Session III, 30 June 2026, 12.00-13.15 (CEST)

Major Depressive Disorder (MDD) is a leading cause of disability worldwide, frequently characterized by anhedonia and deficits in subjective energy. Existing therapeutic approaches are often associated with high barriers to access, stigma, and long waiting times. To promote an easily accessible and individualized therapy approach, the MASE project investigates how a non-exercise activity (NEA; i.e., fast walking) intervention can increase subjective energy levels and lead to a mid-term decrease in depressive symptoms. We aim to recruit 180 participants (acute MDD, remitted MDD and healthy controls; n = 60 each). For the 5-week intervention we set up a within-person encouragement design (WPED) with Ecological Momentary Assessment and movement acceleration sensors. These ambulatory assessment methods offer real-time data acquisition, effectively overcoming limitations of retrospective bias and low ecological validity. The micro-intervention is individually triggered by a participant’s inactivity, i.e. after 30 minutes of inactivity. The prompts to engage in the NEA task or, in an active control condition, to engage in a working memory game are randomized. Momentary state, such as mood and energy levels are assessed repeatedly throughout the day to capture real-time changes. To maximize adherence, intervention triggers will be limited to a maximum of four per day. To study neurobiological moderation, we perform a magnetic resonance imaging scan at baseline to examine the volume of the subgenual anterior cingulate cortex (sgACC). This is a key region implicated in affect regulation and moderates momentary energetic arousal. We hypothesize that sgACC volume will predict the individual susceptibility to NEA-induced increases in subjective energy. For the assessment of long-term effects, additional follow-ups will take place at one month, three months, and six months post-intervention. Recruitment began in September 2025 and is ongoing. Preliminary results of the first enrolled 30 participants will be presented, focusing on adherence and initial trends in NEA - induced momentary changes in subjective energy. We anticipate that these targeted momentary micro-interventions will enable more individualized mental health strategies and foster both primary and secondary prevention of MDD.

Cortical structural alterations are consistently reported across substance use disorders (SUDs), showing shared patterns across different substances and overlapping with abnormalities observed in major psychiatric disorders. Early life adversity can disrupt neurodevelopment, increasing vulnerability to later substance use and contributing to structural deviations that persist into adulthood. Also, the neurotoxic effects of substances may induce cortical changes associated with cumulative exposure and addiction severity. However, it remains unclear how substance-related morphological alterations are organized within large-scale brain networks, and how they relate to early substance exposure and shared neurodevelopmental vulnerability. Here, we investigate the joint embedding of substance-related cortical alteration networks to identify common organizational principles across substances, and examine their correspondence with co-alteration patterns observed in other neuropsychiatric conditions and during early substance use. We analyzed two large cohorts spanning adolescence and adulthood. Adult substance-related alterations were examined using the ENIGMA-SUD dataset, including 4,332 individuals with SUD and 2,756 healthy controls across six substance categories. Adolescent substance effects were assessed in the ABCD cohort, including substance-naive participants and those who initiated substance use between baseline and four-year follow-up. For each substance, case–control cortical thickness effect sizes were estimated across regions, followed by dimensionality reduction to derive low-dimensional embeddings of cortical structural covariance networks. We constructed co-alteration matrices based on the covariance of effect-size maps across substances and across major psychiatric disorders, and applied gradient analysis to extract shared latent dimensions. Spatial correlations were used to assess correspondence among gradients. To assess the impact of early substance use on cortical development, linear mixed-effects models were applied across three ABCD time points to test for baseline differences and altered age-related thinning associated with substance initiation, and cortical thickness change from baseline to four-year follow-up was quantified to compare overall atrophy between substance-initiation and substance-naive adolescents. Finally, ABCD-derived effect maps were correlated with ENIGMA SUD cortical gradients to assess developmental correspondence.

Background: Sleep and physical activity (PA) are bidirectionally linked and have neuroprotective effects. Both affect neurometabolism and cerebral blood flow (CBF), which can be non-invasively quantified with magnetic resonance spectroscopy (MRS) and arterial spin labelling (ASL), respectively. N-Acetylaspartate (NAA) and Glutamate-Glutamine (Glx) are key metabolites associated with sleep and PA. The objective of our study is to examine how particular characteristics of sleep and PA relate to brain physiology, to determine the differential contributions of day-to-day variation versus aggregated measures of sleep and PA, to clarify how NAA, Glx, and CBF are related, and to examine how PA and sleep are related in influencing brain physiology. Methods: Healthy adults (N = 28; age range: 22–43 years) participated in repeated MRI sessions across two consecutive days. Scan times were counterbalanced between morning (08:00) and afternoon (14:00). Point-resolved spectroscopy sequences were used to acquire MRS data from the anterior and posterior cingulate cortices (ACC and PCC). Regional CBF was assessed using pseudo-continuous ASL. The Oura ring continuously monitored total sleep time, sleep efficiency, midpoint of sleep, moderate-to-vigorous physical activity (MVPA), and sedentary time. Participants wore the Oura ring for seven days prior to the first imaging session. Planned analyses: Primary outcomes include concentrations of total NAA, Glx, and CBF in the ACC and PCC. Intra-subject time-of-day effects will be evaluated by comparing neuroimaging measures between morning and afternoon sessions. The independent effects of PA and sleep characteristics will be assessed using correlations. Multiple regression models will examine associations between the metabolites of interest and CBF and compare models that combine sleep and PA characteristics to determine which best explain variability in these outcomes. Exploratory analyses will investigate associations between daily variations in sleep and PA measures and neuroimaging outcomes. Expected Results: Greater MVPA time is expected to be associated with higher levels of NAA and Glx, as well as increased CBF. In contrast, poor sleep is anticipated to have an opposing effect. This study provides a methodological basis for subsequent research assessing the influence of sleep and PA patterns on brain metabolism and CBF in both healthy and clinical populations. Final results will be presented at the conference.

Cardiac responses to central autonomic network (CAN) regions may be reflected in every heartbeat, but the co-occurrence in the brain remains unclear. In this study, we combine fMRI with ECG to explore the role of the CAN in the dynamics of heart-brain interaction. We analyzed participates with both ECG and fMRI scanning in the LIFE dataset. Subjects were without neurologic disease (N = 1100, 507 F; M=69.79; SD=4.63). Modularity is used to quantify the within-module connection densities. Degree centrality (DC) is used to evaluate the significance of CAN regions for the network integrity. CAN components are selected based on a meta-analysis of autonomic correlates. We used the modularity of the CAN to discriminate between two groups with a median cutoff and performed permutation-based t-tests to detect the ECG clusters that show significant differences. As a result, the elevation of PQ and ST segments in the inferior direction is linked to lower modularity of CAN. QT intervals are significantly positively correlated with the DC of the pregenual anterior cingulate cortex (pgACC) (r=0.082, p=0.007). Specific brain connections that related to QT intervals including: 1. negative correlations with ‘the left cerebral arterial territories-amygdala, hypothalamus, and hippocampal formation connections’, ‘the left thalamus-right angular gyrus and supramarginal gyrus connections’, ‘the left parahippocampal gyrus-right frontinsular cortex connections’; 2. positive correlations with ‘the right frontal pole-the pregenual anterior cingulate cortex connections’, ‘the right frontal pole-ventral posterior cingulate cortex, precuneus cortex, and lingual gyrus connections’. No significant brain network correlations were observed for the averaged P wave duration, PR intervals, or QRS duration. The findings indicate the role of the CAN regions in the heart-brain interaction.

Effective locomotor control relies on a fundamental system of orientation relative to the environment [1]. Previous research has shown that gaze behavior plays an important role in orientation, and directing gaze toward a distant location has been reported to stabilize postural control in locomotion [2]. However, it remains unclear whether such gaze strategies facilitate the learning of self-localization in novel locomotor tasks in which coordination has not yet been established. The purpose of the present study was to examine whether gaze guidance promotes the acquisition of self-localization during locomotor learning. We hypothesized that a group instructed to direct gaze toward a distant location would show faster improvement in locomotor learning compared with a control group. To test this hypothesis, a reverse bicycle task was used [3], in which the relationship between handlebar movement and front-wheel direction was inverted, requiring the acquisition of a novel coordination pattern. Practice was conducted for 20 min per day over four days, and a test was administered after each practice session to assess learning progress. Twenty-four participants (age: 24.9 ± 2.7 years; height: 173.3 ± 5.7 cm; 21 males, 3 females) were randomly assigned to either a gaze instruction (EXP) group or a control (CON) group. The gaze instruction group (n = 12) was instructed to fix their gaze on a distant point during practice, whereas the control group (n = 12) received no specific gaze instructions. A linear mixed-effects model was used for analysis, with distance traveled without foot contact as the dependent variable, and group (CON vs. EXP) and time (Pre, Day 1–4) as fixed effects, with participants included as a random effect. A significant group × time interaction was observed. Compared with the pre-test, performance improvements in the CON group were significantly greater on Day 2 (p=.025), Day 3 (p< .001), and Day 4 (p<001) than those in the EXP group. Contrary to the hypothesis, gaze instruction toward a distant location delayed the learning of coordinated locomotor control. One possible explanation is that directing gaze toward a distant point does not necessarily mean that attention is allocated to that location [4]. Future studies should investigate changes in neural activity across different phases of practice, in order to clarify the forms of attentional engagement that contribute to motor learning.

References

[1] Reed E. S. (1982). An outline of a theory of action systems. Journal of Motor Behavior, 14(2), 98–134. https://doi.org/10.1080/00222895.1982.10735267

[2] Kojima, T., & Kokubu, M. (2025). The effect of instructions regarding gaze direction on stability of movements and accuracy of trajectory control during cycling. Experimental Brain Research, 244(1), 21. https://doi.org/10.1007/s00221-025-07213-6

[3] Magnard, J., Macaulay, T. R., Schroeder, E. T., Laine, C., Gordon, J., & Schweighofer, N. (2024). Initial development of skill with a reversed bicycle and a case series of experienced riders. Scientific Reports, 14(1), 4334. https://doi.org/10.1038/s41598-024-54595-8

[4] Posner, M. I. (1980). Orienting of attention. Quarterly Journal of Experimental Psychology, 32(1), 3– 25. https://doi.org/10.1080/00335558008248231

Introduction: Exposure to trauma is highly prevalent and poses a significant risk for psychopathology. Despite the negative effects trauma can bring, some research suggest that trauma may, under certain conditions, be associated with increased prosocial behaviour. The relationship between trauma and prosocial behaviour remains poorly understood, with existing findings scattered across disciplines. Examining the relationship between trauma and prosocial behaviour is therefore critical for advancing theories of adaptation and resilience following adversity. Methods: A PRISMA-compliant, pre-registered meta-analysis was conducted using studies identified from Web of Science, PubMed, and PsycArticles, supplemented by citation searches. Two reviewers independently screened studies and extracted data. Effect sizes were extracted at the independent-sample level and categorized by trauma type, assessment method, and prosocial behaviour measure. Correlational effects were transformed to Fisher’s z and categorical effects to Hedges’ g. Multilevel random-effects models with robust variance estimation were conducted in R. Study quality and publication bias were assessed, and sensitivity analyses supported the robustness of findings. Results: 39 articles comprising 141 effect sizes were included. No overall association between trauma exposure and prosocial behaviour was observed. Subgroup analyses indicated a significant negative association between childhood trauma and prosocial behaviour in correlational studies (β ̂ = −0.14, p = .034) and a trend-level effect in all studies (β ̂ = −0.12, p = .058). Interpersonal trauma was also associated with lower prosocial behaviour at the trend level (β ̂ = −0.10, p = .084). Egger’s test indicated significant publication bias (t = 3.40, p < .001). Discussion: These findings suggest that there is currently no clear evidence for a relationship between trauma exposure and prosocial behaviour. Given the conceptual and methodological heterogeneity of both trauma samples and prosocial behaviour measures, it is possible that no generic answer can be found to conclude such a relationship. Existing research has also disproportionately focused on childhood interpersonal trauma, with limited attention to other trauma types and forms of prosocial behaviour. Future research should more clearly differentiate trauma subtypes and prosocial domains and employ more objective and standardized measures to clarify this complex relationship.

Music- and rhythm-based interventions hold promise for neuropsychological rehabilitation, yet comprehensive studies in individuals with acquired brain injury remain scarce. This study investigated the efficacy of a digital rhythm-based training program on executive functions and rhythmic abilities in participants with structural brain lesions. In a randomized controlled trial, adults with an acquired brain injury (>3 months after injury) and deficits in executive functions or attention, were assigned to an intervention group or a waitlist-control group. Based on power analysis, we recruited 66 participants. The 4-week home-based intervention consisted of training with the "Rhythm Workers" app (five sessions per week, 20–30 minutes) and weekly brief phone support. The game required finger-tapping in synchrony with musical beats, with performance accuracy determining progression through the game. Primary outcomes included executive functions (cognitive flexibility, inhibitory control, working memory) and rhythmic abilities (synchronization consistency, rhythmic-motor variability, beat perception), assessed at pre-, post-intervention, and three-month follow-up. Secondary outcomes included treatment adherence, game acceptance, satisfaction, and subjective ratings of relevant domains. Preliminary data show high adherence to the protocol, with treatment satisfaction rated moderate to high. Interim analyses show improvements in rhythm-production tasks, whereas beat-perception remained unchanged. Final analyses will clarify the intervention’s effects on executive functions and their maintenance over time. Exploratory analyses will examine demographic, clinical, and intervention-related moderators, including lesion characteristics. Overall, the study is expected to provide robust evidence for rhythm-based digital therapies in neurocognitive rehabilitation and to guide the optimization of interventions for individuals with acquired brain injuries.

Autism Spectrum Disorder (ASD) has been associated with speech-processing deficits, potentially linked to disruptions in excitatory/inhibitory (E/I) balance. This study investigated the relationship between E/I balance and neural tracking of speech in autistic and non-autistic children and adolescents. Using resting-state EEG, E/I balance was estimated, while speech-brain coherence was measured during naturalistic speech listening across intonational, syllabic, and phonemic frequency bands. The sample included 124 participants (62 autistic, 62 non-autistic), matched on age, sex, and non-verbal IQ. Results showed that phonemic coherence significantly predicted E/I balance in non-autistic participants, with higher coherence associated with more inhibitory-leaning activity. This relationship was not observed in autistic participants. Additionally, intonational coherence interacted with age in predicting E/I balance across both groups, suggesting a developmental link. These findings highlight differences in neural mechanisms underlying speech processing between autistic and non-autistic participants and suggest that developmental trajectories of E/I balance are linked to intonational speech tracking.

Accurate identification of brain states and pathological transitions remains a central challenge in clinical neuroscience, particularly in epilepsy management. Current neuroimaging modalities provide complementary but incomplete solutions: electroencephalography (EEG) offers high temporal resolution but poor spatial localization, while functional magnetic resonance imaging achieves fine spatial detail at the cost of portability and temporal precision (1). Functional ultrasound imaging (fUSI) has emerged as a promising alternative, offering depth-resolved hemodynamic maps with spatial resolution of 200–400 μm, millisecond-scale sensitivity, and bedside portability, without contrast agents or invasive procedures (2). This work investigates the potential of fUSI for classifying epileptic brain states, with particular focus on the preictal phase, i. e. the transitional period preceding seizure onset. Clinically critical yet notoriously difficult to detect, this phase can last minutes to over an hour with high inter- and intra-patient variability. Reliable identification could enable timely therapeutic intervention and advance real-time seizure forecasting. Our framework integrates simulated and real electrophysiological data. Synthetic neural signals simulating interictal, preictal, and ictal states were generated using an excitatory-inhibitory population model. Real intracranial EEG recordings were drawn from the HUP dataset (3) (34 drug-resistant epilepsy patients), preprocessed at 500 Hz, and segmented into 3-second windows. Synthetic fUSI signals were derived through convolution with empirically established hemodynamic response functions (4). Classification used a 1D convolutional neural network with three convolutional layers, ReLU activations, and max-pooling. On simulated data, fUSI achieved accuracy up to 0.86 for preictal detection; on real recordings, performance remained stable (~0.78–0.79), comparable to sEEG-based models. These results suggest that fUSI-derived vascular signals carry sufficient information to discriminate epileptic brain states, including the elusive preictal phase. Future work will incorporate multimodal EEG–fUSI fusion, larger cohorts, and experimental validation toward robust clinical seizure prediction. (1) ABREU, R.; LEAL, A.; FIGUEIREDO, P., Frontiers in human neuroscience, 2018 (2) DEFFIEUX, T., et al. Current opinion in neurobiology, 2018 (3) BERNABEI, J. M., et al. Brain, 2022 (4)NUNEZ-ELIZALDE, A. O., et al. Neuron, 2022.

The hippocampus and amygdala are medial temporal lobe structures central to cognition and emotion, yet it remains unclear where their cortical connectivity profiles are distinct versus overlapping and how they are embedded within large-scale cortical organization. Because both contain multiple subfields and nuclei, subregion-resolved mapping is needed. Here, we mapped hippocampal and amygdala subregion functional connectivity to the isocortex and contrasted Pearson correlation with graphical lasso (GLASSO). We analyzed 1 hour of resting-state fMRI per subject from HCP-Young Adult (N = 722; 380 females). The hippocampus was segmented with HippUnfold (30 parcels) and the amygdala with the Saygin atlas (18 nuclei), yielding 48 seeds. Targets were 358 Glasser parcels excluding hippocampal cortex. Connectivity was estimated with Pearson and GLASSO, Fisher-z transformed, and averaged across subjects. We applied diffusion map embedding to a seed-seed affinity matrix to estimate joint gradients and projected these gradients to cortex using top-union 10% masks. We also derived dominance and sharedness metrics. Top-10% seed-cortex connectivity revealed hippocampus- and amygdala-biased territories as well as shared hubs, indicating graded dissociation. Hippocampal subregions preferentially coupled with medial temporal, cingulate, sensory, and visual cortex, whereas amygdala nuclei preferentially targeted anterior insula, orbitomedial frontal, superior temporal, midcingulate, lateral occipitotemporal, and motor cortices. GLASSO sharedness peaked in anterior temporal and medial prefrontal/paralimbic cortex, whereas Pearson sharedness was broader and shifted toward somatomotor cortex. GLASSO gradients revealed a principal hippocampus-amygdala axis, a left-right lateralization axis, and a within-BLA ordering in which paralaminar was most hippocampus-like. Overall, Pearson captured broader co-fluctuation structure, whereas GLASSO isolated a sparser, more interpretable backbone.

Background: Transcranial magnetic stimulation (TMS) motor mapping increasingly relies on electric field (E-field) modeling to localize cortical targets with high precision. However, commonly used sampling schemes can generate highly correlated E-field patterns, leading to redundant measurements and inefficient mapping sessions. Objective: To develop and validate a unified prospective sequence-design framework that selects coil placements to maximize information gain per pulse in E-field-informed motor mapping. Methods: We formulated coil position--orientation selection as a modular subset-selection problem parameterized by three design knobs (pattern representation, similarity metric, and set-level aggregator). We compared distance-based objectives (including farthest point sampling) against determinant-based volume objectives (D-optimality), including a kernelized (RBF) variant to address low-rank structure in feasible E-field libraries. Performance was evaluated in virtual mapping simulations on 12 subject-specific head models with four hotspot contexts (crown, wall, fundus, and periphery), using geodesic distance (GD) and normalized RMSD (NRMSD) trajectories and AUC. Results: Prospective objectives accelerated convergence relative to random sampling, but their benefits depended on how diversity was defined. Linear D-optimal selection reduced stimulus-wise redundancy yet exhibited a saturation-and-rebound regime in element-wise separability that coincided with the low effective rank of the E-field libraries. Kernelizing the volume objective mitigated this rebound and improved late-stage stability of map-shape recovery. Conclusion: Determinant-based sequence design provides a statistically grounded route to improving sample efficiency in E-field-informed motor mapping. Accounting for the low-rank structure of feasible E-fields via kernelized similarity can stabilize mapping as sequences grow, yielding practical guidance for prospective stimulation planning.

Cooperative social dynamics typically involve more than two individuals, yet most investigations of Interpersonal Neural Synchronization (INS) have focused on dyadic interactions. As a result, the neural dynamics underlying group-level interactions remain poorly understood. To address this gap, we adopted a hyperscanning EEG paradigm during a cooperative group task in a naturalistic setting. Specifically, triads of participants were invited to discuss and find shared solutions to hypothetical situations provided by us, while their gaze behaviors and their cortical activity were recorded, using three portable EEG systems, simultaneously. Phase synchronization measures were computed to target group and subgroup neural synchrony. Moreover, turn-taking metrics were extracted from triads’ verbal exchanges to quantify their behavioral alignment. Here, we aimed at unraveling how INS and behavioral alignment change over time through three objectives: i) to quantify INS and behavioral alignment during joint cooperative group interactions; ii) if present, to understand whether and how both neural and behavioral synchrony occur at the group and subgroup level; and iii) to explore the relationship between neural synchronization and behavioral alignment. Overall, we observed that behavioral dynamics differentially modulate group- and subgroup-level neural synchrony during cooperative triadic interactions. Additionally, as participants reach a common agreement on their discussion, neural synchronization tends to align with behavioral one, suggesting a temporal alignment of neural and communicative processes. Taken together, these results highlight the importance of a shared and not polarized goal for promoting interpersonal synchronization.

The rapid integration of AI-based conversational agents (AICAs) into healthcare raises a fundamental question: do people respond to AI medical advice differently than to human advice, and if so, what are the underlying behavioral and neural mechanisms? This project addresses these questions through a multi-study programme spanning behavioral experiments and functional neuroimaging. Four online experiments (total N = 589) demonstrated that AICAs exert domain-specific social influence on health decision-making, modulated by perceived expertise. Thirty-four participants underwent 3T fMRI while receiving health advice from either a human or an AI agent across four counterbalanced runs (96 trials per participant). Trial-wise beta maps were estimated using a least-squares separate (LSS) approach. To decode agent identity from whole-brain activation patterns, we will implement a Vision-Language Model (VLM) pipeline: beta map volumes are fed into a 3D Vision Transformer (ViT) encoder to extract high-dimensional spatial features, which are then aligned to a language model embedding space via an HR-LR cross-attention projection layer that fuses global volumetric context with high-resolution anatomical detail, and finally processed by an LLM backbone to classify Human versus AI advisory conditions. We predict that the most discriminative signal to be concentrated in regions associated with social cognition and source evaluation are specifically the temporoparietal junction (TPJ) and medial prefrontal cortex (mPFC), which support agency attribution and belief updating, as well as the anterior insula, implicated in integrating interoceptive signals with external advisory cues. We further predict that the cross-attention projection layer will assign disproportionate weight to prefrontal and parietal tokens, consistent with these regions serving as the neural locus of human–AI advisory differentiation. Behavioral and neural evidence have implications for the responsible integration of AI into clinical and public health communication, and demonstrate the feasibility of applying large-scale vision-language architectures to the decoding of socially-relevant neural representations.

The classroom social environment plays a key role in shaping learning processes, as interactions among students and teachers promote shared attention, engagement, and collective understanding. However, the neural mechanisms underlying these social dynamics in real educational settings remain poorly understood. Traditional single-brain approaches focus on individual cognition and do not capture the reciprocal neural processes that characterize social learning. This study analyzes EEG hyperscanning data recorded from high school students during real-world biology lessons to investigate how shared attentional states influence learning outcomes. The dataset includes neural activity collected across different teaching formats, enabling the comparison of interactive and passive learning conditions. We propose a statistically controlled hyperscanning analysis pipeline designed to isolate genuine inter-brain synchronization from spurious correlations induced by shared sensory input. To mitigate correlations arising from environmental noise or coincidental phase alignment, inter-brain synchrony was estimated using the Circular Correlation Coefficient (CCor), a robust estimator of phase coupling. The statistical significance of the observed inter-brain synchrony (IBS) was assessed through a phase-shuffling procedure that generated a null distribution of surrogate datasets by disrupting temporal relationships while preserving signal structure. Only synchronization values exceeding the 95th percentile of this distribution were retained. Additionally, cognitive indices—Engagement, Attention, Pleasantness, and Memory—were derived from EEG spectral power to characterize students' mental states. Statistical differences in both CCor values and cognitive indices across teaching styles were assessed using Linear Mixed Models (LMM) to account for the hierarchical structure of the data. Our results show that interactive teaching significantly increases IBS, particularly in the alpha band, compared to passive listening. Higher synchronization levels correlate with improved learning performance and shared attention, while lower IBS is associated with reduced engagement. These findings prove that classroom learning involves measurable neural alignment. Inter-brain Synchrony (IBS) serves as an objective marker to evaluate teaching strategies and optimize knowledge transfer, providing evidence-based insights to design pedagogical environments that enhance collective attention.

Characterization of brain tissue microstructure provides important insights into fundamental biological processes in the brain and therefore represents a major focus of neuroimaging research. Quantitative MRI (qMRI) methods measure voxel-wise biophysical properties and, in combination with biophysical modeling, enable the investigation of underlying tissue architecture and microstructural properties. One such quantitative technique is Quantitative Susceptibility Mapping (QSM), which estimates magnetic susceptibility. Magnetic susceptibility describes the degree to which a material becomes magnetized in an external magnetic field and is considered an important biomarker for myelin and iron in the brain. Within each voxel, the bulk susceptibility represents the sum of all microscopic and cellular contributions. It is influenced not only by tissue composition but also by microstructural orientation, which can introduce anisotropic behavior. This anisotropy is mainly attributed to the radially organized myelin lipid sheath surrounding axons. Although both gray and white matter exhibit a high accumulation of non-heme iron-containing molecules with strong paramagnetic effects, their contribution to susceptibility anisotropy remains insufficiently studied and is not yet fully understood. This project investigates the anisotropic behavior of tissue susceptibility in order to assess its sensitivity to microstructural features and to improve the reconstruction of white-matter architecture and fiber connectivity. It provides a conceptual overview of the planned research and the methodological approach.

Linking brain chemistry to cognitive control is an important goal in cognitive neuroscience. Proton magnetic resonance spectroscopy (MRS) can be used to measure glutamate in the brain. However, for individual-differences research, it is important that these measures are stable over time, and that the relationship with behaviour is consistent. This pilot study tests whether glutamate levels in the anterior cingulate cortex (ACC) can be measured reliably, and whether they can be meaningfully linked to cognitive- control performance. Participants, including individuals with ADHD and healthy controls, will complete three 3T MRI sessions. In each session, ACC glutamate will be measured using single-voxel MRS, and participants will perform a Flanker task. A healthy control group serves as a reference for typical stability, which allows us to interpret differences in ADHD. Symptoms will be assessed dimensionally, focusing on inattention and hyperactivity-impulsivity. We will also examine whether differences are due to data quality or reflect meaningful variation. In short, this pilot focuses on whether glutamate, behaviour, and their relationship are stable across sessions. This will help evaluate the feasibility of using ACC glutamate to study individual differences in cognitive control and will inform the design of future studies.

Early-life exposure to an adverse nutritional environment is increasingly recognized as a critical factor influencing neurodevelopmental outcomes; however, sex-dependent behavioral and molecular mechanisms remain incompletely understood. This study investigated whether maternal exposure to a Western diet (WD) during gestation and lactation induces behavioral alterations in offspring are associated with region and age-specific molecular dysregulation in the brain at specific time points. Female Wistar Han rats were fed either a control diet or a (WD) throughout gestation and lactation (14 weeks total). After weaning, both male and female offspring were maintained on a control diet and assessed at postnatal days (PND) 30, 60, and 90. Behavioral phenotyping included the open field, elevated zero maze, novel object recognition, self-grooming, marble burying, and social interaction ,locomotor activity tests. Molecular mechanisms were examined using RNA sequencing of the hippocampus (HIP) and prefrontal cortex (PFC), followed by protein-level validation of selected top regulated gene using ELISA. Maternal WD exposure increased body weight and adiposity in offspring and induced prominent behavioral alterations at early developmental stages. At PND30, WD-exposed male offspring showed significant changes in locomotor activity, anxiety-related behaviors, grooming, and social interaction, with distinct sex-specific patterns. These behavioral alterations were less pronounced or absent at later developmental stages . Molecular analyses revealed (HIP) alterations in male WD-exposed offspring, with EDN1 dysregulation at PND30, and KLF2, AGTRAP, VEGF, and eNOS changes at PND60 but in (PFC), whereas eNOS changes were seen at both PND30 and PND60 IN (HIP) and VEGF dysregulation was noted at PND60 in (HIP) in female WD offspring. To sum up, these findings suggest that maternal exposure to a (WD) leads to early, sex-specific behavioral dysregulation, accompanied by significant molecular changes in brain regions essential for cognition and social behavior, thereby reinforcing the notion of early-life nutritional programming as a key factor in neurodevelopmental susceptibility.

Proprioception, our “6th sense", enables us to perceive limb position and movement without vision of our body (Proske & Gandevia, 2012). Recent evidence suggests proprioceptive acuity is trainable in adults, with improvements of 23-43% observed after single training sessions (Seo et al., 2023). Crucially, enhanced proprioception transfers to improved motor execution (Darainy et al., 2013; Vahdat et al., 2019), but whether it enhances motor imagery, the internal simulation of movement, remains unknown. This gap is particularly intriguing given that motor imagery lacks the bottom-up proprioceptive feedback present during actual movement (Jeannerod, 2001). Approximately 20% of brain-computer interface users struggle with "BCI illiteracy" (Allison & Neuper, 2010), potentially due to poor motor imagery ability. Could proprioceptive training, by strengthening sensory representations, improve motor imagery and address this challenge? We investigate whether single-session proprioceptive discrimination training improves motor imagery performance in healthy adults. Sixty participants undergo passive arm movements via the KINARM robot, discriminating left-versus-right relative to body midline. One group receives accuracy feedback, the control group does not. Pre- and post-training assessments include standardised proprioceptive tests (Arm Position Matching, Arm Movement Matching), goal-directed reaching tasks, and mental chronometry comparing executed and imagined movement times. Continuous EEG and ECG recordings capture neurophysiological changes. We will use EEG recordings to investigate the neural mechanisms by which proprioceptive training enhances motor imagery. Findings hold promise for neurorehabilitation, BCI applications, and will advance knowledge on sensory integration.

Electroencephalography (EEG) stands out for its millisecond-level temporal resolution, making it highly sensitive to fast neuronal events. However, EEG source localization is a notoriously ill-posed inverse problem, as many different source configurations can produce virtually indistinguishable scalp potentials. This challenge is significantly magnified for deep structures, such as the hippocampus, whose signals undergo stronger attenuation and spatial blurring before reaching the scalp. While classical linear inverse solvers (e.g., MNE [1], dSPM [2], sLORETA [3]) have shown excellent performance in cortical reconstructions, they frequently produce smeared or diffuse estimates for deeper subcortical structures. As a consequence, these approaches tend to favor superficial solutions or rely on global regularization priors that may not be optimal for deep dipoles. To address these limitations, we introduce a novel deep learning pipeline specifically designed for subcortical EEG source localization. The proposed architecture utilizes a Long Short-Term Memory (LSTM) architecture to process short (0.2s) raw EEG segments, leveraging temporal dependencies in the signal to directly infer subcortical activation patterns without explicit dipole reconstruction. Because successful training of deep neural networks relies on highly realistic training data, we developed a dedicated simulation pipeline capable of generating physiologically plausible neural activity patterns and corresponding scalp potentials.. Crucially, distinct forward operators were used to generate the training and evaluation datasets, effectively mitigating both data leakage and the so-called “inverse crime”. When benchmarked against nine classical solvers (including MNE, dSPM, and sLORETA) across seven different metrics, our approach demonstrates superior localization accuracy and spatial specificity in both cortical and subcortical scenarios. By integrating end-to-end deep learning with this physiologically inspired simulation pipeline, our framework successfully mitigates the surface bias that plagues standard linear solutions. Ultimately, these results highlight the immense potential of deep learning to address long-standing EEG localization challenges, advancing reliable, non-invasive imaging of deep brain activity. [1] Hämäläinen,Ilmoniemi, 1994 [2] Dale et al., 2000 [3] Pascual-Marqui, 2002.

Social interactions are inherently asymmetric, often characterized by the spontaneous emergence of roles, such as leader and follower, that dictate the flow of the exchange. While these roles are fundamental to social cognition, identifying their neural correlates remains a challenge, especially when they are not assigned experimentally but emerge naturally. Here we introduce a novel framework for aligning multi-subject connectivity networks based on spontaneously emerging social roles. Central to this approach is the development of GraphMatch, a new network similarity metric designed to quantify structural correspondence across multi-subject networks. Using this metric, we validated a realignment procedure driven by the Inter-Imbalance index[LA1.1], a measure of directional asymmetry in inter-brain connectivity that quantifies the imbalance between influences from one subject to the other, and we tested the hypothesis that individuals fulfilling similar social roles exhibit more consistent neural connectivity patterns. We applied this framework to EEG-hyperscanning data from 16 dyads performing a joint action task. Our results show that the proposed realignment significantly increases structural similarity across dyads, as measured by GraphMatch, suggesting that leader-follower dynamics are reflected in systematic asymmetries of inter-brain connectivity. Complementary spectral analyses further supported this interpretation, revealing a predominant directional connectivity flow from leaders to followers. By enabling the identification of spontaneously emerging role differentiation, this work opens new avenues for investigating asymmetric social dynamics and their possible alterations in atypical populations.

The computation of a forward electromagnetic field simulation is a core step for MEEG source estimation. To accomplish this, the real head anatomy is abstracted to a finite-resolution mesh model and the field pattern of a given source configuration is frequently computed using the boundary element method (BEM). Higher-resolution meshes may generally improve the quality of the source estimate, but with common implementations, the mesh grid size cannot be made finer than about 3 mm due to numerical stability constraints. The BEM-FMM forward algorithm overcomes these constraints and permits processing meshes with sub-millimeter resolutions. Even when very high model resolution are thus numerically feasible, working with these fine-grained conductor models would require the acquisition and processing of suitably high-resolution MRI volumes, i.e. ones acquired on 7 T systems. This may introduce further complications into the source reconstruction workflow. Therefore, we seek to empirically investigate the utility of high-resolution models for a common source estimation problem and assess the practical advantage for the source estimates. To this end, we performed MEG recordings of somatosensory fields evoked by painless electrical digit stimulation of participants' right-hand fingers excluding the thumb. In the absence of ground truth for absolute source location, the somatotopic ordering of the digit representations serves as a focal source discrimination problem that we test various model configurations against to determine the model resolution at which such focal sources can be best separated.

Oligodendrocytes are the glial cells responsible for myelination and require sufficient iron levels, especially during development. It is not surprising that myelin and iron are colocalized on the macroscopic scale. Yet, their developmental trajectories diverge: myelin reaches adult levels within a few years, whereas brain-iron continues to accumulate for decades. Quantitative MRI parameters, including the effective relaxation rate R2*, are sensitive to iron as well as myelin. This offers the possibility to monitor brain iron deficiency in early childhood causing hypomyelination and impairing cognition. However, the relative contribution of oligodendrocyte-iron to R2* and related relaxation mechanisms are unexplored. Lifespan post mortem studies are required to understand the contrast changes, but human infant tissue is scarce and rodent models are not informative due to low iron levels and short lifespans. Here, we combined ultra-high resolution quantitative MRI at 3, 4.7, 7 and 9.4 T with quantitative iron mapping by X-ray fluorescence microscopy (at PETRA III, DESY, Hamburg and ID16B, ESRF, Grenoble) in five ethically-collected post mortem chimpanzee brains within the Evolution of Brain Connectivity Project (1.7, 6, 12, 44, 52 y; 2 females). The resulting quantitative iron concentrations were used in biophysical models to estimate the iron-induced contribution to R2*. Across the lifespan, R2* increased in the primary visual cortex. Adult specimen showed layer-specific variation and pronounced cellular-scale heterogeneity in R2*, absent in younger brains. The Stria of Gennari was visible as a hyperintense stripe from 12 years onward similar to what has been observed in humans. Cortical iron profiles mirrored these changes. Young brains displayed a relatively homogeneous iron concentration, whereas adult cortex exhibited a heterogeneous iron distribution. Across the lifespan, iron accumulated in oligodendrocyte somata and in fiber‑like structures located in the mid-cortical layers. Interestingly, in these structures myelin and iron were found to only partly colocalize on the sub-micrometer scale.

Recent high-resolution MRI studies suggest links between mesoscale functional domains and myelin variations in human secondary visual cortex (V2). Reduced myelination has been reported in color- and stereo-selective sites (Haenelt et al., 2023), and increased myelination in regions responsive to high temporal frequencies (Dumoulin et al., 2017). However, the three major extrastriate domains—color, stereo, and motion—have not been mapped within the same individuals, leaving their exact relationship to intracortical myelination in V2 and downstream visual areas unresolved. We studied five participants (30.4 ± 9.3 years, 1 female) with normal vision, for whom motion-, stereo-, and color-selective sites and the boundaries of V2, V3, V3A, and V4 were previously identified using fMRI (1.0mm isotropic) (Kennedy et al., 2023). Participants underwent high-resolution multi-parametric MRI (0.6mm isotropic) (Vaculčiaková et al., 2022). Linear mixed-effects models tested differences in longitudinal relaxation rates (R1), a surrogate for myelin density, across functional domains in V2 and explored patterns in downstream areas. We also compared correlations between R1 values and functional MRI response amplitudes. R1 maps revealed mesoscale patches along the mid-cortical surface. In V2, these patches formed stripe-like patterns orthogonal to the V1/V2 border, often extending into V3 and V4, whereas V3A showed larger, less stripe-like patches. In V2, motion-selective sites exhibited higher R1 than color-selective sites (t = 2.22, p = 0.047) and tended to exceed stereo-selective sites (t = 1.80, p = 0.097), indicating greater myelination within motion-processing domains. In V3A, where color-selective activity was minimal, motion-selective sites again showed higher R1 than stereo-selective sites (t = 5.26, p < 0.001). No domain-specific differences emerged in V3 or V4. Correlation analyses showed that in V2, R1 was more positively associated with motion-selective response amplitude than with color- (t = 2.24, p = 0.045) or stereo-selective responses (t = 2.64, p = 0.022). In V3A, R1 correlated more strongly with motion- than stereo-selective activity (t = 3.57, p < 0.01), while no differences were observed in V3 or V4. These findings show that higher myelination aligns with motion-selective domains in V2 and V3A and correlates with motion-selective response amplitude, revealing a mesoscale structure–function relationship in human visual cortex.