data from a cognitively-challenging N-back task from the Human Connectome Project (N = 457) into information theoretic time series. Our results show that cognitive task performance alters the whole-brain information processing landscape. We next established that the information theoretic patterns were spatially coincident with patterns of dynamic task-based network topology. Finally, we modulated central noradrenaline levels in a double-blind, cross-over atomoxetine pharmacological fMRI study (N = 19). We found that potentiating the noradrenergic system altered information dynamics, shifting the frontoparietal cortices from a storage to transfer processing mode. Together, our results provide a conceptual bridge between cognitive function, network topology, information theory and the ascending neuromodulatory arousal system.

The Brain Dynamics Toolbox for Matlab

Stewart Heitman (Victor Chang Cardiac Research Institute)
Click here for abstract
The Brain Dynamics Toolbox is an open-source toolbox for researchers, engineers and students who wish to simulate dynamical systems in neuroscience. It allows modellers to build interactive Matlab simulations for any custom set of differential equations with minimal programming effort -- often less than 100 lines of code. The graphical interface fosters intuitive exploration of the dynamics for research and teaching purposes. The accompanying command-line tools allow large-scale simulations to be run in batch mode. The toolbox supports a dozen solver routines and has a growing list of specialised display panels. The solvers span the major classes of differential equations that typically arise in computational neuroscience. Specifically, Ordinary Differential Equations (ODEs), Delay Differential Equations (DDEs) and Stochastic Differential Equations (SDEs). Each of these can further be extended to Partial Differential Equations (PDEs) using the method of lines. The software is extensively documented with a 100 page Handbook and online training courses are provided too (

Open source and deployable BCI

Johan van der Meer (QIMR Berghofer Medical Research Institute)
Click here for abstract
A brain–computer interfaces (BCI) is a direct connections between the brain and and a computer. BCIs can enable people who are severely paralyzed to communicate their wishes, write using word processing programs, or even control a neuroprosthesis. In addition, using a Neurofeedback Training protocol, people can learn how to control aspect of their own brain activity. This can be used either to enhance performance in sports, or for treating mental health disorders such as attention deficit hyperactivity disorder (ADHD), Depression, Fatigue or post-traumatic stress disorder (PTSD). Elon Musk’s Neuralink company even aims to develop a direct ‘Brain-to-Brain’ communication technology ( BCI research is an interesting craft that requires troubleshooting on the level of engineering, mathematics and experimentation. furthermore, it holds a great future promise in facilitating human function. It is not easy to set up a BCI system yourself. You need an electroencephalography (EEG) – a combination of scalp electrodes and an amplifier (which is expensive) and specialized software (which can be obscure without programming experience, or expensive when in need of a software license). Due to these hurdles, adoption of BCI systems by enthusiasts or university curriculum is limited. To explore a BCI is usually the prerogative of specialized labs or companies aiming to market their device. It would be great if BCI is more easy to set up. To this end, the hardware and software side need to be more easily accessible. On the hardware side, the initiative of openBCI ( allows the acquisition of cost-effective EEG hardware, with open source software that can easily interface with any analysis platform. On the software side, new open-source python packages such as nfb ( and pycnbi ( and c-libraries such as bci2000 ( and openvibe ( become available allowing for the interface with amplifiers and real-time analysis. These hardware and software packages need to be combined to allow a BCI system to be easily deployable across different labs. In addition, someone new to BCI needs not to have to re-invent the wheel, and be provided a a starting point (with examples) from which one could continue. In the brain twitter conference I would showcase some of my own work on a BCI system that uses these open-source libraries. The goal of the system is be as easily deployable as possible across multiple sites or set up by anyone who has an interest in BCI.

Concurrent brain waking modes: predicting cortical hyperarousal

Paula Sanz-Leon (QIMR Berghofer )
Click here for abstract
In the space of six tweets I will give you the TL;DR version of my recent analytic work in which I have predicted the existence of concurrent brain waking states. These two coexistent states are termed the low- and high-waking modes.

How did I find these states? I performed a numerical analysis of the steady-state solutions of a neural field model of the thalamocortical system. A neural field model is a widely adopted type of mean-field model in computational neuroscience. This family of models enables the prediction of the ensemble activity of thousands of neurons, and of the whole brain. Really? The whole brain? Yes, the whole brain! You can ask me more about the model during the conference.

How are steady-state solutions related to brain states? The frequency of the fluctuations around a stable steady state (also known as a fixed point) of the model, is characteristic of a global brain state (e.g., resting with the eyes open, eyes closed, REM sleep, slow wave sleep). The first part of the analysis I did consisted of a systematic identification of multistable regions, for physiological parameter ranges representing normal arousal waking states in adult humans. In case you are wondering, parameter values for the model have been derived from human electroencephalographic recordings (EEG).

The key analytical results are the confirmation of the existence of up to three linearly stable steady states. Why is this important? Because the presence of multiple stable steady-state solutions implies that the model has enough degrees of freedom to capture multiple operating points. That is, it can capture multiple brain states that coexist. And in simple English terms? Two out of the three steady states, which are the low- and high-waking modes, are physiologically plausible in terms of their frequency characteristic. The low-waking mode has been previously identified with normal brain activity during quiet wakefulness and its signature frequency is the alpha rhythm (~10Hz). However, the high-waking mode had not been fully characterized. In this presentation I will show you (i) the frequency features of the high-waking mode; (ii) the nonlinear attractors between the two waking modes (yes, there is chaos involved); and, (iii) the switching dynamics between the low- and high-waking modes.

I argue that the high-waking mode may represent cortical hyperarousal waking states. Do you know that feeling in the evening, when you are trying to sleep but you brain will not stop? Yes, that one. That is the hyperaroused cortical state. For real world applications, these modelling results open up the possibility to predict and identify subtypes of primary insomnia in which cortical hyperarousal is the main hallmark from human neuroimaging data.

The possibility and reliability of the unattended memory retrieval under relaxation state

Kwon Nayeon (Seoul National University, CogSci)
Click here for abstract
In the field of forensic investigation, statements of witnesses or victims under the hypnotic state would be cues for deeper examinations. The points are whether the statements have reliability. For inspecting the reliability of memory retrieval under hypnotic state, I conducted a pilot study about retrieving unattended stimuli. I have the participants take word categorization tasks for 4000 ms and provided the masked picture under the word-tasks for 100 ms, which they couldn’t notice during the word categorization tasks. After 30 tasks in total, one group (experiment group) had hypnotic suggestion by the expert and the other group took a relax. After the suggestion or relaxation, they were asked to choose a picture out of three feeling the most familiar. The results showed the participants under the hypnotic(relaxation) suggestion tended to select the picture presented with masked during the word categorization tasks. The scores of word categorization tasked were not different between the two groups, which indicates both groups focused on the tasks. This results suggested that it would be valuable to investigate whether hypnotic state would affect the retrieval of unattended stimuli through further sophisticated methodologies such as fMRI or EEG devices.


Discussions can freely continue under the hashtag #brainTC.

Decoding EEG signals reveals weak auditory neuronal representations in tinnitus

Hyun Seok Kim (Institute for Brain and Cognitive Engineering, Korea University)
Co-authors: Dimitrios Pantazis (McGovern Institute for Brain Research, Massachusetts Institute of Technology), Byoung-Kyong Min (McGovern Institute for Brain Research, Massachusetts Institute of Technology; Department of Brain and Cognitive Engineering, Korea University)
Click here for abstract
Tinnitus is the conscious perception of internally generated noise or ringing in the ears in the absence of external sound. Although tinnitus is usually associated with maladaptive neuroplasticity resulting from deafferentation of the auditory nerve, the neurophysiological correlates underlying this symptom remain unclear. In this study, we investigated whether tinnitus patients could be distinguished from healthy controls based on event-related electroencephalography (EEG) signals using a support vector machine (SVM) decoding method. EEG data was recorded from 14 tinnitus patients and 14 healthy age/sex-matched volunteers. During EEG acquisition, participants performed two tasks: an auditory oddball task and a passive listening task. In the oddball task, two sound stimuli (standard: 500 Hz, target: 8 kHz for healthy participants and an individual tinnitus pitch-matched frequency for patients with tinnitus) were presented in random order, and participants were required to discriminate the rare stimulus (target) from the frequent one (standard) by pressing a button. In the passive task, the same stream of auditory stimuli was passively heard by the participants without requiring a response. In the oddball paradigm, we could decode target vs. standard stimuli in both the tinnitus and the control condition. However, the tinnitus group had decreased decoding results than the control group during 180-680 ms poststimulus, suggesting the tinnitus condition is associated with weakened auditory neuronal representations (tinnitus: 92.0% vs controls: 97.4 %, p < 0.05 by the cluster-based permutation test). Corroborating these results, when decoding oddball vs. passive task from the EEG responses to the standard stimuli, we again observed weaker decoding results for the tinnitus group (maximum 86.0%) than the control group (maximum 92.9%) roughly in the same time window 180-610 ms poststimulus as before. Together, these observations could provide a potential neurophysiological biomarker with increased diagnostic, and maybe prognostic, accuracy of tinnitus symptoms when combined with other established clinical procedures.

Functional analysis with BrainQ

Qin He (Tampere University)
Co-authors: Timo Hämäläinen, Sampsa Pursiainen (Tampere University)
Click here for abstract
Goal: Signal processing and machine learning is commonly applied in the cross disciplinary study of neuro science and cognitive system. Combine signal processing and machine leaning, which derives from cognitive control, to study the EEG and MEG signal of finger stimulation, realizing the source localization prediction evaluation as well as the functional connectivity and coherence analysis. Methods: Cognitive control is usually applied in information processing and behavioral response. In the preprocessing, baseline correction is plemented to analyze the pre-stimuli, combining ERP to mark the event related potential, studying the time-locked only behavior. Z-score transform, coherence and spectrum are calculated and analyzed in the functional connectivity analysis. In addition to the functional analysis, Bayes Optimizer evaluates the neuro imaging according to the hierarchical Bayes. Results: The coherence on time domain recorded by EEG and MEG is studied according to the ERP corrected baseline stimuli, which shows significantly the time-locked behavior.When it comes to neuro imaging, the basic calculation related to forward simulation and inverse dipole reconstructions are conducted in the parent application 'zeffiro' with the synthetic data exploring the single dipole source and paired dipole source in thalamus and somatosensory. The theta0 prediction works better with SVM. Brain Q is suitable for preprocessing for the EEG and MEG data, being capable of functional analysis including coherence and spectral related computation. Machine learning techniques are conducted as well to analyze the result of the dipole reconstruction and help to predict the better model parameters and the localization of the origin dipoles.

A stochastic approximation expectation-maximization algorithm for joint estimation of source and connectivity from MEG data

Narayan Puthanmadam Subramaniyam (Dept. of Neuroscience and Biomedical Engineering, Aalto University)
Co-authors: Filip Tronarp and Simo Särkkä (Dept. of Electrical Engineering and Automation, Aalto University), Lauri Parkkonen (Dept. of Neuroscience and Biomedical Engineering, Aalto University)
Click here for abstract
Current methods to estimate connectivity from magnetoencephalographic (MEG) / electroencephalographic (EEG) methods use a two-step approach; first the MEG/EEG inverse problem is solved to estimate the source activity, and second, connectivity is estimated between the sources. This suffers from the limitation that the estimation of the source activity is not informed by the connectivity structure. Also, most of the two-step approaches need regions-of-interest (ROIs) to be specified a priori for connectivity estimation. In this work, we present a novel algorithm to jointly estimate the source locations, their activity and directed functional connectivity between them using Bayesian filtering within a stochastic approximation expectation-maximization framework. Our results from simulations show that the proposed method outperforms the traditional two-step approach. When applied to real MEG data, our approach provides physiologically plausible estimates for the connectivity network underlying face perception in humans.

Does the attention processes affect children with attentional difficulties and reading difficulties?

Praghajieeth Raajhen Santhana Gopalan (University of Jyväskylä)
Co-authors: Otto Loberg, Kaisa Lohvansuu, Jarmo Arvid Hämäläinen and Paavo H. T. Leppänen (University of Jyväskylä, Department of Psychology)
Click here for abstract
Background: Visual attention-related processes include three functional sub-components: alerting, orienting, and inhibition. Previous studies have proposed that selective visual attention in the normal adult population is a complex cognitive process involving both ventral and dorsal visual networks, posterior parietal cortex and prefrontal cortex that mediates top-down attentional systems. Here we examine these components using brain event-related potentials and their neuronal source activations during the Attention Network Test (ANT) in children with attentional difficulties (AD) and reading difficulties (RD) to localize these processes.

Methods: During the ANT test, EEG was measured with 128 electrodes and combined with simultaneous eye-tracking and reaction time data, from three groups of Finnish sixth-graders aged 12-13 years (control, N = 83; AD, N = 15; RD, N = 23). In the ANT test, participants were asked to detect the direction of a middle target fish out of a group of five fish. The target stimulus was either preceded by a cue (centre, double, or spatial) or without a cue, in order to manipulate the alerting and orienting sub-processes of attention. The direction of the target fish can either be congruent or incongruent in relation to the flanker fish, thereby manipulating the inhibition sub-processes of attention. Behaviorally, the attention network effects were studied using reaction time (RT) differences between the stimulus conditions. The neuronal source activations were examined using pre-established spatial filter model.

Results: Reaction time performance of AD group showed reduced orienting effects (centre cue - spatial cue) compared to other groups. No differences were found between the three groups on RT of alerting and inhibition effects. No differences were found between the groups at the sensor level for the ERPs. Neuronal source analysis revealed significant group effects in the left occipital lobe associated with alerting network (double cue vs no cue; control > AD; AD < RD) and orienting network (spatial cue vs centre cue; control > AD; control > RD). In inhibition network (incongruent vs congruent target stimuli), medial frontal lobe and left medial temporal lobe showed the difference in control > AD, right medial temporal lobe showed the difference in control < RD and AD < RD.

Conclusion: The results provide evidence for the differences in orienting of attention in AD and RD children compared to the control children. This could be related to the limited ability to maintain the spatial attentional focus. Our data indicate the functional differences in the occipital, medial temporal and medial frontal lobes in AD and RD groups.

What determines the aesthetical appreciation of music

Alice Proverbio (“Neuro-Mi” Milan Center for Neuroscience, University of Milano-Bicocca)
Besides personal taste, culture and musical expertise, some intrinsic harmonic or melodic properties can be identified in the architecture of a musical piece, able to interact with innate neurobiological structures of the brain in a predictive and quite universal manner: e.g., complexity, familiarity, melodic profile, harmonic content, dissonance.Neuroimaging and psychophysiological studies have shown the neural correlates of music emotional sensations (e.g., joy, nostalgia, tension), but how does the brain really extract the emotional content of music? Why major chords are «bright» and minor chords are «dark»? Why atonal contemporary music is not appreciated by the average public, whereas pop music it is? We know how the brain reacts to dissonant vs consonant intervals. Perceptual dissonance has been ascribed to the fact that dissonant chords contain frequency components that are too closely spaced to be resolved by the cochlea. Two harmonics close in frequency (e.g. a second minor interval) shift in and out of phase over time, producing an interaction that oscillates, so that the amplitude of the combined physical waveform alternately waxes and wanes. These amplitude modulations are called ‘beats’ and result in an unpleasant sensation defined as ‘roughness’. If the partials are close enough they excite the same set of auditory fibers: the neural signal is noisy and the information delivered is not sufficiently clear for frequency recognition. On the other hand, consonant chords (which share superior harmonics) combine to produce an aggregate spectrum that is typically harmonic, resembling the spectrum of a single sound that is recognized as a unitary object by the auditory cortex. This leads to the positive sensation linked to listening to harmonic vs. disharmonic chords. But besides this specific effect, how does the brain recognize emotions in music? How can a sad or happy music be universally recognized as such regardless of style? How can a fearful music really be able to scare us, even if we are sitting on the couch and know that there is no danger in reality? How can film music make the various movie characters more or less lovable? These issues are studied by Neuroaesthetics. We will show how the brain is able to extract and comprehend emotional cues by means of specialized neural populations within the middle and superior temporal cortex devoted to understanding the prosodic and affective content of human vocalizations and speech.


Discussions can freely continue under the hashtag #brainTC.

Neurocenter Finland - promoting neuroscience and innovation

NEUROSCIENCE MAKING AN IMPACT: Mikael von und zu Frauenberg (Neurocenter Finland)
Co-authors: Merja Jaronen, Emmi Reijula and Antti Kotimaa (Neurocenter Finland)
Click here for abstract
Neurocenter Finland is a national network, which brings together clinical and preclinical researchers, health sector companies and financiers. In the future, the center is aiming to operate as a one-stop-shop creating an internationally competitive operational environment to attract partners for collaborative and innovative projects as well as clinical studies. The center also seeks to advance the growth and exports of companies working within the area, and finally, to increase the number of jobs and investments in Finland.

The European Brain Council: towards sustained and better coordinated brain research in Europe

NEUROSCIENCE MAKING AN IMPACT: Frédéric Destrebecq (Executive Director of EBC)
Click here for abstract
The European Brain Council (EBC) serves as a pan-European platform bringing together brain researchers, patient advocates, scientists and clinicians around a common vision for the ‘brain space’. Accelerating brain research, raising awareness of the challenges associated with brain disorders, reducing stigma and discrimination and supporting science are at the heart of our advocacy work. Our mission revolves around promoting brain research and brain health at the European level, to the benefit of people living with neurological and psychiatric disorders.

Brain disorders are highly prevalent. In 2010, it was estimated that more than 1 in 3 European citizens were affected and that brain disorders alone amount to ca. 45 % of the annual health budget in Europe, totalling around €800 billion every year.

A number of scientific research projects and initiatives are true showcases of the work of EBC:

The European Brain Research Area (EBRA) was created as a catalysing platform for brain research stakeholders to streamline and better co-ordinate brain research across Europe while fostering global initiatives. EBRA is coordinated by EBC, partnered with the Joint Programming on Neurodegenerative Diseases, the Human Brain Project and the ERANET Neuron.

The EBC conducted a study on the Value of Treatment for brain disorders in Europe, designed to address the existing ‘treatment gaps’ in Europe. It is estimated that out of 10 people living with a brain condition, from 3 to 8 do not receive adequate treatment even when it is available and the burden of diseases is mainly linked to inadequate care or a lack of access to care.

The EU-funded MULTI-ACT project aims to increase the impact of health research on people with brain diseases. It will create and implement a new model to engage stakeholders in defining metrics for a given mission and agenda. MULTI-ACT is focused on Brain Diseases Research Agenda and uses Multiple Sclerosis as the first case study and foresees patients as a key stakeholder in the Responsible Research Innovation process.

With the launch of the EBC “Brain Mission”, EBC called for increased recognition of the burden of brain disorders and the continued need for funding of brain research. In light of “Horizon Europe” and the intent to “adopt a mission-oriented impact focused approach to address global challenges”, it is on that basis that EBC proposed policy makers to endorse its proposed brain mission.

Lastly, EBC has launched its Election Manifesto in light of the upcoming European Elections 2019. The document calls on incoming Members of the European Parliament to guide the EU to the role of global leader in brain research, sustain EU funding to expand and boost brain research, improve the quality of care for patients across the EU, recognize brain health as a priority for European society and encourage awareness, early preventative measures and early intervention to address the growing burden of brain conditions.

Educating a new generation of scientists: An interdisciplinary approach

NEUROSCIENCE MAKING AN IMPACT: HPB Education Programme (Medical University Innsbruck)
Click here for abstract
As one of the largest research projects ever funded by the European Union with a planned duration of ten years, the Human Brain Project (HBP) unites more than 500 scientists at universities, teaching hospitals and research institutes all over Europe in the pursuit of a unique goal: To connect knowledge and insights from computer science, brain-related medicine and neuroscience to approach a better understanding of the complexity of the brain, its inherent processes, and the diseases that put its functionality at risk.

Six prototype platforms have been developed to form the basis of the world’s first integrated infrastructure for academic and industrial brain research and development. This joint effort combines advanced neuroinformatics software and high-performance computing resources that support analytics, modelling and high-end simulations at all levels of brain organisation in an innovative and unprecedented way.

To support the novel transdisciplinary approach and distribute information about the research conducted and the tools developed in the framework of this EU Flagship Project, a dedicated education programme has been established. The HBP Education Programme comprises various formats targeted at researchers in the early stages of their career who work in neuroscience, medicine and computation as well as in related fields, such as biology, psychology or robotics. Courses and events range from basic lessons in neuroscience, ICT and medicine, over workshops on complementary topics such as research ethics or innovation, interdisciplinary student conferences and specialised training on specific tools, to advanced summer schools.

The teaching and training strategy of the HBP Education Programme has been tailored to the extraordinary needs of a research environment with an exceptional scale both in terms of headcount and variety of scientific backgrounds and approaches. Its main objective is to provide students and young researchers with a skillset that enables them to think outside the box, communicate about their work and results in a comprehensible manner also outside their subject area, collaborate across disciplines, and to use the different tools and services developed in the HBP for their own research. Guided by pioneers of the fields currently involved, these young scientists mature together with the infrastructure and will emerge as experts when the project is finalised.

Per definition, the field of neuroscience is one of the most interdisciplinary scientific fields. In a time when technology advances at an increasing pace, the mutual exchange of different perspectives plays an important role for the emergence of new ideas and paves the way for scientific breakthroughs. Through its innovative and interdisciplinary training formats, the HBP Education Programme contributes to the education of a new generation of scientists that rise to the challenges posed by this ever-changing environment.

Effects of spatial smoothing on group-level differences in functional brain networks

Ana Triana (Aalto University, Department of Computer Science)
Co-authors: Enrico Glerean (Aalto University, Department of Neuroscience and Biomedical Engineering), Jari Saramäki (Aalto University, Department of Computer Science), Onerva Korhonen (Université de Lille, CNRS, UMR 9193 - SCALab)
Click here for abstract
The human brain is comprised of spatially separated groups of specialized neurons. Brain function is based on the interactions between these neuronal groups that produce complex behaviors. Such interactions may vary in different populations, for example between healthy and clinical groups. Therefore, investigating group-level differences in brain network structure has become increasingly popular, leveraging on different neuroimaging techniques, such as functional Magnetic Resonance Imaging (fMRI). Due to the nature of the fMRI signal, several preprocessing steps must be applied to the time-series to control for artifacts – such as head motion and physiological noise – before constructing the functional networks. These steps aim to increase the signal-to-noise ratio and improve data quality. However, the preprocessing choices may have undesired effects on the results. For example, spatial smoothing induces changes in functional network structure of individuals. Yet, its effects on group-level differences remain unknown.

Here, we investigate the effects of spatial smoothing on group-level differences between Autism Spectrum Disorder (ASD) patients and Typical Controls (TC). We compare resting-state fMRI connectivity patterns of N=33 age-matched subject pairs from the ABIDE initiative. We consider spatially unsmoothed data as well as filtered data with 8 different Gaussian kernels, spanning from 4mm to 18mm full-width at half-maximum (FWHM). We compare the network structure using the Network Based Statistic Toolbox, which identifies networks comprising the connectome associated with a between-group difference. We find that spatial smoothing affects network differences. Especially, the network density increments with increasing kernel size (F-statistic>16). Moreover, the network nodes change at different levels of smoothing and few links are permanently detected across all kernels. These few links are found between the sensorimotor and subcortical areas such as thalamus and striatum, which are commonly reported in the literature related to ASD. These effects are present independently of the ROI size or physical link length.

Due to the lack of consensus on the differences between ASD and TC brain connectivity patterns, it is difficult to say whether spatial smoothing is highlighting true differences between the networks due to a data quality improvement, or if its effects are distorting non-existing differences as significant findings. In general, the effects of spatial smoothing are non-trivial and difficult to predict. Hence, spatial smoothing should be considered carefully, as it alters network differences when comparing functional brain networks of different groups.

Sub-systems of phase-synchronisation from human intracerebral recordings comprise functionally related, spatially contiguous regions

Nitin Williams (Neuroscience Center, Helsinki Institute of Life Science, University of Helsinki)
Co-authors: Gabriele Arnulfo (Neuroscience Center, Helsinki Institute of Life Science, University of Helsinki; Dept. of Informatics, Bioengineering, Robotics and System Engineering, University of Genoa), Sheng H. Wang (Neuroscience Center, Helsinki Institute of Life Science, University of Helsinki; Doctoral Programme Brain & Mind, University of Helsinki), Lino Nobili (Claudio Munari Epilepsy Surgery Centre, Niguarda Hospital; Child Neuropsychiatry, IRCCS, Gaslini Institute, DINOGMI, University of Genoa), Satu Palva (Neuroscience Center, Helsinki Institute of Life Science, University of Helsinki; BioMag laboratory, HUS Medical Imaging Center; Center for Cognitive Neuroimaging, Institute of Neuroscience and Psychology, University of Glasgow), J. Matias Palva (Neuroscience Center, Helsinki Institute of Life Science, University of Helsinki; Center for Cognitive Neuroimaging, Institute of Neuroscience and Psychology, University of Glasgow)
Click here for abstract
The brain in resting-state exhibits inter-regional correlations between amplitude envelopes of neuronal oscillations as well as synchronisation between oscillation phases. Connectomes are the set of correlations between every pair of brain regions and sub-systems are sets of strongly inter-correlated regions. Sub-systems in the connectome of amplitude envelope correlations have been identified (so-called Resting State Networks), but little is known about sub-systems in the connectome of phase-synchronisation. In this study, we identified sub-systems in connectomes of phase-synchronisation at 18 frequencies from 3 to 320 Hz, using stereo-EEG (SEEG) recordings from 64 subjects. After generating connectomes of phase-synchronisation from white-matter referenced measures of LFP (Local Field Potential) activity, we identified sub-systems using Louvain community detection, combined with consensus clustering to counteract sparse SEEG coverage. First, we found evidence for sub-systems of phase-synchronisation at multiple spatial scales. Next, we found that sub-systems of phase-synchronisation are grouped into sets of adjacent frequencies, where the sets correspond to frequency bands reported in the literature. The specific bands were 3-4 Hz (delta), 5-10 Hz (theta/alpha), 12-20 Hz (beta), 28-80 Hz (gamma), 113-135 Hz (high gamma) and 160-320 Hz. Finally, we found that sub-systems of phase-synchronisation up to 80 Hz comprised spatially contiguous regions that are known to be functionally related. The sub-systems respectively comprised occipital areas, sensorimotor areas, posterior parietal areas, superior temporal areas, inferior temporal areas, superior frontal areas and inferior frontal areas. These findings reveal sets of regions that interact in resting-state, over and above those identified by other neuroimaging technologies.

Genome-wide association study of brain connectivity changes for Alzheimer’s disease

Samar Salah Mohamedahmed Elsheikh (University of Cape Town)
Co-authors: Emile R. Chimusa (Division of Human Genetics, Department of Pathology, Institute of Infectious Disease and Molecular Medicine, Faculty of Health Sciences, University of Cape Town), Nicola J. Mulder (Computational Biology Group, Department of Integrative Biomedical Sciences, Institute of Infectious Disease and Molecular Medicine, Faculty of Health Sciences, University of Cape Town), Alessandro Crimi (University Hospital of Zurich; African Institute for Mathematical Sciences) and Alzheimer's Disease Neuroimaging Initiative (Data used in preparation of this article were obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database ( As such, the investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in the analysis or writing of this report. A complete listing of ADNI investigators can be found at:\url{
Click here for abstract
Variations in the human genome have been found in the literature to be an essential factor that affects the susceptibility of Alzheimer’s disease (AD). Genome-wide association studies (GWAS) have identified genetic loci that significantly contribute to the risk of Alzheimer's disease. The availability of the genetic data, coupled with brain images technologies have left rooms for more discovery, data integration methodologies and allowed further study designs. Although methods have been proposed towards integrating images characteristics and genetic information; for such kind of studies, the measurement of disease is often taken at a single time point, therefore, not allowing the disease progression to be taken into consideration. In longitudinal settings, we analyzed neuroimaging and single nucleotide polymorphisms datasets obtained from Alzheimer’s Disease Neuroimaging Initiative for three groups of participants at different clinical stages, including controls, early mild cognitive impairment and AD patients. We conducted a GWAS regressing the absolute change of specific global network metrics on the genetic variants. Moreover, we used the GWAS summary statistics to compute the gene scores, and we observed significant associations between the change in the brain network metrics and genes previously reported to manipulate certain brain disorders, AD, as well as the progression of some cancers.

Automatic multi--dipole estimation from M/EEG data in the time or frequency domain

Sara Sommariva (Department of Neuroscience and Biomedical Engineering, Aalto University)
Co-authors: Gianvittorio Luria (Istituto Neurologico Carlo Besta, Milano), Alberto Sorrentino (Dipartimento di Matematica, Università di Genova and CNR-Spin)
Click here for abstract
Multi-dipole modeling of magneto/electro-encephalografic (M/EEG) data is an effective method for characterizing the activity of the brain, under the hypothesis of the active sources being focal. This happens, e.g, in non-invasive localization of epileptic spikes [1].

Estimating dipolar sources from M/EEG data is a complex and time consuming task, that usually requires expert initialization, subjective decisions about the number of active sources, and careful double-check of the results. This is due mostly to the non-linearity of the problem, that generates local minima in which optimization algorithms tend to get trapped, leading to implausible solutions that need to be discarded.

Here we present a Bayesian approach [2-3] to this problem for automated estimation of the number of dipoles and their locations, under the assumption that they are static over a set of M/EEG topographies; the method also estimates the corresponding dipole moments, which are instead allowed to vary. The final result of the algorithm is an approximation of the posterior distribution of all the unknowns (number, locations and moments of the dipoles), which is computed through a Sequential Monte Carlo sampler [5]. The posterior distribution can be used to obtain point estimates as well as to quantify the uncertainty of such estimates.

In [2,3,4], we used both simulated and experimental data (somatosensory stimulation and visual go-no-go task) to exhibit the reliability of the proposed approach when the input consists of either a set of time points of the recorded data [2-3], or a set of frequencies in which the Fourier transform of the data is evaluated [4]. We showed that the method is able to estimate dipolar configurations comprising up to 4 sources with a location error in average below 2 mm even when the sources have perfectly correlated time-courses [2-3]; in [4] we applied the method in the frequency domain, and showed that the method localizes correctly not only dipolar sources, but also moderately extended active areas.

The algorithm is currently part of the BESA research software, under the name SESAME ( and a free, open-source python package is under development (alpha version at

[1] Merlet and Gotman, Clinical Neurophysiology, 1999 [2] Sorrentino et al., Inverse Problems, 2014 [3] Sommariva and Sorrentino, Inverse Problems, 2014 [4] Luria et al. Journal of Neuroscience Methods, 2019 [5] Del Moral et al. Journal of the Royal Statistical Society B, 2006

InVesalius Navigator, a free neuronavigation software for transcranial magnetic stimulation

Victor Souza (Aalto University)
Co-authors: Renan Matsuda and Oswaldo Baffa (Universidade de São Paulo), André Peres (Instituto Santos Dumont), Thiago Moraes, Paulo Amorim and Jorge Silva (Centro de Tecnologia da Informação Renato Archer)
Click here for abstract
We developed an open-source, free neuronavigation software with tools for transcranial magnetic stimulation (TMS) experiments. Neuronavigation is a valuable tool in clinical and research environment. Specifically for TMS, neuronavigation improve the reliability of physiological outcomes and accuracy in targeting cortical structures. However, commercial systems are expensive, not compatible with multiple tracking devices, and do not provide an easy-to-implement platform for custom tools. To overcome such limitations, we developed the InVesalius Navigator for navigated TMS ( A guiding interface were designed for tracking any TMS coil relative to an individual’s anatomy. The InVesalius Navigator provides communication with multiple tracking devices, tools for structural image processing, and online TMS coil tracking, all combined in a user-friendly interface. Localization, precision errors, and repeatability were measured for the spatial trackers Patriot (Polhemus Inc.) and MicronTracker Sx60 (ClaroNav Inc.). Errors were measured and compared to the commercial navigated systems NBS 3.2 and 4.3 (Nexstim Plc.). InVesalius Navigator provided a localization error of about 1.5 mm, and repeatability of about 1 mm for translation and 1° for rotation angles. Our results are within the limits established in the literature and similar to those achieved with the NBS systems. The developed software provides a flexible platform aiming to fulfill the needs of research and clinical requirements, expanding the use of navigated TMS throughout the community. Finally, InVesalius Navigator might improve the reliability and spatial accuracy of non-invasive brain stimulation measurements.


Discussions can freely continue under the hashtag #brainTC.

Cathodal tDCS impacts the modulatory effects of learning in fear extinction: a dynamic causal modelling study.

Raquel Guiomar (Cognition Brain and Behavior, CINEICC, Faculty of Psychology and Educational Sciences, University of Coimbra)
Co-authors: Ana Ganho-Ávila (Cognition Brain and Behavior, CINEICC, Faculty of Psychology and Educational Sciences, University of Coimbra)
Click here for abstract
Transcranial direct current stimulation (tDCS) has been studied as an add-on enhancer of extinction-based treatments for anxiety disorders. Previous studies have showed that although anodal tDCS enhances fear extinction it also leads to decreased stimuli discrimination and consequent generalization of the fear response. Here we aim to 1) understand the impact of cathodal tDCS in neural activity and connectivity patterns between the tDCS target site (the right dorsolateral prefrontal cortex; rDLPFC) and the extant fear network; 2) see if cathodal stimulation also increases generalization.

In a fear conditioning procedure (day 1 – fear acquisition; day 2 – 20-min 1 mA cathodal tDCS (for tDCS group) + fear extinction, thirty-four female participants were randomly assigned to either the tDCS or control group. All participants showed fear acquisition in day 1.

Group-level psychophysiological interaction analysis (PPI) showed that cathodal tDCS effectively interferes with the fear response neural network. To further explore the PPI results, we selected the significant regions (the vmPFC, the bilateral DLPFC, and right amygdala) and performed an effective connectivity analysis using DCM (Dynamic Causal Modelling). We separately entered the CS+ and the CS- as driving inputs, and the early and late phases of extinction session as modulatory inputs. To understand the effects of cathodal tDCS on the network’s connectivity, we performed independent samples t-tests on the connectivity parameters. The results suggested that in the early phase, when the CS+ was the driving input, the connectivity between the rDLPFC and the vmPFC was reduced for the tDCS group (p < .05, uncorr.). Thus, cathodal tDCS leads to an inhibitory effect of the rDLPFC over the vmPFC activity. Moreover, when the CS- was the driving input, the connectivity between the vmPFC and the right amygdala was enhanced for the tDCS group (p < .005, Bonferroni corr.). Hence, cathodal tDCS leads to an excitatory effect of the vmPFC over the right amygdala, impacting the processing of neutral cues through an augmented activity of the amygdala and consequent increased plasticity. This may be the mechanism of action by which cathodal tDCS boosts stimuli discrimination as we have seen in our previous studies. In fact, previously we have seen reduced implicit avoidance tendencies towards the CS- after cathodal tDCS, suggesting that an enhanced distinctiveness between threatening cues and perceptively similar neutral cues.

Our results show that 1mA cathodal tDCS effectively interferes with the cortico-subcortical fear network during extinction, suggesting that it can be a potential add-on strategy to enhance extinction-based treatments. Future studies should test the potential therapeutic effect of cathodal tDCS in increasing CSs discrimination in clinical samples defined by intolerance to uncertainty, such as are patients diagnosed with generalized anxiety disorder.

Experiencing the self through touch in healthy participants and ADHD patients

Rebecca Boehme (Center for Social and Affective Neuroscience, Linköping University)
Co-authors: Andrea Johansson Capusan (Center for Social and Affective Neuroscience, Linköping University; Adult psychiatry, Linköping University Hospital), Håkan Olausson (Center for Social and Affective Neuroscience, Linköping University; Department of Clinical Neurophysiology, Linköping University Hospital)
Click here for abstract
Differentiation between self-produced tactile stimuli and touch by others is necessary for social interactions and for a coherent concept of “self”. We have recently shown robust self-other-distinction in brain areas related to somatosensory, social cognitive, and interoceptive processing (Boehme et al., PNAS, 2019). Attention-deficit-hyperactivity-disorder (ADHD) is associated with sensory processing problems and social cognition impairments (Ghanizade 2011, Uekerman et al., 2010). Here, we compare the neural signatures of affective self- and other-touch between adult ADHD patients and healthy controls. We hypothesized to find altered processing in regions of interest related to affective touch (S1, Insula) and self-other-cognition (cortical-midline-structures). 28 AHDH patients and 30 age- and gender-matched healthy controls (HC) performed a self-other-touch-task during functional imaging: they stroked their own arm, an object or were stroked by the experimenter. In ADHD patients, we found enhanced responses to both touch conditions in the posterior insula, i.e. stronger deactivation during self-touch and increased activation during other-touch. Furthermore, we found increased activation in response to other-touch in S1 and a stronger deactivation during self-touch in the insula. We conclude that ADHD-patients show altered processing of social tactile stimuli, which could relate to deficits in social cognition. The more pronounced differentiation between self- and other-touch might indicate a clearer self-other-distinction, which could relate to lower empathy and tactile hypersensitivity.

Inter-brain synchrony during parent-child interaction and its potential link to attachment

Pascal Vrticka (Max Planck Institute for Human Cognitive and Brain Sciences)
Co-authors: Jonas G. Miller (Center for Interdisciplinary Brain Sciences Research, Stanford University; sharing the first authorship with P. Vrtička), Xu Cui, Sharon Shrestha, S.M. Hadi Hosseini and Joseph M. Baker (Center for Interdisciplinary Brain Sciences Research, Stanford University), Allan L. Reiss (Center for Interdisciplinary Brain Sciences Research, Stanford University; Department of Radiology, School of Medicine, Stanford University)
Click here for abstract
Coordinated brain activity between individuals, or inter-brain synchrony, has been shown to increase during cooperation and correlate with cooperation success. However, few studies have examined parent-child inter-brain synchrony and whether it is associated with meaningful aspects of the parent-child relationship. Here, we measured inter-brain synchrony in the right prefrontal (PFC) and temporal cortices in mother-child dyads while they engaged in a cooperative and independent task. We tested whether inter-brain synchrony in mother-child dyads (i) increases during cooperation, (ii) differs in mother-son versus mother-daughter dyads, and (iii) is related to cooperation performance and the attachment relationship. Overall inter-brain synchrony in the right hemisphere, and the right dorsolateral and frontopolar PFC in particular, was higher during cooperation. Mother-son dyads showed less inter-brain synchrony during the independent task as compared to cooperation than mother-daughter dyads. Lastly, we found preliminary evidence for potential links between inter-brain synchrony and child attachment. Mother-child cooperation may increase overall inter-brain synchrony, although differently for mother-son versus mother-daughter dyads. More research is needed to better understand the potential role of overall inter-brain synchrony in mother-child cooperation, and the possible link between inter-brain synchrony and attachment.

Polygenic architecture of human neuroanatomical diversity

Nicolas Traut (Institut Pasteur)
Co-authors: Anne Biton (Institut Pasteur), Jean-Baptiste Poline (Montreal Neurological Institute and Hospital, McGill University; Henry H. Wheeler, Jr. Brain Imaging Center, Helen Wills Neuroscience Institute, University of California), Thomas Bourgeron and Roberto Toro (Institut Pasteur)
Click here for abstract
We analysed the genomic architecture of neuroanatomical diversity using magnetic resonance imaging and SNP data from > 20,000 individuals. Our results replicate previous findings of a strong polygenic architecture of neuroanatomical diversity. SNPs captured from 40% to 54% of the variance in the volume of different brain regions. We observed a large correlation between chromosome length and the amount of phenotypic variance captured, r~0.64 on average, suggesting that at a global scale causal variants are homogeneously distributed across the genome. At a more local scale, SNPs within genes (~51%) captured ~1.5-times more genetic variance than the rest. We also observed extensive pleiotropy across regions, with an average genetic correlation of rG~0.45. Across regions, genetic correlations were in general similar to phenotypic correlations.

Evolution of neocortical folding: A phylogenetic comparative analysis of MRI from 34 primate species

Katja Heuer (Max Planck Institute for Human Cognitive and Brain Sciences; Institut Pasteur)
Co-authors: Omer Faruk Gulban (Maastricht University Cognitive Neuroscience Department), Pierre-Louis Bazin (Max Planck Institute for Human Cognitive and Brain Sciences, Department of Neuropsychology; Spinoza Centre for Neuroimaging; Netherlands Institute for Neuroscience), Anastasia Osoianu (Groupe de Neuroanatomie appliquée et théorique, Unité de Génétique humaine et fonctions cognitives, Département de neuroscience, Institut Pasteur), Romain Valabregue (Institut du Cerveau et de la Moelle Épinière, CENIR, ICM, Sorbonne Universités, UPMC Univ Paris 06, Inserm U 1127, CNRS 7225), Mathieu Santin (Institut du Cerveau et de la Moelle Épinière, CENIR, ICM, Sorbonne Universités, UPMC Univ Paris 06, Inserm U 1127, CNRS 7225), Marc Herbin (Département Adaptations du Vivant, UMR MECADEV 7179 Équipe FUNEVOL Sorbonne Universités-MNHN-UPMC-CNRS-IRD, Muséum National d'Histoire Naturelle), Roberto Toro (Groupe de Neuroanatomie appliquée et théorique, Unité de Génétique humaine et fonctions cognitives, Département de neuroscience, Institut Pasteur)
Click here for abstract
We present a comparative analysis of cerebral size and neocortical folding. Magnetic resonance imaging data was collected from 65 individuals belonging to 34 different primate species. We measured several neocortical folding parameters and studied their evolution using phylogenetic comparative methods. Our results suggest that the most likely model is one where phenotypical differences vary randomly through evolution (the Brownian Motion model). We present estimations of the ancestral primate phenotypes as well as estimations of the rates of phenotypic change.

Potential of magnetoencephalography in the identification of functional brain network alternations in gliomas patients

Fatemeh Shekoohi-Shooli (Department of Neuroscience, Imaging and Clinical Sciences, “G. d’Annunzio” University of Chieti-Pescara)
Co-authors: Federico Chella (Department of Neuroscience, Imaging and Clinical Sciences and Institute for Advanced Biomedical Technologies (ITAB), “G. d’Annunzio” University of Chieti-Pescara; sharing first authorship with F. Shekoohi-Shooli), Massimo Caulo (Department of Neuroscience, Imaging and Clinical Sciences and Institute for Advanced Biomedical Technologies (ITAB), “G. d’Annunzio” University of Chieti-Pescara), Vincenzo Palatino (Department of Neuroscience, Imaging and Clinical Sciences, “G. d’Annunzio” University of Chieti-Pescara; Department of Diagnostic Imaging, University of Foggia, Ospedali Riuniti Hospital), Riccardo Navarra (Department of Neuroscience, Imaging and Clinical Sciences and Institute for Advanced Biomedical Technologies (ITAB), “G. d’Annunzio” University of Chieti-Pescara), Vittorio Pizzella (Department of Neuroscience, Imaging and Clinical Sciences and Institute for Advanced Biomedical Technologies (ITAB), “G. d’Annunzio” University of Chieti-Pescara), Laura Marzetti (Department of Neuroscience, Imaging and Clinical Sciences and Institute for Advanced Biomedical Technologies (ITAB), “G. d’Annunzio” University of Chieti-Pescara)
Click here for abstract

Gliomas consist of 29%–35% of the central nervous system tumors in adolescents and young adult and are divided in low-grade or fast-growing gliomas (LGGs) and high-grade or fast-growing gliomas (HGGs) [1]. LGGs and HGGs can disturb brain network architecture, contributing to abnormal neuronal activity that can be detected with magnetoencephalography (MEG) [2]. In this preliminary study, we evaluated abnormalities in the large–scale topology of resting state functional connectivity networks in gliomas patients by using MEG as well alterations of functional connectivity restricted to the tumor area.


Six gliomas patients (3F, 3M; mean age: 56 y) before treatment options were recruited for this study. MEG recordings of spontaneous cortical activity were obtained using a 153-channel whole-head MEG system installed in University of Chieti, and that data from two 5 minutes eye-closed resting state sessions were acquired. Data were beamformed onto a 4-mm spaced grid covering the whole brain by using a linearly constrained minimum variance (LCMV) beamformer [3]. Source-level functional connectivity between a set of regions of interest (ROIs) – i.e., tumorregion , peri-tumoralregion , control region, and their respective contralateral regions (Figure1: )-and the whole brain was carried out by using the imaginary part of coherency [4] and for frequencies ranging from 1 to 90 Hz; this resulted in a value for global functional connectivity (GFC) [2] for each ROI. Source activity at the same frequencies and for the same ROIs was analyzed in terms of the neural activity index (NAI) [3]. Comparisons between GFC and NAI values in the different ROIs were assessed by using t-test.


In the NAI in the upper alpha band, i.e. 11-13Hz range, we observed a significant difference (p<0.05) between the tumor and contralateral regions. Furthermore, there was a decrease of GFC in the tumor region in comparison to the contralateral region in the beta band, i.e. 13-30Hz range, although this decrease did not reach statistical significance for the small number of patients (p = 0.05).


Preliminary results on this small sample of patients reveal that NAI and GFC are altered in patients with gliomas in a frequency specific manner. In particular, NAI appears to be altered in the upper alpha frequency range while GFC features changes in the beta frequency range, suggesting a dissociation of the frequencies involved in local and long-range oscillations induced by the tumor. Moreover, it will be interesting to assess if and how these indices differ in additional patients with further differentiation with respect to tumor grade.


[1] Diwanji, T. (2017), Adolesc Health Med Ther, 2017(8):99-113. [2] Guggisberg, A.G. (2008), Ann Neurol, 63(2):193-203. [3] Van Veen, B.D. (1997), IEEE Trans Biomed Eng, 44(9):867-880. [4] Nolte, G. (2004), Clin Neurophysiol,115(10):2292-2307.


Discussions can freely continue under the hashtag #brainTC.

Power spectra predicts survival in comatose patients after cardiac arrest during the first day of coma

Thomas Kustermann (Laboratoire de Recherche en Neuroimagerie (LREN), University Hospital (CHUV) & University of Lausanne)
Co-authors: Nathalie Ata Nguepnjo Nguissi (Laboratoire de Recherche en Neuroimagerie (LREN), University Hospital (CHUV) & University of Lausanne), Christian Pfeiffer (Department of Psychology, University of Zürich), Matthias Hänggi (Department of Intensive Care Medicine, Inselspital, Bern University Hospital, University of Bern), Rebekka Kurmann and Frédéric Zubler (Department of Neurology, Inselspital, Bern University Hospital, University of Bern), Mauro Oddo and Andrea O. Rossetti (Department of Intensive Care Medicine, University Hospital (CHUV) & University of Lausanne), Marzia De Lucia (Department of Neurology, University Hospital (CHUV) & University of Lausanne)
Click here for abstract

Prediction of outcome in comatose patients following cardiac arrest remains challenging [1]. Prognostication is performed using a multimodal approach, including clinical examinations and interpretation of EEG patterns. Evaluations must be performed by expert physicians and remain subject to inter-rater reliability concerns. Quantification of EEG activity recorded during coma could serve to ameliorate prognosis [2]. We set out to investigate the putative prognostic value of comatose patients’ power spectra.


We recorded 8 to 20 minutes of 63-channel resting state EEG from post anoxic patients (n = 129; 80 survivors) under targeted temperature treatment during the first day of coma in four hospitals in Switzerland. For each patient, the spectral power from 1-40 Hz was computed and normalized by the total channel-wise power. Data were split into a training (n = 46) and test (n = 83) set to allow for independent validation of predictors. We first computed a statistical contrast of survivors and non-survivors in the training set and extracted the power spectral values from the resulting cluster. Based on these power spectral values we identified a threshold maximizing the positive predictive value (PPV) for good outcome in the training set. Using the same threshold and frequencies, we aimed to validate the prediction results in the test set.


A cluster-based permutation test [3] comparing survivors’ and non-survivors’ power spectra revealed significant (p = 0.02) differences across all electrodes. By extracting the average power spectral value along the cluster (5.2-13.2 Hz), threshold optimization for positive predictive power resulted in the correct identification of 13 survivors with no false positives, i.e. PPV: 1.00, 95% confidence interval (CI): 1.00-1.00. Applying the same threshold to the hold-out test set, we correctly identified 18 survivors with no false positives (PPV: 1.00; CI: 1.00-1.00). Across both datasets, we obtained negative predictive value: .50 (CI .40-.60), sensitivity: .39 (CI: .28-.49), specificity: 1.00 (CI: 1.00-1.00) and accuracy: .62 (CI: .54-.70)


Our analysis presents a viable and easy-to-implement marker for outcome prognostication in coma after cardiac arrest using portable EEG at bedside. The presence of prominent alpha peak in the power spectrum of survivors can be interpreted as mirroring preserved cortico-thalamic loops in patients with good clinical evolution.

1. Rossetti, A.O., A.A. Rabinstein, and M. Oddo, Neurological prognostication of outcome in patients in coma after cardiac arrest. Lancet Neurol, 2016. 15(6): p. 597-609. 2. Tzovara, A., et al., Prediction of awakening from hypothermic post anoxic coma based on auditory discrimination. Ann Neurol, 2016. 79: p. 748-757. 3. Maris, E. and R. Oostenveld, Nonparametric statistical testing of EEG- and MEG-data. J Neurosci Methods, 2007. 164(1): p. 177-90.

Cardio-audio synchronization induces neural surprise response in comatose patients

Marzia De Lucia (University Hospital (CHUV) & University of Lausanne)
Co-authors: Nathalie Ata Nguepnjo Nguissi and Thomas Kustermann (Laboratoire de Recherche en Neuroimagerie (LREN), University Hospital (CHUV) & University of Lausanne), Matthias Hänggi (Department of Intensive Care Medicine, Inselspital, Bern University Hospital, University of Bern), Frédéric Zubler and Rebekka Kurmann (Department of Neurology, Inselspital, Bern University Hospital, University of Bern), Christian Pfeiffer (Department of Psychology, University of Zürich)
Click here for abstract

Previous research on the neural processing of unexpected sensory signals or omissions within regular series has mainly used exteroceptive signals (i.e., from outside the body) presented to awake individuals. Recently, we introduced a new paradigm for studying the sensitivity of the human brain to statistical properties of auditory sequences temporally coupled to heartbeat signals (i.e., interoceptive, from inside the body). Using this paradigm in awake subjects, we observed a neural surprise response induced by omission of cardio-audio synchronous stimulation [1]. Here we investigated whether such cardio-audio surprise responses can also evoked in the absence of consciousness.


High-density electroencephalography (EEG) was recorded in a group of seventeen post-anoxic comatose patients during the first day after cardiac arrest while delivering sound sequences that were either occurring at fixed time-lags from the ongoing heartbeat (synchronous condition) or pseudorandomly (asynchronous condition), while matching the sequence statistical properties. In addition, we considered a condition where we presented sounds occurring at fixed and individually adjusted pace following mean heartbeat (isochronous condition).


A cluster permutation analysis comparing the omission response during the synchronous vs the asynchronous condition revealed a significant differential EEG response similar to previous observations in healthy subjects [1]. In the same patients we found no evidence of omission responses during the isochronous sequence.


Temporal regularity across heartbeats and sounds enabled prediction of sound onsets and thus detection of sound omission in comatose patients. As in isochronous auditory regularity we observed no evidence of omission detection, these results suggest that in unconscious individuals external stimuli are preferentially processed with respect to interoceptive as compared to exteroceptive temporal regularity.

1. Pfeiffer, C. and M. De Lucia, Cardio-audio synchronization drives neural surprise response. Sci Rep, 2017. 7(1): p. 14842.

Simultaneous electroencephalography and transcranial electric stimulation using innovative electrode materials integrated in a flexible cap

Alexander Hunold (Technische Universität Ilmenau)
Co-authors: Sophia Wunder (neuroConn GmbH), Patrique Fiedler (Technische Universität Ilmenau), Klaus Schellhorn (neuroConn GmbH), Jens Haueisen (Technische Universität Ilmenau)
Click here for abstract
The neuro-modulatory effects of transcranial electric stimulation (TES) by means of tDCS, tACS or tRNS indicate promising potential in neuroscience and clinical practice. However, the underlying mechanisms are not yet fully understood. Simultaneous application of TES and electroencephalography (EEG) as intervention and diagnostic neurophysiological modalities open significant room to investigate the neuro-modulatory mechanisms. Conventional electrode materials, i.e. rubber electrodes for TES and wet EEG electrodes, introduce restrictions on the electrode and the usability level when applied simultaneously. In order to overcome these restrictions, we designed a cap incorporating innovative electrode concepts for simultaneous TES-EEG applications. We aimed to verify the cap’s functionality in a visually evoked potential (VEP) paradigm. The cap was made of flat knitted highly flexible fabric. The stimulation electrodes of conductive fabric with silver coated polyamide threads were knitted as second layer inside the cap. This procedure generated a textile pocket between the inner electrode and the outer cap layer. Saline soaked sponges were placed in the pockets as electrolyte reservoirs. Silicone infiltrating the fabric and surrounding the stimulation electrodes acted as diffusion barrier. We incorporated dry EEG electrodes based on PU substrates with Ag/AgCl coating as recording electrodes in the cap. The resulting TES-EEG cap was tested on 10 healthy volunteers (mean age 24.8 ± 2.5 years, 2 women) in a checkerboard pattern reversal experiment. VEPs were recorded at POz, PO1, PO2, PO5, PO6, PO9 and PO10 with a reference at Fz. The stimulation electrodes were positioned at Cz (anode, 4.5 cm x 4.0 cm) and Oz (cathode, 4.5 cm x 3.5 cm). TDCS with 1 mA was applied for 10 min. Six VEP measurements with 300 pattern reversals each were performed: 1 before, 2 during and 3 after tDCS. In accordance to the previously reported stimulation effect in this VEP paradigm performed with interleaved conventional tDCS, the N75 component was significantly decreased post-stimulation. This demonstrated the feasibility of the new cap for simultaneous TES and EEG applications. Additionally, we observed a significant reduction of the P100 component occurring during tDCS. This indicated a different neuro-modulatory effect only observable in simultaneous TES-EEG application. In conclusion, the TES-EEG cap represents a new tool for simultaneous TES-EEG applicable in multiple scenarios such clinical research, therapy monitoring and closed-loop stimulation. The TES-EEG cap overcomes limitations of conventional equipment applied to combine TES and EEG.

The function and timing of EEG α-power in attention alerting and orienting

Alberto Zani (Institute for the History of Philosophical and Scientific Thought in Modern Age (ISPF); National Research Council (CNR))
Co-authors: Alice Mado Proverbio (“Neuro-Mi” Milan Center for Neuroscience, University of Milano-Bicocca)
Click here for abstract
Previous literature has indicated that brain EEG α-frequency power is closely involved in the functional regulation of visuospatial alerting and orienting of attention functions. The most recent evidence points to a determining role of occipito-parietal α-band oscillations (8–14 Hz) in anticipatory orienting facilitation (α-power decreases) vs. suppression (α-power increases). Yet, while such modulations are a common finding, the direction of modulation and its timing varies to a great extent across studies implying dependence on task demands. Furthermore, scarce knowledge is available about the power of these oscillations in the endogenous and exogenous orienting modes and about its timing.

In this study, nineteen volunteers were administered a modified version of Posner’s ANT (Attention Network Test) including three diverse cue-target visuospatial attention orienting tasks: namely, a CC (Central Cue; an alerting, but not orienting cue) task, a NC (No Cue; a no alerting and no orienting cue) task, and a peripheral LC (Local Cue; an alerting and attention orienting cue) priming tasks. On each trial, the EEG was recorded from 128 scalp sites going from 100 ms before the cue to 1300 ms after it, a target being delivered 500 ms after the cue, to which the participants had to give a press button motor response. Offline, we computed LC-CC and CC-NC difference waves to get access to attention orienting and alerting systems activation modes, respectively. EEG α-band oscillations power (in uV2/Hz) between 8 - 12 Hz was computed by means of the Morlet’s wavelet-based Compressed Spectral Density Array (CSDA), so to investigate its relationships with the indicated attention networks and its time progression.

A much greater α-power was observed at parietal-occipital scalp sites for the attention alerting than the exogenous orienting condition. This was observed both before and after the target delivery over the left hemis phere, and after the target appearance only over the right hemisphere. The results also pointed out that α-power progressively decreased over time, reaching the lowest power peak in the 300-800 ms post-target latency, independent of the functional activity mode.

Overall, our findings provide support for the view of a facilitating vs. inhibiting role of α-power decreases and increases for visual attention orienting and suggests that these changes are differentially deployed during attention orienting and alerting to modulate vs. maintain the parietal-occipital cortex cells discharging frequency for optimization of information processing.

Pre-stimulus feedback connectivity biases the content of visual experiences

Elie Rassi (Centre for Cognitive Neuroscience, University of Salzburg)
Co-authors: Andreas Wutz (Centre for Cognitive Neuroscience, University of Salzburg; The Picower Institute for Learning & Memory and Department of Brain & Cognitive Sciences, Massachusetts Institute of Technology), Nadia Müller-Voggel (Center for Biomagnetismus, Department of Neurosurgery, University Hospital Erlangen; Center for Mind/Brain Sciences (CIMeC), University of Trento), Nathan Weisz (Centre for Cognitive Neuroscience, University of Salzburg; Center for Mind/Brain Sciences (CIMeC), University of Trento)
Click here for abstract
Ongoing fluctuations in neural excitability and in network-wide activity patterns before stimulus onset have been proposed to underlie variability in near-threshold stimulus detection paradigms, i.e. whether an object is perceived or not. Here, we investigated the impact of pre-stimulus neural fluctuations on the content of perception, i.e. whether one or another object is perceived. We recorded neural activity with magnetoencephalography before and while participants briefly viewed an ambiguous image, the Rubin face/vase illusion, and required them to report their perceived interpretation on each trial. Using multivariate pattern analysis, we showed robust decoding of the perceptual report during the post-stimulus period. Applying source localisation to the classifier weights suggested early recruitment of V1 and ~160 ms recruitment of category-sensitive FFA. These post-stimulus effects were accompanied by stronger oscillatory power in the gamma frequency band for face vs vase reports. In pre-stimulus intervals, we found no differences in oscillatory power between face vs. vase reports in V1 nor in FFA, indicating similar levels of neural excitability. Despite this, we found stronger connectivity between V1 and FFA prior to face reports for low-frequency oscillations. Specifically, the strength of pre-stimulus feedback connectivity (i.e. Granger causality) from FFA to V1 predicted not only the category of the upcoming percept, but also the strength of post-stimulus neural activity associated with the percept. Our work shows that pre-stimulus network states can help shape future processing in category-sensitive brain regions and in this way bias the content of visual experiences.

Mesoscopic scale functional interactions in the human brain

Gabriel Kreiman (Children's Hospital, Harvard Medical School)
Co-authors: Jerry Wang (Harvard University)
Click here for abstract
Functional interactions between brain regions play a central role in cognitive computations. Evaluating such functional interactions in the human brain has been challenging due to the difficulties inherent to interrogating human brain activity at adequate spatiotemporal resolutions and with sufficient signal-to-noise ratio. We investigated pairwise interactions at a mesoscopic scale by quantifying the degree of coherence in intracranial field potential recordings from 4432 electrodes in 48 patients with pharmacologically intractable epilepsy over the course of 6360 hours. After correcting for artifacts and removing seizure events, we defined putative interactions by computing the coherence in different frequency bands between electrode pairs within each patient. We observed functional interactions that are consistent with known anatomical connectivity in the human brain, with macaque anatomical connections, and with neurophysiological interactions documented in macaque monkeys. These interactions showed strong stability across days. The interactions were also were also consistent across subjects. These results provide the first mesoscopic functional interactome of the human brain and constitute an important database to study modulations by state, by cognitive function, as well as by impairments due to neurological disorder.

The social neuroscience of cooperation

Jay van Bavel (New York University)
We review literature from several fields to describe common experimental tasks used to measure human cooperation as well as the theoretical models that have been used to characterize cooperative decision-making, as well as brain regions implicated in cooperation. Building on work in neuroeconomics, we suggest a value-based account may provide the most powerful understanding the psychology and neuroscience of group cooperation. We also review the role of individual differences and social context in shaping the mental processes that underlie cooperation and consider gaps in the literature and potential directions for future research on the social neuroscience of cooperation. We suggest that this multi-level approach provides a more comprehensive understanding of the mental and neural processes that underlie the decision to cooperate with others.

Strategies for integrative analysis in Imaging Genetic studies

Natalia Vilor-Tejedor (Center for Genomic Regulation (CRG); BarcelonaBeta Brain Research Center (BBRC))
Click here for abstract
Imaging Genetic (IG) studies integrate neuroimaging and genetic data from the same individual, offering the opportunity to deepen our knowledge of the biological mechanisms of neurodevelopmental domains and neurological disorders. For instance, genetic factors are of potential interest to research on preventive practices. Preventive practices make personalized medicine possible by targeting those risk factors, allowing an appropriate treatment or preventive strategies for people who are at increased genetic risk of developing a specific neurodevelopmental disease. Moreover, the evolution of neuroimaging techniques and sequencing methods could translate into their use as fast and profitable tools in a clinical context by enabling the characterization – specifically the pathology – of the individual.

Although the literature on IG studies has sustained exponential growth during these last years, the majority of studies have mainly analyzed individual associations of candidate brain regions with genetic variants. However, this strategy was not designed to deal with the complexity of neurobiological mechanisms underlying behavioral and neurodevelopmental domains. Moreover, larger sample sizes and multidimensionality of these data represents a challenge for the standardization of modeling procedures in IG research field.

We provide a systematic update of current methods and strategies used in IG studies, which may serve as an analytical framework for researchers working in the neuroepidemiological field. In addition, we present an overview of how these methodological approaches are applied for the integration of neuroimaging and genetic data. We hypothesize that relevant IG findings in relation to biological mechanisms (e.g., the identification of new target genes involved in brain structure and functioning) might assist with the design of personalized disease-modifying drugs and treatments.

Mother of unification studies, a 204-subject multimodal neuroimaging dataset to study language processing.

Julia Uddén (Max Planck Institute for Psycholinguistics)
Co-authors: Jan-Mathijs Schoffelen (Radboud University, Donders Institute for Brain, Cognition and Behaviour), Robert Oostenveld (Radboud University, Donders Institute for Brain, Cognition and Behaviour; NatMEG, Karolinska Institutet), Nietzsche H.L. Lam (Radboud University, Donders Institute for Brain, Cognition and Behaviour), Annika Hultén (Radboud University, Donders Institute for Brain, Cognition and Behaviour; Max Planck Institute for Psycholinguistics; Department of Neuroscience and Biomedical Engineering, Aalto University), Peter Hagoort (Radboud University, Donders Institute for Brain, Cognition and Behaviour; Max Planck Institute for Psycholinguistics)
Click here for abstract
The Mother Of Unification Studies (MOUS) dataset contains multimodal neuroimaging data that has been acquired from 204 healthy human subjects. The neuroimaging protocol consisted of magnetic resonance imaging (MRI) to derive information at high spatial resolution about brain anatomy and structural connections, as well as functional data during task, and at rest. In addition, magnetoencephalography (MEG) was used to obtain high temporal resolution electrophysiological measurements during task, and at rest. All subjects performed a language task, during which they had to process linguistic utterances that either consisted of normal or scrambled sentences. Half of the subjects were reading the stimuli, the other half listened to the stimuli. The resting state measurements consisted of 5 minutes eyes-open for the MEG and 7 minutes eyes-closed for fMRI. The neuroimaging data, as well as the information about the experimental events are shared according to the Brain Imaging Data Structure (BIDS) format. This unprecedented neuroimaging language data collection allows for the investigation of various aspects of the neurobiological correlates of language.

Probabilistic TFCE: a generalised combination of cluster size and voxel intensity to increase statistical power

Tamas Spisak (Department of Neurology, University Hospital Essen)
Co-authors: Zsófia Spisák, Matthias Zunhammer and Ulrike Bingel (Department of Neurology, University Hospital Essen), Stephen Smith (Wellcome Centre For Integrative Neuroimaging (FMRIB), University of Oxford), Thomas Nichols (Wellcome Centre For Integrative Neuroimaging (FMRIB), University of Oxford; Department of Statistics, University of Warwick), Tamás Kincses (Department of Neurology, University of Szeged)
Click here for abstract
Simple voxel-wise inference in neuroimaging might be suboptimal for two reasons. First, correcting for multiple comparisons in typical mass-univariate analyses typically results in large Type II errors (i.e. missing true effects). Second, the inherent spatial smoothness of typical neuroimaging data does not allow for taking advantage of the high (voxel-level) localising power of this approach. Boosting belief in extended areas of signal, as performed e.g. by the threshold-free cluster enhancement (TFCE) approach take advantage of image smoothness and enhance detectability of signal. Applications of this technique are limited by several factors: (i) generalisation to data structures, like brain network connectivity data is not trivial, (ii) TFCE values are in an arbitrary unit, therefore, P-values can only be obtained by a computationally demanding permutation-test.

Here, we introduce a probabilistic approach for TFCE (pTFCE), that gives a simple general framework for topology-based belief boosting.

The core of pTFCE is a conditional probability, calculated based on Bayes’ rule, from the probability of voxel intensity and the threshold-wise likelihood function of the measured cluster size. In this paper, we provide an estimation of these distributions based on Gaussian Random Field theory. The conditional probabilities are then aggregated across cluster-forming thresholds by a novel incremental aggregation method. pTFCE is validated on simulated and real fMRI data. The results suggest that pTFCE is more robust to various ground truth shapes and provides a stricter control over cluster “leaking” than TFCE and, in many realistic cases, further improves its sensitivity.

Correction for multiple comparisons can be trivially performed on the enhanced P-values, without the need for permutation testing, thus pTFCE is well-suitable for the improvement of statistical inference in any neuroimaging workflow. Implementation of pTFCE is available at

Diagnosing dementia when it matters: AI-based Integrated Cognitive Assessment for Earlier Diagnosis

Seyed-Mahdi Khaligh-Razavi (Cognetivity Neurosciences )
Co-authors: Chris Kalafatis (Department of Old Age Psychiatry, King’s College London), Mohammad Hadi Modarres (Cognetivity ltd), Haniye Marefat (School of Cognitive Sciences, Institute for Research in Fundamental Sciences (IPM)), Hamed Karimi (Amir Kabir University of Technology), Mahdiyeh Khanbagi (Department of Brain and Cognitive Sciences, Cell Science Research Center, Royan Institute for Stem Cell Biology and Technology, ACECR), Zahra Vahabi (Tehran University of Medical Sciences)
Click here for abstract
Despite significant progress in biomarker development for the diagnosis of dementia, these have yet to translate into efficient early diagnostic tools primarily due to limited sensitivity, cost considerations and ease of access. We present the Integrated Cognitive Assessment (ICA), a 5-minute, self-administered, computerised test that is independent of language, cultural background and education. The ICA addresses standard assessment limitations by targeting cognitive domains that are affected in the initial stages of the disease, specifically before the onset of memory symptoms and by utilising Artificial Intelligence (AI) to analyse high-dimensional clinical and demographic data to continuously improve its predictive power.

We carried out head-to-head studies with widely used cognitive assessment tools (i.e. MoCA, ACE, BICAMS) in participants with mild Alzheimer’s Disease (AD), mild cognitive impairment (MCI), and cognitive impairment secondary to MS. Furthermore, we compared ICA results with Serum NfL, and ran a task-based fMRI study to identify brain areas engaged by the ICA test.

The ICA shows convergent validity with MoCA, ACE-III, and BICAMS cognitive assessments, and achieves high accuracy (94%, AUC) in detecting cognitive impairment. Within those with cognitive impairment, ICA discriminates MCI from mild AD with 96% accuracy. Furthermore, in the task-based fMRI investigation, we find that the ICA task engages brain areas, such as parahippocampal, fusiform gyrus, inferior and middle temporal, that are anatomically identified among the earliest areas affected by tau-pathology in pre-symptomatic stages of AD, as shown by Braak and colleagues, 2006. The ICA also shows a strong correlation with NfL and severity of cognitive impairment; and demonstrates excellent test-retest reliability and no evidence of a learning bias (Khaligh-Razavi et al, 2019), factors which affect standard cognitive assessments.

The ICA platform can impact dementia diagnosis and research on a large scale. With the advent of new AI tools (such as the ICA platform) that have the potential to identify health conditions earlier, pharmaceutical companies will need to shift their focus on treating early-stage diseases, rather than late-stage ones. This can fundamentally change the way the industry operates today towards more preventative rather than curative care.


Discussions can freely continue under the hashtag #brainTC.

Somatosensory responses to nothing: an MEG study of expectations during omission of tactile stimulations

Lau Andersen (Karolinska Institutet)
Co-authors: Daniel Lundqvist (NatMEG, Karolinska Institutet)
Click here for abstract

The brain builds up expectations to future events based on the patterns of past events. This function has been studied extensively in the auditory and visual domains using various oddball paradigms, but only little exploration of this phenomenon has been done in the somatosensory domain. In this study, we explore how expectations of somatosensory stimulations are established and expressed in neural activity as measured with magnetoencephalography.


Using tactile stimulations to the index finger, we compared conditions with actual stimulation to conditions with omitted stimulations, both of which were either expected or unexpected.


Our results show that when a stimulation is expected but omitted, a time-locked response occurs ∼135 ms subsequent to the expected stimulation. This somatosensory response to “nothing” was source localized to the secondary somatosensory cortex and to the insula. This provides novel evidence of the capability of the brain of millisecond time-keeping of somatosensory patterns across intervals of 3000 ms.

Our results also show that when stimuli are repeated and expectations are established, there is associated activity in the theta and beta bands. These theta and beta band expressions of expectation were localized to the primary somatosensory area, inferior parietal cortex and cerebellum. Furthermore, there was gamma band activity in the right insula for the first stimulation after an omission, which indicates the detection of a new stimulation event after an expected pattern has been broken.

Finally, our results show that cerebellum play a crucial role in predicting upcoming stimulation and in predicting when stimulation may begin again.


The refractory beta band activity found in the cerebellum and the inferior parietal cortex after a stimulation may be interpreted as the beta band signalling the status quo – that a predictable sequence of stimulations is expected. The cerebellum may thus act as a time-keeper.

Cerebral mechanisms underlying time production in healthy adults: An fNIRS study

Ségolène Guérin (SCALab, University of Lille)
Co-authors: Marion Vincent and Yvonne Delevoye-Turrell (SCALab UMR 9193, Lille University)
Click here for abstract
Humans are able to modify the spontaneous pace of their actions to interact with their environment and peers. Developing with age, the ability to adapt the pacing of actions would depend on high level cognitive functions (e.g., inhibition) and thus, may be underpinned by frontal activations. However, little is known concerning the brain areas underlying such temporal control. The objective here was to investigate the differences in cerebral oxygenation in prefrontal and motor areas when adult participants were instructed to speed up or slow down a sequential motor task. For this purpose, participants performed a spatial finger tapping task in synchronization with an external metronome; the motor complexity of the task was controlled. Cerebral haemodynamic responses were recorded using the functional near infrared spectroscopy (fNIRS) optical imaging technique, which is a non-invasive imaging method that makes use of optical proprieties of light to evaluate local haemodynamic responses in a given cortical area. In the present study, a particular caution was taken to avoid hair contamination, as it induces light absorption which is not associated to brain activity. The results confirmed that the fNIRS tool is particularly salient to motor paradigms given that it is far less sensitive to movement than traditional scanning techniques such as fMRI or EEG. In addition, results showed that complex motor sequences globally engaged more cerebral activation compared to simple motor ones. Slow-rate sequences were characterized by more prefrontal and motor activations than movements performed at fast-pace. In addition, with motor complexity, activations were greater in the prefrontal areas than in the motor areas. Overall, these findings provide strong support for the involvement of cognitive functions in the pacing of motor sequences, and especially when required to produce movements slower than the spontaneous tempo. We also highlight the advantages of fNIRS whole-brain models to dissociate the involvement of different cognitive mechanisms in motor paradigms as a function of task complexity.

Perceptual decision-making: Attractor dynamics explain post-error adjustments.

Kevin Berlemont (Laboratoire de Physique, ENS PSL University, Université Paris Diderot, Université Sorbonne Paris Cité, Sorbonne Université CNRS)
Co-authors: Jean-Pierre Nadal (LPENS, PSL University Université Paris Diderot, Université Sorbonne Paris Cité, Sorbonne Université, CNRS and CAMS, EHESS, PSL University, CNRS)
Click here for abstract
Perceptual decision-making is the subject of many experimental and theoretical studies. Most modeling analyses are based on statistical processes of accumulation of evidence. In contrast, very few works confront attractor network models’ predictions with empirical data from continuous sequences of trials. Recently however, numerical simulations of a biophysical competitive attractor network model have shown that such network can describe sequences of decision trials and reproduce repetition biases observed in perceptual decision experiments. Here we get more insights into such effects by considering an extension of the reduced attractor network model of Wong and Wang (2006), taking into account an inhibitory current delivered to the network once a decision has been made. We make explicit the conditions on this inhibitory input for which the network can perform a succession of trials, without being either rapped in the first reached attractor, or losing all memory of the past dynamics. We study in details how, during a sequence of decision trials, reaction times and performance depend on the nonlinear dynamics of the network, and we confront the model behavior with empirical findings on sequential effects. Here we show that, quite remarkably, the network exhibits, qualitatively and with the correct orders of magnitude, post-error slowing and post-error improvement in accuracy, two subtle effects reported in behavioral experiments in the absence of any feedback about the correctness of the decision. Our work thus provides evidence that such effects result from intrinsic properties of the nonlinear neural dynamics.

Realistic real-time modelling of the TMS-induced electric field

Matti Stenroos (Aalto University, Department of Neuroscience and Biomedical Engineering)
Co-authors: Lari M. Koponen (Aalto University, Department of Neuroscience and Biomedical Engineering)
Click here for abstract
In transcranial magnetic stimulation (TMS), a current pulse in a coil gives rise to a time-varying magnetic field that causes an electric field (E-field) that stimulates the brain. In navigated TMS, the stimulation is targeted and dosed using a model of E-field that takes also the head conductivity into account. State of the art in real-time experimental TMS navigation is to approximate the head as a spherically symmetric conductor, omitting the effect of cerebrospinal fluid (CSF) on the E-field. The spherical approximation can cause an error of tens of percents to the estimated E-field. So far, realistic head models have been too slow for TMS navigation. Here, we present novel computational methods that enable real-time computation of E-field in a realistic head model that contains the CSF.

Using reciprocity (Heller & van Hulsteyn, 1992), we convert the TMS forward problem to a form similar to the forward problem of magnetoencephalography. Then, we split the computations to coil-independent and coil-dependent parts using discretized Geselowitz boundary-integral equation (Geselowitz 1970). The coil-independent part is solved offline; we did it using Galerkin boundary element method (Stenroos & Sarvas 2012). To solve the coil-dependent part in real time, we developed a moment-matching method for modelling the TMS coil with a small number of dipoles and a new, fast numerical quadrature for approximating the Geselowitz integrals.

We verified and benchmarked the new methods using a four-compartment head model (brain, CSF, skull, scalp; altogether 21000 vertices).We computed the E-field on the whole cortex (20000 vertices) for 251 different TMS coil positions. As reference we used a high-density coil model (1 mm spacing, 85288 dipoles) with analytical boundary integrals. All computations were carried out in a standard PC using Matlab; the speed tests used Nvidia GeForce GTX 1060 GPU.

The fast quadrature introduced a relative error of 1.1%. The total error of the quadrature and coil model was 1.43% and 1.15% for coils with 38 and 76 dipoles, respectively. For comparison, the difference between our head model and a simpler realistic model that omits the CSF was 29%. Using a standard PC and a 38-dipole coil, our solver computed one component of the E-field in 84 coil positions per second in 20000 points on the cortex. The full 3D E-field was computed at the speed of 28 cps.

The presented methods enable real-time solving of the TMS-induced E-field in a realistic head model that contains the CSF. The new methodology allows more accurate targeting and precise adjustment of intensity during experimental or clinical TMS mapping.

Identifying cellular and molecular substrates of mouse brain feature maps

Tina Segessemann (ETH Zurich)
Co-authors: Horea-Ioan Ioanas (ETH Zurich)
Click here for abstract
In the effort of identifying better drug and treatment targets, it is of great scientific interest to compute highest-likelihood cellular and molecular targets for the mechanism underlying a given brain feature map. While researchers increasingly release high-level brain feature maps as part of their publications, the Allen Brain Institute (ABI) publishes maps for the expression of over 20.000 genes and the projections of over 2.000 cell clusters in the mouse brain. We facilitate the high-throughput usability of this extensive data set, and its integration with the increasing number of phenotypically and behaviourally interesting high-level feature maps produced by individual researchers. We present an interface and package distribution format, which allows researchers to compare a feature map of interest against all available ABI mouse brain maps using the full flexibility of their own data analysis system, and identify highest-likelihood candidates for underlying cellular and molecular features. We further examine the performance and limits of spatial-correlation-driven feature identification, and present a conceptual framework of interpreting results.

Stimulation of vagal afferents shapes reinforcement learning

Nils Kroemer (University of Tübingen)
Co-authors: Anne Kühnel (Department of Translational Research in Psychiatry, Max Planck Institute of Psychiatry and International Max Planck Research School for Translational Psychiatry (IMPRS-TP)), Vanessa Teckentrup and Monja P. Neuser (Department of Psychiatry and Psychotherapy, University of Tübingen), Quentin J. M. Huys (Division of Psychiatry, University College London; Max Planck UCL Centre for Computational Psychiatry and Ageing Research), Caroline Burrasch (Department of Psychiatry and Psychotherapy, University of Tübingen), Martin Walter (Department of Psychiatry and Psychotherapy, University of Tübingen; Otto-von-Guericke University Magdeburg, Department of Psychiatry and Psychotherapy; Clinical Affective Neuroimaging Laboratory; Leibniz Institute for Neurobiology), Nils B. Kroemer (Department of Psychiatry and Psychotherapy, University of Tübingen)
Click here for abstract
The vagus nerve plays a vital role in the regulation of appetitive behavior according to homeostatic needs. Whereas the impact of metabolic states such as hunger on motivation are well documented, whether and how vagal signals shape instrumental action learning is unknown. Here, we investigated the effect of non-invasive auricular transcutaneous vagus nerve stimulation (tVNS) versus sham stimulation on approach and avoidance behavior using a go/no-go reinforcement learning task. To this end, 39 healthy participants received stimulation after an overnight fast following a single-blind randomized cross-over design. tVNS acutely impaired decision-making which was reflected in a reduced accuracy of action choices in a mixed-effects logistic regression analysis. Critically, by using computational reinforcement learning models, we identified that the cause of this reduction was a lowered learning rate through tVNS, particularly after punishment. In contrast, tVNS had no effect on response biases or response time indicating that it primarily affected learning rather than action execution. These results highlight a novel role of vagal afferent input in modulating reinforcement learning by tuning the learning rate according to homeostatic needs. Collectively, our findings add to the growing literature showing a tight interplay between appetitive behavior and peripheral feedback signals that are transmitted via the vagus nerve.

Attenuated beta rebound to proprioceptive afferent feedback in Parkinson’s disease

Mikkel C. Vinding (NatMEG, Karolinska Institutet)
Co-authors: Panagiota Tsitsi (Section of Neurology, Department of Clinical Neuroscience, Karolinska Institutet), Harri Piitulainen (Aalto NeuroImaging, Department of Neuroscience and Biomedical Engineering, Aalto University), Josefine Waldthaler (Section of Neurology, Department of Clinical Neuroscience, Karolinska Institutet), Veikko Jousmäki (NatMEG, Department of Clinical Neuroscience, Karolinska Institutet; Aalto NeuroImaging, Department of Neuroscience and Biomedical Engineering, Aalto University; Cognitive Neuroimaging Centre, Nanyang Technological University), Martin Ingvar (NatMEG, Department of Clinical Neuroscience, Karolinska Institutet), Per Svenningsson (Section of Neurology, Department of Clinical Neuroscience, Karolinska Institutet) & Daniel Lundqvist (NatMEG, Department of Clinical Neuroscience, Karolinska Institutet)
Click here for abstract
Motor symptoms are defining traits in the diagnosis of Parkinson’s disease (PD). A crucial component in motor function is the integration of afferent proprioceptive sensory feedback. Previous studies have indicated abnormal movement-related cortical oscillatory activity in PD, but the role of the proprioceptive afferents on abnormal oscillatory activity in PD has not been elucidated. We examine the cortical oscillations in the mu/beta-band (8-30 Hz) in the processing of proprioceptive stimulation in PD patients, ON/OFF levodopa medication, as compared to that of healthy controls (HC). We used a proprioceptive stimulator that generated precisely controlled passive movements of the index finger and measured the induced cortical oscillatory responses following the proprioceptive stimulation using magnetoencephalography. Both PD patients and HC showed a typical beta-band desynchronization during the passive movement. However, the subsequent beta rebound after the passive movement that was almost absent in PD patients compared to HC. Furthermore, we found no difference in the degree of beta rebound attenuation between patients ON and OFF levodopa medication. The results demonstrate a disease-related deterioration in cortical processing of proprioceptive afference in PD.

How much do you trust a model? - Rigor in neuroscientific modeling and simulation through validation

Robin Gutzen (Institute of Neuroscience and Medicine (INM-6) and Institute for Advanced Simulation (IAS-6) and JARA-Institut Brain Structure-Function Relationships (INM-10), Jülich Research Centre)
Co-authors: Michael von Papen (Institute of Neuroscience and Medicine (INM-6) and Institute for Advanced Simulation (IAS-6) and JARA-Institut Brain Structure-Function Relationships (INM-10), Jülich Research Centre), Guido Trensch (Simulation Lab Neuroscience, Jülich Supercomputing Centre, Institute for Advanced Simulation, JARA, Jülich Research Centre), Pietro Quaglio (Institute of Neuroscience and Medicine (INM-6) and Institute for Advanced Simulation (IAS-6) and JARA-Institut Brain Structure-Function Relationships (INM-10), Jülich Research Centre), Sonja Grün (Institute of Neuroscience and Medicine (INM-6) and Institute for Advanced Simulation (IAS-6) and JARA-Institut Brain Structure-Function Relationships (INM-10), Jülich Research Centre,Theoretical Systems Neurobiology, RWTH Aachen University), Michael Denker (Institute of Neuroscience and Medicine (INM-6) and Institute for Advanced Simulation (IAS-6) and JARA-Institut Brain Structure-Function Relationships (INM-10), Jülich Research Centre)
Click here for abstract
Modeling and the simulation of the activity in neuronal networks is an essential part of modern neuroscience and represents a powerful vehicle to combine insights from experiments and theory into a coherent understanding of brain function. The only measure to assess how much trust we can place in a given model is how well it can predict the biological reality it aims to describe. Validation testing formalizes the comparison between measured and simulated data and quantifies their similarity. The resulting test scores characterize the model and determine its validity with respect to predictions concerning the experimental reference (Thacker et al., 2004). However, it is sometimes useful to directly compare two models by means analogous to validation testing. Such direct comparisons are not constrained by the scarcity and specificity of experimental data. In contrast to validation, direct comparisons between two models are not able to determine the descriptive power of a model regarding its reference to reality. They can, however, be greatly beneficial in evaluating the model’s consistency, robustness with respect to parameter variation, and directed improvements in the model development process (Gutzen et al., 2018). In either scenario, several aspects must be considered. Any validation test only considers a specific statistic of a certain aspect of a finitely sampled data set. Therefore, in order to gain a more complete and less biased evaluation, it is necessary to apply multiple validation tests, taking into account different aspects and statistical measures (Forrester & Senge, 1980). For example, for a neural network model the dynamics on the single-cell and network activity level are not trivially related, and thus should be regarded individually.

Here, we present a workflow and a software solution to provide the means to perform such validation tests for the activity on the network level: NetworkUnit (RRID:SCR_016543). This Python package is based on the open-source projects Elephant (RRID:SCR_003833) and SciUnit (RRID:SCR_014528).

For further formalization and reproducibility of neuroscientific modeling, it is beneficial to separate the model implementation from the simulation engine and make use of available simulators. Notably, different simulators are not necessarily equivalent (especially when using e.g. neuromorphic systems as simulators (van Albada et al., 2018)) and even small differences in the numerics have been shown to influence the simulated model behavior (Trensch et al., 2018). To evaluate and control for such influences, validation techniques can also be analogously applied for the quantitative comparison between simulators, as we demonstrate as well.

Funding by: EU's Horizon 2020 Framework Programme for Research and Innovation under Specific Grant Agreements No. 720270 and No. 785907 (HBP SGA1, SGA2); the Helmholtz Association Initiative and Networking Fund, No. ZT-I-0003; and HDSLEE.


Discussions can freely continue under the hashtag #brainTC.

Can biophysically principled neural modeling bring us beyond correlation to causation of human EEG/MEG signals?

Stephanie Jones (Brown University, Human Neocortical Neurosolver
Electro- and Magneto- encephalography (EEG/MEG) are currently the only methods to non-invasively record human brain activity with millisecond resolution. EEG/MEG signals are correlated with nearly all healthy and pathological brain activity. Yet, how these signals are generated at the cell and circuit level is still largely unknown. This lack of interpretability limits our understanding of the causal role of these signals in brain function, as well as our ability to develop novel treatments for neuropathology based on changes in these signals. Biophysically principled neural modeling can provide the missing link between the extracranially measured “macro” scale signal and the underlying “meso” scale cell and network level generators, because models can have specificity at both scales. I will discuss the development and application of a new neural modeling software created to provide this bridge: the “Human Neocortical Neurosolver (HNN)”, HNN’s foundational model is unique in its circuit level details and is embedded in a user-friendly graphical user interface designed for researchers and clinicians without neural modeling experience to develop and test hypotheses on the neural origin of their EEG/MEG data. The goal of HNN is to provide an unprecedented link from the measured EEG/MEG signals to their neural generators in service of connecting to complimentary animal research and ultimately improving our understanding of the causal role of EEG/MEG signals in health and disease.

Post-hoc modification of linear models: combining machine learning with domain information to make solid inferences from noisy data

Marijn van Vliet (Aalto University)
Co-authors: Riitta Salmelin (Aalto University)
Click here for abstract
Machine learning is a powerful technique which is increasingly used in neuroscience to explore ambitious research questions. The ability to “fit” the parameters of a model to training data allows the models to adapt to the unique signal and noise characteristics of each dataset. However, there are reasons why one would sometimes prefer a balance between aspects of the model that are fitted to the data, and aspects that one can define based on prior knowledge.

We have developed a framework that can be used to place restrictions on linear machine learning models that are in common use today (e.g. linear regression, logistic regression, linear support vector machines). This is achieved by combining the insight of Haufe et al. (Neuroimage 2014) that a pattern matrix can be computed for any linear model, with the insight from source estimation methods that priors that are formulated on the pattern matrix can be translated into priors on the weight matrix. These priors can be based, for example, on domain knowledge of the data, experimental paradigm and research question. This allows for an approach to data analysis where machine learning becomes less of a “black box” and more of a dialog between the re-searcher and the learning algorithm: one providing domain knowledge and the other providing the resulting optimal fit to the data.

As an example use case, we demonstrate how this framework can be used to address the “curse of dimensionality”, where the quantity of available training data (e.g. the number of trials) is insufficient to properly estimate all model parameters, leading to overfitting. It is a problem often faced in neuroscience studies. Our example problem consists of a decoding task, where the model aims to decode the associative strength between words, presented using a semantic priming paradigm, from EEG data. We show that by incorporating domain knowledge about the spatio-temporal nature of the data, data from other participants and information about the N400 potential, overfitting of the model could be reduced, leading to increased decoding performance.

However, increasing decoding performance is not the only reason to wish to apply restrictions to a general purpose machine learning algorithm. For example, by fixing aspects of the model, such as the signal of interest, while still allowing the model to estimate and cancel out noise sources, we can draw more robust conclusions about the results.

Preprint: Code:

Functional homogeneity of Regions of Interest predicts their topological roles as nodes of functional brain networks

Onerva Korhonen (Université de Lille, CNRS, UMR 9193 - SCALab)
Co-authors: Jari Saramäki and Elisa Ryyppö (Aalto University, Department of Computer Science)
Click here for abstract
Connectivity between brain areas is essential for normal human brain function. Therefore, a complex network is a natural model for the brain. Nodes of functional brain networks are brain areas (often referred to as Regions of Interest, ROIs), and links represent the temporal co-activation of ROIs. Depending on their function, ROIs play different roles in network topology. For example, the most central nodes of the network, hubs, divide into provincial and connector ones. Provincial hubs orchestrate the local activity around them, while connector hubs mediate connections between more distant parts of the brain.

In networks constructed from functional magnetic resonance imaging (fMRI) data, ROIs are collections of measurement voxels, typically defined a priori based on anatomical landmarks or functionality observed in earlier studies. Earlier, we have shown that ROIs consist of voxels that behave differently in time, yielding a rich voxel-level connectivity structure inside ROIs and low functional homogeneity. In other words, significant amounts of information get lost when the time series of voxels are averaged to obtain the ROI time series. Besides, functional homogeneity of ROIs changes in time, indicating that the connectivity of voxels changes in both within the fixed ROI boundaries and across these boundaries.

Here, we investigate how the time-dependent functional homogeneity and internal connectivity structure of ROIs relate to the topology of functional brain networks. To this end, we quantify internal connectivity with several measures. Then, we show that these measures, together with ROI size, can predict if a ROI is a network hub and if an identified hub ROI plays a provincial or connector role, as well as the ROI’s role in a more detailed schema of seven roles.

Our prediction models outperform random classification in all three classification tasks. In particular, high functional homogeneity predicts hubness in general, whereas among hubs low and fluctuating functional homogeneity together with unstable internal connectivity structure predicts a connector role. ROI size is an important predictor of hubness in general and provincial hubness in particular. However, we observe classification accuracies higher than random also after excluding ROI size from the model.

The low and fluctuating functional homogeneity is often interpreted as a technical flaw of the current ROI definition approaches. However, the clear connection between ROIs’ internal connectivity structure and their roles in brain network topology speaks against this interpretation. Instead, the fluctuating connectivity inside ROIs and the variation of functional homogeneity are genuine features of brain function that should not be omitted from brain network analysis. Indeed, network neuroscience would greatly benefit from new tools that are able to combine information about the connectivity both inside and between ROIs.

White matter changes in offspring of patients with Alzheimer’s Disease

Stella Maris Sanchez (INAAC Group; Buenos Aires University; CONICET)
Co-authors: Gabriela De Pino (INAAC Group; Centro Universitario de Imágenes Médicas, Escuela de Ciencia y Tecnología, Universidad Nacional de San Martín; Laboratory of Neuroimaging, Department of Imaging, FLENI Foundation), Hernán Bocaccio (INAAC Group; CONICET; FCEyN, University of Buenos Aires), Mariana N. Castro (INAAC Group; CONICET; Department of Physiology, University of Buenos Aires School of Medicine; Department of Mental Health, University of Buenos Aires School of Medicine), Bárbara Duarte-Abritta (INAAC Group), Carolina Abulafia (INAAC Group; CONICET; BIOMED, Pontifical Catholic University of Argentina), Daniel E. Vigo (CONICET; BIOMED, Pontifical Catholic University of Argentina), Salvador M. Guinjoan (INAAC Group; CONICET; Department of Physiology, University of Buenos Aires School of Medicine; Department of Mental Health, University of Buenos Aires School of Medicine; Service of Psychiatry, FLENI Foundation;Neurophysiology I, University of Buenos Aires School of Psychology), Mirta F. Villarreal (INAAC Group; CONICET; FCEyN, University of Buenos Aires)
Click here for abstract
The earliest preliminary neuropathological anomalies characteristic of Alzheimer’s disease involve intracellular Tau accumulation ultimately affecting white matter tracts. The microstructure of these tracts can be studied through metrics derived from diffusion MRI (dMRI), such as fractional anisotropy (FA) mean diffusivity (MD), and fiber density (FD). In this work, we analysed the microstructure of white matter tracts in a group of 23 middle-aged (40-60 years), cognitively intact individuals who present a higher risk of developing late-onset Alzheimer’s disease (LOAD) by virtue of having at least one parent diagnosed and 22 age- and sex-matched control subjects (CS) without family history of AD. In addition, we explore the relationship between metrics derived from dMRI and cognitive performance of participants. We observed a decreased fiber density in specific areas of white matter in the offspring of patients with LOAD (O-LOAD) group compared with CS group. This compromise in white matter integrity was associated with decreased semantic fluency in O-LOAD group. These results suggest that white matter integrity metrics could reveal changes in cognitively normal persons who have family history of LOAD.

Automated segmentation in the diagnosis of focal cortical dysplasias with magnetic resonance imaging.

Gonzalo Rojas (Laboratory for Advanced Medical Image Processing, Department of Radiology, Health Innovation Center and Advanced Epilepsy Center, Clínica la Condes)
Co-authors: M Magdalena Sepúlveda (Department of Radiology, Clínica las Condes), Evelyng Faure (Advanced Epilepsy Center and Department of Radiology, Clínica las Condes), Claudio R Pardo (Department of Radiology, Clínica las Condes), María de la Iglesia-Vayá (Unidad Mixta de Imagen Biomédica FISABIO-CIPF), José Molina Mateo (Center for Biomaterials and Tissue Engineering, Universitat Politècnica de València), Marcelo Gálvez (Health Innovation Center, Advanced Epilepsy Center and Department of Radiology, Clínica las Condes)
Click here for abstract
Objective: Focal cortical dysplasias (FCD) are a frequent cause of epilepsy. It has been reported that up to 40% of them cannot be visualized by means of conventional magnetic resonance imaging. The main objective of this work is to evaluate whether the automated brain segmentation is useful for detecting FCD.

Materials and Methods: through a retrospective, descriptive observational study, 155 patients who underwent surgery, registered in the database of the Advanced Epilepsy Center between the years 2009 and 2016, were reviewed. 20 patients with FCD confirmed by histology and a preoperative segmentation study (with FreeSurfer), with ages ranging from 3 to 43 years (14 men) were analyzed. Three expert neuroradiologists analyzed conventional and advanced MRI imaging with automated segmentation. They were classified into positive and negative concerning the visualization of FCD by consensus.

Results: We defined: Cortical true thickening: visible increase in cortical thickness as evidenced in FreeSurfer segmentation, defined by visualization of the same cortical thickness in surface segmentation (lines) and volume (color). Cortical pseudothickening: an increase in cortical thickening visualized in volume segmentation (color) which is not represented in surface segmentation (lines).

Of the 20 patients evaluated with conventional MRI, 12 were positive for FCD. Of the negative studies for FCD with conventional MRI, 25% were positive when they were analyzed with automated segmentation. In 13 of the 20 patients (38.5%), cortical true thickening was observed and pseudothickening in the rest of patients in the anatomical region of the brain corresponding to the dysplasia.

Conclusion: This work demonstrated that automated brain segmentation helps to increase detection of focal cortical dysplasias that are unable to be visualized in conventional MRI images.

Neuroanatomy of serial position effects in dementia at learning, short and delayed recall

Nancy Foldi (Department of Psychology, Queens College; The Graduate Center, City University of New York)
Co-authors: Kirsten I. Taylor (Neuroscience, Ophthalmology, and Rare Diseases (NORD), Roche Pharma Research and Early Development, Roche Innovation Center Basel, F. Hoffmann-La Roche Ltd; Faculty of Psychology, University of Basel), Andreas U. Monsch (Memory Clinic, University Center for Medicine of Aging Basel, Felix-Platter Hospital), Marc Sollberger (Memory Clinic, University Center for Medicine of Aging Basel, Felix Platter-Hospital; Department of Neurology, University Hospital Basel), Sasa L. Kivisaari (Department of Neuroscience and Biomedical Engineering, Aalto University)
Click here for abstract
Serial Position Effects (SPE) in wordlist learning provide a rich set to of metrics of cognitive functioning. In particular, the recall of primacy and recency items from the beginning and end of the word list, respectively, putatively have a distinct neuroanatomical basis. The two goals of this study were to (1) systematically map the neuroanatomical correlates of SPEs after across Learning (across Trials 1-5) (acc, Short delay (SD) and Long delay (LD) and (2) test whether progression of in performance accuracy from Learning to LD serves as a sensitive measure of dementia-related pathology. Primacy, middle, and recency SPE scores of the California Verbal Learning Test at Learning (across trials 1-5), SD, and LD in healthy controls and patients with dementia-related pathology were correlated with MRI gray matter signal intensities (voxel-based morphometry, VBM). As expected, the VBM analyses revealed distinct patterns of brain-behavioral correlations depending on both the SPE (primacy, middle, recency) and the time-point (Learning, SD, LD). Interestingly, we found that the healthy controls’ performance improved with primacy items from Learning to SD, and to LD (i.e., “primacy progression”), whereas the patients’ did not. Moreover, the proportion of primacy items recalled at LD compared to Learning correlated with bilateral MTL regions which commonly bear the brunt of pathology in dementia. We did not observe a similar effect for either middle or recency region. The findings demonstrate that the SPEs have distinct functional neuroanatomy. The results suggest the primacy progression score should be used as an accessible and sensitive measure for dementia detection, progression, and therapeutic response.


Discussions can freely continue under the hashtag #brainTC.

Toward a fully open ecosystem for neuroimaging analysis and data sharing

Russ Poldrack (Stanford University)
I will describe the ongoing development of a set of open-source resources for the analysis and sharing of neuroimaging data. I will discuss the various levels of data sharing and the role of the Brain Imaging Data Structure in supporting effective sharing of raw fMRI data. I will also discuss the emerging Python-based ecosystem for neuroimaging data analysis.

Decrease of long-range temporal correlations during loss of consciousness

Dominik Krzemiński (Cardiff University)
Co-authors: Maciej Kamiński (Faculty of Physics, University of Warsaw), Artur Marchewka and Michał Bola (Laboratory of Brain Imaging, Nencki Institute of Experimental Biology of Polish Academy of Sciences)
Click here for abstract
Consciousness has been hypothesised to emerge from complex neuronal dynamics, which prevails when brain operates in a critical state. In this study we investigate a relation between critical dynamics and consciousness by means of long-range temporal correlations (LRTCs). We utilise electrocorticography signals recorded from 4 macaque monkeys during resting wakefulness and general anaesthesia. The dataset analyzed in the study was recorded at the RIKEN Institute (Japan) and is publicly available on the Neurotycho database. The experimental protocol was approved by the RIKEN Ethics Committee. Detrended Fluctuation Analysis was used to estimate LRTCs in amplitude fluctuations of band-pass filtered signals. We found that during conscious states all lateral cortical regions are characterised by significant LRTCs of alpha-band activity (7-14 Hz). LRTCs are stronger in the eyes-open than eyes-closed state, but in both states they form a spatial gradient, with anterior brain regions exhibiting stronger LRTCs than posterior regions. Furthermore, we observed a substantial decrease of LRTCs during loss of consciousness, the magnitude of which was associated with the pre-anaesthetised state of the brain. Therefore, our results suggest that brain activity is characterised by robust LRTCs, which are disrupted by a loss of consciousness.

The brain’s somatosensory mu rhythms: a case study for where individual and social meet

Emmanuele Tognoli (Center for Complex Systems and Brain Sciences @ FAU)
Co-authors: Basak Kocaoglu and Erik Engeberg (FAU)
Click here for abstract
Mu rhythms are important neuromarkers believed to relate to feeling own’s and others’ movements; and a promising yet incompletely-fulfilled signal to control computers, robotic and prosthetic technologies. Their dynamics remain poorly understood, and especially how left and right mu interact with each other and why? [2] We previously demonstrated a spectral dissociation of left and right mu and their metastable dynamics (Tognoli & Kelso, 2013). [3] Here we provide an ultra-detailed analysis of the neural coordination dynamics from two subjects (a leader and a follower) performing intrinsic movement/arrest; and observing a partner’s movement in a reanalysis of data from (Suutari et al., 2010). Dual-EEG provided electrophysiological information on both brains, behavior consisted of repetitive self-paced right index finger movement at about 1-3Hz. Subjects performed 40 trials consisting of 5 consecutive intervals during which they (1) move without viewing each others; (2) leader continues to move, follower stops; (3) both stay motionless; (4) follower imitates leader’s movement, leader observes; (5) both stay motionless again. Each phase is 8s. [4] Follower’s EEG spectrum was strongly dominant for left mu; leader’s had a greater balance of left and right mu to start with and right mu occurrence increased further in later trials. In follower, two very similar mu dynamics emerged in phase 2 (stop moving, observe other’s movement) and in phase 5 (after cessation of own movement), both about 1-3s after arrest. In leader, a compound of left and right mu emerged after cessation of own movement (phase 3) and continued throughout the remainder 24s. [5] Our pervasive observation of loose time locking with arrest of self or other’s movement suggest that mu pertains to some endogenous processes, not a response to behavioral states. We also observed greater difference in dynamical organization of left and right mu as a function of assigned roles (leader, follower) than agent (self, other) which highlights the possibility of their causal roles in terms of perception~action cooperation (perception of movement and producing action accordingly). [6] Our results suggest that the dynamical organization of left and right mu is complex, probably endogenous, context dependent, and can be deciphered by detailed analysis of the oscillations’ dynamics. They aim to support our future efforts to built a neuroprosthesis that “feel”.

The relationship between white matter and reading acquisition, refinement and maintenance

Kulpreet Cheema (Department of Neuroscience, University of Alberta)
Co-authors: Jacqueline Cummine (Faculty of Rehabilitation Medicine, University of Alberta)
Click here for abstract

Reading requires efficient communication between brain regions that are situated all over the brain. These brain areas are structurally connected by white matter connections that develop over the period of reading acquisition. Notably, white matter is highly plastic during school years due to age-related maturation, and are prone to changes in response to experiential factors like reading acquisition. Here, we conducted a cross-sectional study to examine the relationship between reading performance and the stages of developmental progression of white matter pathways associated with: reading acquisition (i.e., 3-6 year olds), reading skill development (6-10 year olds), reading refinement (11-14 year and 15-17 year olds), and skilled readers (18-21 year olds).


Behavioural (reading performance) and neuroimaging (diffusion tensor imaging (DTI)) data were collected from participants aged 3-21 years as a part of the multi-site project called Pediatric Imaging Neurocognition Genetics (PING) study. DTI measures (fractional anisotropy and mean diffusivity) of bilateral dorsal tracts (arcuate fasciculus and parietal superior longitudinal fasciculus) and ventral (uncinate fasciculus, inferior fronto occipital fasciculus and inferior longitudinal fasciculus) were extracted. Reading performance was calculated as the number of items correctly read (words for older children and letters in the case of young children).Correlational and regression analyses were conducted between the DTI measures and reading scores.


During the early stages of reading acquisition (ages 3-6), dorsal tracts are positively related to reading performance (as FA goes up, reading performance goes up). For ages 6-10, a relationship between ventral tracts and reading performance begins to emerge. This brain-behaviour relationship in the ventral tracts changes into a negative association with reading from age 10 onwards. In addition to the involvement of left hemispheric tracts, this study revealed the initial engagement of right hemispheric tracts during the early stages of reading acquisition. These different associations of white matter tracts with reading during development will be discussed in terms of the biological processes of myelination and pruning.


As this is the first cross-sectional study to look into the emergence of white matter pathways for reading from childhood to adulthood, findings can be used to inform the structural and functional developmental models of brain development during reading acquisition.

Transcranial Direct Current Stimulation (tDCS) paired with a decision-making task reduces risk-taking in a clinically impulsive sample

Casey Gilmore (Defense and Veterans Brain Injury Center; Minneapolis VA Medical Center)
Co-authors: Patricia J. Dickmann and Greg J. Lamberty (Minneapolis VA Health Care System; University of Minnesota), Kelvin O. Lim (Defense and Veterans Brain Injury Center; Minneapolis VA Health Care System; University of Minnesota)
Click here for abstract
Impulsivity is a multidimensional construct that includes lack of premeditation, sensation-seeking, and impaired cognitive control. Impulsivity is often present in many clinical conditions, including traumatic brain injury (TBI), post-traumatic stress disorder (PTSD), and substance use disorder. Impulsivity typically manifests as poor decision-making or excessive risk-taking, and is difficult to treat and manage. Transcranial direct current stimulation (tDCS) applied over dorsolateral prefrontal cortex (DLPFC) - an area involved with cognitive control functions - has shown promise as an intervention to reduce impulsivity. This was a randomized, single-blind, sham-controlled study to investigate the effects of tDCS paired with a decision-making task on risk-taking in Veterans with a clinical history of impulsive behavior. Veterans were referred to the study because of observed problems with impulsivity. Participants were randomized to either active or sham tDCS, and completed two tDCS sessions per day for five days. During a session, participants trained on a Balloon Analogue Risk Task (BART), an interactive computer task in which subjects try to inflate a virtual balloon as large as possible without bursting it to gain points. tDCS was applied at 2 mA current with two 25 cm2 saline soaked electrode sponges (anode at F4 (over right DLPFC) and cathode at F3 (left DLPFC)) for 25 minutes concurrent with performing the BART. To evaluate generalization, an untrained Risk Task was performed before and after the five days of training. To evaluate durability, the BART and Risk Task were administered again at one and two month follow-up sessions. Fifteen Veterans received active tDCS (mean age 60.4±6.6 years, 1 woman) and 15 received sham tDCS (mean age 58.3±7.6 years, 2 women). All participants self-reported a history of exposure to sub-concussive or concussive events at some point in their lives; 10 participants receiving active tDCS and 8 receiving sham met criteria for having sustained a mild TBI (no significant difference between groups). For the trained BART task, growth curve analysis (GCA) examining individual variation of the growth rates over time showed no significant variations in individual trajectory changes. For the untrained Risk Task, GCA showed that the active tDCS group had a significant 46% decrease in risky choice from pre- to post-intervention, which persisted through the one and two month follow-up sessions. The sham tDCS group showed no significant change in risky choice from pre- to post-intervention. Active tDCS over DLPFC paired with a decision-making task effectively reduced risk-taking behavior in a group of Veterans with clinically-relevant impulsivity. Results suggest that this approach may be an effective non-pharmacological, neuroplasticity-based intervention for patients affected by impulsivity.

Meta-brain computational models for representing cultural diversity

Bradly Alicea (Orthogonal Research and Education Laboratory)
Click here for abstract
With great relevance to the naturalistic study of behavior, the characterization of both the neural substrate and cognitive underpinnings of cultural content remain elusive. One approach is to define a network of neural correlates related to expressiveness of culture. Yet this does not capture the great variation between brains and, more specifically, across cultures. To meet these criteria, we introduce a representational construct called the meta-brain model. The meta-brain model characterizes the structure of collective neural and cognitive activity for specific concepts across individuals in a single context. This allows us to work from a characterization of all possible behaviors, with the most plausible being generated through individual experiences.

Presented here is a computational model of how simple stimuli and complex concepts are represented across brains shaped by cultural context. This particular meta-brain model consists of two parts: 1) a n-dimensional kernel that is used to classify stimuli according to a blend of physical features of the perceptual world and higher-level concepts, and 2) a connectionist model used as a complement to the kernel. Taken together, they allow is to model and even simulate how the conceptual and perceptual world are translated into cultural practices and culturally-mediated behaviors. This includes the representation of both ritual processes and cultural differences in classification of domains such as colors.

The n-dimensional kernel (geometric structure) provides a geometric approach that draws parallels to the work of Roger Shepard, Peter Gardenfors, and Mark Turner. Each geometric structure consists of a soft classification space bounded by all possible values for a given concept or phenomenon. These bounds serve as dimensions, and a single event is represented by a set of scalar values. A series of these scalar values represents a set of observations across a particular cultural population. Different populations exhibiting similar cultural behaviors can be plotted in the same space, and will yield differences in how the same set of factors are subjectively understood.

As a model of behavior rather than representational coding, the connectionist model is meant to represent decision-making based on subjective criteria. This is done using a three-level network with nodes representing cultural structures (top layer), beliefs or epistemic propositions (middle layer), and discrete empirical phenomena (bottom layer). This allows for us to characterize empirical observations according to principles rather than rational criteria, leading to suboptimal (and ultimately more realistic) behaviors.


Discussions can freely continue under the hashtag #brainTC.

ENIGMA & Global Neuroscience – Worldwide studies of 22 brain diseases across 40 Countries

Paul Thompson (University of Southern California)

The ENIGMA Consortium ( is a worldwide scientific alliance that conducts coordinated global studies of brain diseases. Since 2009, ENIGMA published the largest neuroimaging studies of 9 major brain diseases, using harmonized analysis of brain MRI, diffusion imaging, EEG, and resting state functional MRI. 51 dedicated working groups study 22 disorders across psychiatry (schizophrenia, bipolar disorder, and major depression, anxiety, PTSD, suicidality, and substance use disorders), neurology (epilepsy, Parkinson’s disease, brain injury, stroke, ataxia, eating disorders) and neurodevelopment (OCD, ASD, ADHD). ENIGMA’s World Aging Center coordinates brain aging studies across 40 countries, charting brain changes across the lifespan, and performing large scale genome-wide association studies and epigenetic studies of the brain. This talk will highlight some major findings of the consortium, including large-scale studies of major brain diseases.


ENIGMA’s Genomics Workgroups are organized into (1) large scale genome-wide association studies (GWAS), seeking loci that affect brain development, aging and psychiatric disease risk, (2) epigenetic studies, relating genome-wide methylation to brain metrics, and (3) ENIGMA-CNV, discovering effects of rare genetic variants on the brain. Partnering with the CHARGE Consortium and UK Biobank, ENIGMA uses harmonized protocols to compute brain metrics from MRI (e.g. cortical thickness, subcortical volumetrics), DTI, EEG and resting state fMRI. In parallel, ENIGMA’s 22 neurological and psychiatric workgroups process neuroimaging data worldwide, meta-analyzing disease effects, seeking disease modulators in the genome and environment, and predictors of clinical outcomes. Standardized analysis protocols are run remotely by consortium members, who propose secondary analyses of the aggregated data to study factors that affect the brain (e.g., specific medications, childhood trauma, environmental exposure, co-morbidities). Several projects use machine learning for clinical subtyping, outcome prediction, and biomarker discovery. Additional working groups perform some of the largest worldwide imaging studies of brain laterality, evolution, transgender persons, infant brain development, HIV/AIDS, schizotypy, antisocial personality disorder, panic disorder, and early onset psychosis.


Partnering with CHARGE and UK Biobank, ENIGMA analyzed 50,000+ brain scans, revealing over 100 genetic loci associated with brain metrics. The genetic architecture of the human cortex and subcortical regions overlaps with genetic profiles associated with Alzheimer’s, Parkinson’s, and psychiatric risk. Regionally selective effects of disease risk loci suggest possible mechanisms.


Global pooling of imaging data can eludicate genomic, epigenetic and disease effects on the human brain. Cross-disorder studies are charting how 22 brain diseases impact the brain, yielding normative metrics of brain aging, and relating them to key genetic markers and biomarker profiles. In aggregate, ENIGMA addresses the reproducibility crisis in brain research, empowering the study of factors that affect the brain worldwide.

Global and regional white matter development in early childhood

Jess Reynolds (University of Calgary)
Co-authors: Melody Grohs, Deborah Dewey and Catherine Lebel (University of Calgary)
Click here for abstract
White matter development continues throughout childhood and into early adulthood, but few studies have examined early childhood. The specific trajectories and regional variation in this age range remain unclear. The aim of this study was to characterize developmental trajectories and sex differences of white matter in typically developing young children. Three hundred and ninety-six diffusion tensor imaging datasets from 120 children (57 male) were analyzed using tractography. Participants were aged 1.95-6.97 years at intake (mean = 4.04 ± 1.07 years), and had 1-20 scans each (mean time between scans = 7.99 ± 4.63months). Semi-automated tractography was run using ExploreDTI to delineate ten major white matter tracts. Fractional anisotropy (FA) increased and mean diffusivity (MD) decreased in all white matter tracts by 5-15% over the 6-year period, likely reflecting increases in myelination and axonal packing. Males showed steeper slopes in a number of brain areas. Overall, early childhood is associated with substantial development of all white matter and appears to be an important period for the development of occipital and limbic connections, which showed the largest changes. Callosal and pyramidal tracts had high initial maturity and slow rates of change, which likely reflects progression towards a plateau, and is consistent with earlier maturation of these tracts demonstrated in adolescent and early adult literature. The limbic tracts (fornix, cingulum) and most association areas had low initial FA and higher rates of change, reflecting that early childhood is an important period in their development. Frontal-temporal connections showed slower rates of change and initial immaturity, which suggests considerable development is yet to come. This study provides a detailed characterization of age-related white matter changes in early childhood, offering baseline data that can be used to understand cognitive and behavioural development, as well as to identify deviations from normal development in children with various diseases, disorders, or brain injuries.

Establishing function of brain signals using a pattern learning task during MEG and pupillometry

Silvia Isabella (University of Toronto; Hospital for Sick Children)
Co-authors: Douglas Cheyne (University of Toronto; Hospital for Sick Children)
Click here for abstract
Although studies involving use of cognitive control demonstrate changes in brain activity, the functional roles of these signals remain unknown. The objective of this study is to characterize brain oscillations related to increases in cognitive and motor demands. We hypothesize that frontal θ and sensorimotor γ power will increase with cognitive effort (speed + load) required by the task (as measured by reaction time, RT, and pupil diameter, PD) while post-movement β rebound (PMBR) will decrease with effort. We measured neuromagnetic (MEG) brain activity in 16 right-handed healthy adults performing 6 blocks of a go/switch task (Switch=25%). We manipulated stimulus predictability using fixed stimulus sequences (P=90%, d=10%) that were unknown to the participants. There was a main effect of response hand (Go/Switch) and pattern (all p<0.01) on RT and PD. Given that PD was more sensitive than RT to the task parameters, subjects likely increased cognitive load (PD) to reduce differences in performance (RT). We found that θ activity was more sensitive to the pattern (p<0.001) than to the Switch response (p<0.02), indicating a role in surprise detection. Furthermore, PMBR and γ increased with cognitive load, but did not correlate with RT or PD, indicating a role in integrating cognitive and motor parameters together. This study is the first to distinguish these functions for θ, PMBR andγ, and demonstrates a role for frontal activity in surprise detection over cognitive control, and a role for sensorimotor cortex in integrating cognitive and motor control.

Comparison of MEG functional connectivity measures for detection of Alzheimer’s disease

Han Bao (McGovern Institute for Brain Research at MIT; Department of Radiology at Harvard Medical School)
Co-authors: Yasaman Bagherzadeh (McGovern Institute for Brain Research, Massachusetts Institute of Technology), David Lopez, Pilar Garces and Fernando Maestu (Department of Experimental Psychology, Complutense University of Madrid), Quanzheng Li (Department of Radiology, Harvard Medical School), Dimitrios Pantazis (McGovern Institute for Brain Research, Massachusetts Institute of Technology)
Click here for abstract
Magnetoencephalography offers high temporal resolution and captures a direct measure of intraneuronal electrical activity instead of indirect hemodynamic responses, making it uniquely suited to study the subtle changes of large-scale functional brain networks associated with the pathological cascade of Alzheimer’s Disease (AD). Synaptic dysfunction is a well-established finding in AD, however it remains unclear which MEG functional connectivity measures are more sensitive to capture pathological patterns associated by AD. In this work, we investigated different connectivity measures both in the sensor-level and source-level to identify which biomarkers best predict AD.

We used MEG data from 78 prodromal AD patients and 54 age-matched clinically normal subjects from the Madrid cohort database. MEG resting-state data were band-pass filtered in different canonical frequency bands and then split into 2s epochs. Dynamic statistical parametric mapping was used to map sensor data on cortex and then different connectivity measures in both sensor-level and source-level brain networks were constructed, including power correlation; amplitude correlation; coherence connectivity; cross-spectral density; phase-locking value; weighted pairwise phase consistency; orthogonal correlation; and mutual information. For each connectivity measure, an artificial neural network was trained to discriminate prodromal AD patients from normal subjects using features selected from a Kolmogorov-Smirnov test. Results were validated by using Monte Carlo cross-validation.

Our findings showed that sensor-level data resulted in higher classification performance than source-level data. However, all connectivity measures could predict prodromal AD from normal subjects. Mutual information was the most reliable measure with classification accuracy ~90%.

Taken together, our results show that selection of MEG connectivity measures plays a critical role in identifying biomarkers for AD, and motivate future studies that explore the robustness of these measures in cross-site investigations.

Attenuating anger and aggression with neuromodulation of the vmPFC: a simultaneous tDCS-fMRI study

Gadi Gilam (Systems Neuroscience and Pain Laboratory, Division of Pain Medicine, Department of Anesthesiology, Perioperative and Pain Medicine, Stanford University School of Medicine)
Co-authors: Rany Abend (Section on Development and Affective Neuroscience), Guy Gurevitch, Alon Erdman, Halen Baker, Ziv Ben-Zion and Talma Hendler (Tel Aviv Center for Brain Function; Wohl Institute for Advanced Imaging; Tel Aviv Sourasky Medical Center)
Click here for abstract
Angry outbursts during interpersonal provocations may lead to violence and prevails in numerous pathological conditions. In the anger-infused Ultimatum Game (aiUG), unfair monetary offers accompanied by written provocations induce anger. Rejection of such offers relates to aggression, whereas acceptance to anger regulation. We previously demonstrated the involvement of the ventro-medial prefrontal cortex (vmPFC) in accepting unfair offers and attenuating anger during an aiUG, suggestive of its role in anger regulation. Here, we aimed to enhance anger regulation by facilitating vmPFC activity during anger induction, using anodal transcranial direct current stimulation (tDCS) and simultaneously with functional Magnetic Resonance Imaging to validate modulation of vmPFC activity. In a cross-over, sham-controlled, double-blind study, participants (N=25; 15 females, age=26.16±3.63) were each scanned twice, counterbalancing sham and active tDCS applied during adinistration of the aiUG. Outcome measures included the effect of active vs. sham stimulation on vmPFC activity, unfair offers’ acceptance rates, self-reported anger, and aggressive behavior in a subsequent reactive aggression paradigm. Results indicate that active stimulation led to increased vmPFC activity (coupled by changes in ACC and insula activity) during the processing of unfair offers, increased acceptance rates of these offers, and mitigated the increase in self-reported anger following the aiUG. Notably, stimulation did not influence other emotion categories (fear, sadness, and happiness). We also noted a decrease in subsequent aggressive behavior following active stimulation, but only when active stimulation was conducted in the first experimental session. Finally, an exploratory finding indicated that participants with a stronger habitual tendency to use suppression as an emotion regulation strategy, reported less anger following the aiUG in the active compared to sham stimulation conditions. Findings support a potential causal link between vmPFC functionality and the experience and expression of anger, supporting vmPFC's role in anger regulation, and providing a promising avenue for reducing angry and aggressive outbursts during interpersonal provocations in various psychiatric and medical conditions.

Familiarity and implicit memory, an EEG study.

@G4_gul (Wilfrid Laurier University)
Co-authors: Jeffery Jones (Wilfrid Laurier University)
Click here for abstract
At present, there is a great controversy regarding how familiarity operates in order to support recognition. Despite much work regarding familiarity, it is unclear whether familiarity is an expression of explicit memory or implicit memory. In our present study, we tested several hypotheses regarding the cognitive processes reflected in the frontal FN400 ERP potentials that is elicited in the time window (300-500 ms) and fluency, which is linked to perceptual fluency and elicited at left parietal sites during the time window (200-400 ms) post-stimulus. One, the FN400 component reflects familiarity elicited by the stimulus that emanates from recent exposure or repetition. Two, the FN400 component reflects the associated conceptual implicit memory initiated by the stimulus, which may or may not emanate from recent exposure. Three, fluency effect is independent of old/new effect and it reflects perceptual fluency elicited by the stimulus. Four, whether there is an association between the FN400 effect elicited during the first exposure (during the encoding phase) and during test phase (recognition). The present study aims to extend the Gul & Jones (2018) findings using different stimuli but same methodology i.e. meaningless novel stimuli (fractals) were used instead of the pictures of common objects (Gul & Jones, 2018). Our ERP results suggest that not only the fluency effect was independent of the repetition but also the FN400 effect. Our ERP results suggest that the FN400 potentials, which were driven by conceptual implicit memory, correlated with the behavioral indicators of recognition whereas the recollection effect was absent. Thus, our behavioral and ERP results suggest that neural correlates of conceptual implicit memory process can influence the decisions driven by explicit memory. Moreover, fluency effect plays a strong role in setting up the encoding strategies and together with FN400 effect these two processes support recognition.

For the program of #BrainTC 2018, see here.