Talk Series

Brain Space Initiative: A Social and Technical Resource for Advancing Brain Research

Brain dynamics and flexible behaviors

Executive control processes and flexible behaviors rely on the integrity of, and dynamic interactions between, large-scale brain networks. The right insular cortex is a critical component of a salience/midcingulo-insular network that is thought to mediate interactions between brain networks involved in externally oriented (central executive/lateral frontoparietal network) and internally oriented (default mode/medial frontoparietal network) processes. How these brain systems reconfigure with development is a critical question for cognitive neuroscience, with implications for neurodevelopmental pathologies affecting brain connectivity. I will describe studies examining how brain network dynamics support flexible behaviors in typical and atypical development, presenting evidence suggesting a unique role for the dorsal anterior insular from studies of meta-analytic connectivity modeling, dynamic functional connectivity, and structural connectivity. These findings from adults, typically developing children, and children with autism suggest that structural and functional maturation of insular pathways is a critical component of the process by which human brain networks mature to support complex, flexible cognitive processes throughout the lifespan.

Updates from ENIGMA: Worldwide collaborative methods

The Enhancing Neuroimaging Genetics through Meta-Analysis (ENIGMA) consortium is a worldwide, largely volunteer research collaboration that recently celebrated its 10th year. In that time, it has become a model of large-scale, imaging data re-use for meta- and mega-analysis, leading to numerous results in genetic effects on brain structure, clinical effects on brain structure and function, and methods for analysis of heterogenous structural and functional data. Clinical research domains range from schizophrenia to bipolar disorder to Parkinson's disease, sleep disorders, HIV and addiction, while methodological developments span gray matter volumes, subcortical shape, sulcal and gyral depth, white matter tract based statistics, and resting state analyses, with recommendations for applying an ENIGMA approach to task-based fMRI are in development. These efforts highlight both the value of combining data analyses, along with the challenges of combining data from studies collected for different reasons in different environments. I will review some of the current efforts underway across the more than 50 working groups within ENIGMA, along with the implications for these kinds of large-scale analyses in the future.

Leveraging biological knowledge: From Brain Mapping to predictive models

The long predominant paradigm in neuroimaging has been to compare (mean) local volume or activity between groups, or to correlate these to behavioral phenotypes. Such approach, however, is intrinsically limited in terms of possible insight into inter-individual differences and application in clinical practice. Recently, the increasing availability of large cohort data and tools for multivariate statistical learning, allowing the prediction of individual cognitive or clinical phenotypes in new subjects, have started a revolution in imaging neuroscience.

The transformation of systems neuroscience into a big data discipline poses a lot of new challenges, yet the most critical aspects is the still sub-optimal relationship between the extremely wide feature-space from neuroimaging and the comparably low number of subjects. This, however, is only true when approaching neuroimaging machine-learning in a naïve fashion, i.e., when ignoring the large body of existing work on human brain mapping. The regional segregation of the brain into distinct modules as well as the large-scale, distributed networks provide the fundamental organizational principles of the human brain and hence the basis for cognitive information processing. Importantly, both can now be mapped in a highly robust fashion by integrating information on hundreds or even thousands of individual subjects to provide a priori information.

This talk will outline the fundamental principles of topographic organization in the human brain and the robust mapping of functional networks. I will then illustrate, how this knowledge on human brain organization can be leveraged for inference on socio-affective or cognitive traits in previously unseen individual subjects or psychopathology in mental disorders. Providing a bidirectional translation, such application will in turn provide information on the respective brain regions and networks.

New Maps of Activation, Connectivity, and Hierarchy using Ultra High Resolution fMRI

Layer fMRI, requiring high field, advanced pulse sequences, and sophisticated processing methods, has emerged in the last decade. The rate of layer fMRI papers published has grown sharply as the delineation of mesoscopic scale functional organization has shown success in providing insight into human brain processing. Layer fMRI promises to move beyond being able to simply identify where and when activation is taking place as inferences made from the activation depth in the cortex will provide detailed directional feedforward and feedback related activity. This new knowledge promises to bridge invasive measures and those typically carried out on humans. In this talk, I will describe the challenges in achieving laminar functional specificity as well as possible approaches to data analysis for both activation studies and resting state connectivity. I will highlight our work demonstrating task-related laminar modulation of primary sensory and motor systems as well as layer-specific activation in dorsal lateral prefrontal cortex with a working memory task. Lastly, I will present recent work demonstrating cortical hierarchy in visual cortex using resting state connectivity laminar profiles.

Everything you wish to know about critical brain dynamics but are afraid to ask

We discuss recent attempts to understand complex brain dynamics at large scale using approaches from statistical physics. The motivation to adopt this approach is rooted in a more general question: "Why life is complex and –most importantly– what is the origin of the over abundance of complexity in nature?" This fundamental problem “is screaming to be answered but seldom is even being asked”, paraphrasing the late Per Bak.

In this lecture we will justify the approach by reviewing our attempts across several scales to understand the origins of complex biological problems from the perspective of critical phenomena. We will then offer an overview of the experimental and numerical results pertaining to complex aspects of large scale brain dynamics.

Disentangling the intrinsic functional architecture of the brain

The brain has all of the hallmarks of a complex system, with meaningful activity occurring at a wide range of spatial and temporal scales. When measured with resting state fMRI, all of this activity is compressed into a single measurement of the resulting hemodynamic response for each voxel at each time point. However, by leveraging the spatial, temporal and spectral properties of different types of activity, we may be able to identify signatures in the rs-fMRI signal. In this talk, I will describe some of the types of activity that we expect to contribute to the rs-fMRI signal and features that might allow us to selectively extract them for use in research or the clinic.

• Zhang X, Pan WJ, Keilholz SD. The relationship between BOLD and neural activity arises from temporally sparse events. Neuroimage. 2020 Feb 15;207:116390. doi: 10.1016/j.neuroimage.2019.116390. Epub 2019 Nov 27. PMID: 31785420; PMCID: PMC7252681
• Kashyap A, Keilholz S. Dynamic properties of simulated brain network models and empirical resting-state data. Netw Neurosci. 2019 Feb 1;3(2):405-426. doi: 10.1162/netn_a_00070. PMID: 30793089; PMCID: PMC6370489
• Abbas A, Belloy M, Kashyap A, Billings J, Nezafati M, Schumacher EH, Keilholz S. Quasi-periodic patterns contribute to functional connectivity in the brain. Neuroimage. 2019 May 1;191:193-204. doi: 10.1016/j.neuroimage.2019.01.076. Epub 2019 Feb 10. PMID: 30753928; PMCID: PMC6440826
• Yousefi B, Shin J, Schumacher EH, Keilholz SD. Quasi-periodic patterns of intrinsic brain activity in individuals and their relationship to global signal. Neuroimage. 2018 Feb 15;167:297-308. doi: 10.1016/j.neuroimage.2017.11.043. Epub 2017 Nov 22. PMID: 29175200; PMCID: PMC5845807
• Thompson GJ, Pan WJ, Keilholz SD. Different dynamic resting state fMRI patterns are linked to different frequencies of neural activity. J Neurophysiol. 2015 Jul;114(1):114-24. doi: 10.1152/jn.00235.2015. Epub 2015 Jun 3. Erratum in: J Neurophysiol. 2015 Dec;114(6):3400. PMID: 26041826; PMCID: PMC4507971

Mapping early brain network changes in neurodegenerative and cerebrovascular disorders: a longitudinal perspective

The spatial patterning of each neurodegenerative disease relates closely to a distinct structural and functional network in the human brain. This talk will mainly describe how brain network-sensitive neuroimaging methods such as resting-state fMRI and diffusion MRI can shed light on brain network dysfunctions associated with pathology and cognitive decline from preclinical to clinical dementia. I will first present our findings from two independent datasets on how amyloid and cerebrovascular pathology influence brain functional networks cross-sectionally and longitudinally in individuals with mild cognitive impairment and dementia. Evidence on longitudinal functional network organizational changes in healthy older adults and the influence of APOE genotype will be presented. In the second part, I will describe our work on how different pathology influences brain structural network and white matter microstructure. I will also touch on some new data on how brain network integrity contributes to behavior and disease progression using multivariate or machine learning approaches. These findings underscore the importance of studying selective brain network vulnerability instead of individual region and longitudinal design. Further developed with machine learning approaches, multimodal network-specific imaging signatures will help reveal disease mechanisms and facilitate early detection, prognosis and treatment search of neuropsychiatric disorders.

Clarifying the Neurocognitive Systems that Support Reading and Language Development

Connectivity through dynamics: surprise and variance reduction

I will recall how reduction in variance and reduction in surprise are two similar (and sometimes identical) ways of investigating dependencies in dynamical systems.

I will present how this framework can be used to investigate higher order dependencies to find information-based multiplets and to allow for a more precise characterization of multivariate patterns of connectivity.

I will then show how we can look at interactions across temporal scales, and present some applications in neuroscience.

• Reid, A. T., Headley, D. B., Mill, R. D., Sanchez-Romero, R., Uddin, L. Q., Marinazzo, D., ... & Calhoun, V. (2019). Advancing functional connectivity research from association to causation. Nature neuroscience, 22(11), 1751-1760. https://www.nature.com/articles/s41593-019-0510-4
• He, B., Astolfi, L., Valdés-Sosa, P. A., Marinazzo, D., Palva, S. O., Bénar, C. G., ... & Koenig, T. (2019). Electrophysiological brain connectivity: theory and implementation. IEEE transactions on biomedical engineering, 66(7), 2115-2137. https://ieeexplore.ieee.org/document/8708690/
• Faes, L., Marinazzo, D., & Stramaglia, S. (2017). Multiscale information decomposition: Exact computation for multivariate Gaussian processes. Entropy, 19(8), 408. https://www.mdpi.com/1099-4300/19/8/408

Causal models of brain function

This talk will focus on the modelling of resting state time series or endogenous neuronal activity. I will survey recent developments in modelling distributed neuronal fluctuations – spectral dynamic causal modelling (DCM) for functional MRI [1, 2] – and how this modelling rests upon functional connectivity. The dynamics of brain connectivity has recently attracted a lot of attention among brain mappers. I will also show a novel method to identify dynamic effective connectivity using spectral DCM [3]. Further, I will summarise the development of the next generation of DCMs towards large-scale, whole-brain schemes which are computationally inexpensive [4], to the other extreme of the development using more sophisticated and biophysically detailed modelling based on the canonical micro circuits [5]

1. Friston, K. J., Kahan, J., Biswal, B., & Razi, A. (2014). A DCM for resting state fMRI. NeuroImage, 94(100), 396–407. https://doi.org/10.1016/j.neuroimage.2013.12.009
2. Razi, A., Kahan, J., Rees, G., & Friston, K. J. (2015). Construct validation of a DCM for resting state fMRI. NeuroImage, 106, 1–14. https://doi.org/10.1016/j.neuroimage.2014.11.027
3. Park, H. J., Friston, K. J., Pae, C., Park, B., & Razi, A. (2018). Dynamic effective connectivity in resting state fMRI. NeuroImage, 180(Pt B), 594–608. https://doi.org/10.1016/j.neuroimage.2017.11.033
4. Razi, A., Seghier, M. L., Zhou, Y., McColgan, P., Zeidman, P., Park, H. J., Sporns, O., Rees, G., & Friston, K. J. (2017). Large-scale DCMs for resting-state fMRI. Network neuroscience (Cambridge, Mass.), 1(3), 222–241. https://doi.org/10.1162/NETN_a_00015
5. Friston, K. J., Preller, K. H., Mathys, C., Cagnan, H., Heinzle, J., Razi, A., & Zeidman, P. (2019). Dynamic causal modelling revisited. NeuroImage, 199, 730–744. https://doi.org/10.1016/j.neuroimage.2017.02.045

Graph signal processing for computational neuroimaging

State-of-the-art magnetic resonance imaging (MRI) provides unprecedented opportunities to study brain structure (anatomy) and function (physiology). Based on such data, graph representations can be built where nodes are associated to brain regions and edge weights to strengths of structural or functional connections. In particular, structural graphs capture major neural pathways in white matter, while functional graphs map out statistical interdependencies between pairs of regional activity traces. Network analysis of these graphs has revealed emergent system-level properties of brain structure or function, such as efficiency of communication and modular organization.

In this talk, graph signal processing (GSP) will be presented as a novel framework to integrate brain structure, contained in the structural graph, with brain function, characterized by activity traces that can be considered as time-dependent graph signals. Such a perspective allows to define novel meaningful graph-filtering operations of brain activity that take into account smoothness of signals on the anatomical backbone. This allows to define a new measure of “coupling” between structure and function based on how activity is expressed on structural graph harmonics. To provide statistical inference, we also extend the well-known Fourier phase randomization method to generate surrogate data to the graph setting. This new measure reveals a behaviorally relevant spatial gradient, where sensory regions tend to be more coupled with structure, and high-level cognitive ones less so. In addition, we also make a case to introduce the graph modularity matrix at the core of GSP, in order to incorporate knowledge about graph community structure when processing signals on the graph, but without the need for community detection. Finally, recent work will highlight how the spatial resolution of this type of analyses can be increased to the voxel level, representing a few hundredth thousands of nodes.

Recommended Articles:

• W. Huang*, T. A. W. Bolton*, J. D. Medaglia, D. S. Bassett, A. Ribeiro, D. Van De Ville, A Graph Signal Processing Perspective on Functional Brain Imaging, Proceedings of the IEEE, 2018, 106, 868-885
• M. G. Preti, D. Van De Ville, Decoupling of Brain Function from Structure Reveals Regional Behavioral Specialization in Humans, Nature Communications, 2019, 10, 4747
• A. Tarun, H. Behjat, T. Bolton, D. Abramian, D. Van De Ville, Structural Mediation of Human Brain Activity Revealed by White-Matter Interpolation of fMRI, NeuroImage, 2020, 213, 116718
• M. Petrovic*, R. Liégeois*, T. A. W. Bolton, D. Van De Ville, Community-Aware Graph Signal Processing, IEEE Signal Processing Magazine, in press

Representation learning and alignment in biological and artificial neural networks

Good representations are critical to the success of both biological and artificial information processing systems. In this talk, I will highlight new approaches that my lab is developing for representation learning and alignment, and demonstrate their applications in the analysis and interpretation of both biological and artificial neural networks. Being able to align neural representations, promises meaningful ways of comparing high-dimensional neural activities across times, subsets of neurons, individuals, and potentially across disease.

Brain Metastability

Two paradigms continue to spar in the neuroscience at the local and system levels. At the elemental level of neurons, and grounded in the prominent work of many cellular physiologists including Eccles, Hodgkin and Huxley, overwhelming evidence implicate directional flows of information from functional unit to functional unit. At another system level, and thrusted by giants of system Neuroscience like Freeman, Eckhorn, Gray and Singer, focus is shifted to ensemble activities that yield evidence of functional synchronization. In their extreme forms, both paradigms eschew one important property of complex functional systems. A strictly serial model of synaptic propagation labors to achieve mass action. And a completely collective system presents many roadblocks to a dynamics of its self-organized activities, making for a brain frozen in time and inefficient at adaptation and flexibility. I will place in the reconciliatory middle ground the theory of brain metastability initially pioneered by Kelso. Its spatiotemporal complexity is permissive of information flows at the same time as transient coordination provides collective power at multiple spatial scales. I will provide mathematical bases for its study in models that prepare for the empirical encounter of its phenomenology; I will outline empirical evidence of its pervasiveness. Although metastability is conceptually contiguous with the two aforementioned paradigms, I will describe the pitfalls that follow from analyzing it as approximations of them, and I will argue that a shift in perspective and methods is required to fully understand brain complexity.

• Tognoli, E. and Kelso, J.A.S., 2014. Enlarging the scope: grasping brain complexity. Frontiers in systems neuroscience, 8, p.122. https://dx.doi.org/10.3389%2Ffnsys.2014.00122
• Tognoli, E. and Kelso, J.A.S., 2014. The metastable brain. Neuron, 81(1), pp.35-48. https://dx.doi.org/10.1016%2Fj.neuron.2013.12.022

Emotion Representation in the Human Brain

Emotion is central to human experience; it influences cognition, behavior, mental health, and well-being. Despite their importance and self-evidence, emotions are notoriously difficult to define scientifically. In this talk, I will present research combining approaches from cognitive neuroscience and machine learning to build models of human brain activity that track the engagement of distinct affective processes. I will discuss evidence suggesting that brain representations of emotional states are high-dimensional, distributed, and are inconsistent with intuitive psychological models that organize emotions along a small number of dimensions, such as valence and arousal. Building and validating quantitative models of human brain activity promises to provide insight into the nature of emotion provide novels targets for cognitive and therapeutic interventions.

• Kragel, P. A., Reddan, M. C., LaBar, K. S., & Wager, T. D. (2019). Emotion schemas are embedded in the human visual system. Science advances, 5(7), eaaw4358.
• Kragel, P. A., Koban, L., Barrett, L. F., & Wager, T. D. (2018). Representation, pattern information, and brain signatures: from neurons to neuroimaging. Neuron, 99(2), 257-273.
• Kragel, P. A., & LaBar, K. S. (2016). Decoding the nature of emotion in the brain. Trends in cognitive sciences, 20(6), 444-455.

Nonlinear Independent Component Analysis, with Applications to EEG/MEG

Unsupervised learning, in particular learning general nonlinear representations, is one of the deepest problems in machine learning. Estimating latent quantities in a generative model provides a principled framework, and has been successfully used in the linear case, e.g., with independent component analysis (ICA) and sparse coding. ICA is well established in analysis of brain imaging data. However, extending ICA to the nonlinear case has proven to be extremely difficult: A straight-forward extension is unidentifiable, i.e., it is not possible to recover those latent components that actually generated the data. Here, we show that this problem can be solved by using additional information, in particular in the form of temporal structure. Our methods are related to 'self-supervised' learning increasingly used in deep learning. Application of such methods to EEG/MEG analysis is a promising avenue for research.

• Hyvarinen, Aapo, and Hiroshi Morioka. 'Unsupervised Feature Extraction by Time-Contrastive Learning and Nonlinear ICA.'
• Morioka, Hiroshi, Hermanni Hälvä, and Aapo Hyvarinen. 'Independent Innovation Analysis for Nonlinear Vector Autoregressive Process.'
• Banville, Hubert, et al. 'Uncovering the structure of clinical EEG signals with self-supervised learning.'

The Bilingual Brain: Uncovering Neural Architecture for Learning to Read across Languages

Learning to read change's mind and brain. How might bilingual experience influence children’s neural architecture for learning to read? Words have sounds and meanings. The neural architecture for learning to read includes the formation of sound-to-print and meaning-to-print neurocognitive pathways. Importantly, there is also significant cross-linguistic variation in how children form these associations. In phonologically transparent such as Italian, children develop stronger sound-to-print networks, whereas learners of Chinese form stronger meaning-to-print associations. To understand how bilingual experiences influence children’s developing neural architecture for learning to read we use fNIRS with Spanish-English and Chinese-English bilingual children in the US. Several key findings emerge from these data that we will discuss during the presentation. First, the findings reveal principled bilingual transfer effects of children’s phonological, morphological, and lexical processes for learning to read. Second, the findings reveal the Universal aspects of literacy development across these typologically-distinct languages. The findings are discussed in light of theoretical perspectives on learning to read, bilingual development, as well as the universal and language-specific aspects of language, literacy, and dyslexia.

• Arredondo MM, Hu XS, Seifert E, Satterfield T, Kovelman I. “Bilingual exposure enhances left IFG specialization for language in children.” Bilingualism: Language and Cognition. 2019.
• Ip KI, Hsu LS, Arredondo MM, Tardif T, Kovelman I. “Brain bases of morphological processing in Chinese-English bilingual children.” Developmental Science. 2016.

“In the wild“ brain activity decoding

To map the neural substrate of mental function, cognitive neuroimaging relies on controlled psychological manipulations that engage brain systems associated with specific cognitive processes. In order to build comprehensive atlases of cognitive function in the brain, it must assemble maps for many different cognitive processes, which often evoke overlapping patterns of activation. Such data aggregation faces contrasting goals: on the one hand finding correspondences across vastly different cognitive experiments, while on the other hand precisely describing the function of any given brain region.

In this talk I will present two analysis frameworks that tackle these difficulties and thereby enable the generation of brain atlases for cognitive function. The first one uses deep-learning techniques to extract representations—task-optimized networks—that form a set of basis cognitive dimensions relevant to the psychological manipulations. This approach does not assume any prior knowledge of the commonalities shared by the studies in the corpus; those are inferred during model training.

The second one leverages ontologies of cognitive concepts and multi-label brain decoding to map the neural substrate of these concepts. Crucially, it can accurately decode the cognitive concepts recruited in new tasks. These results demonstrate that aggregating independent task-fMRI studies can provide a more precise global atlas of selective associations between brain and cognition.

• “NeuroQuery, comprehensive meta-analysis of human brain mapping.” Jérôme Dockès, Russell Poldrack, Romain Primet, Hande Gözükan, Tal Yarkoni, Fabian Suchanek, Bertrand Thirion, Gaël VaroquauxeLife, eLife Sciences Publication, 2020. https://hal.archives-ouvertes.fr/hal-02485642v2
• “Atlases of cognition with large-scale human brain mapping.” Gaël Varoquaux, Yannick Schwartz, Russell Poldrack, Baptiste Gauthier, Danilo Bzdok, Jean-Baptiste Poline, Bertrand Thirion, PLoS Computational Biology, Public Library of Science, 2018, 14 (11), pp.e1006565. https://www.hal.inserm.fr/inserm-02146700
• “Extracting Universal Representations of Cognition across Brain-Imaging Studies.” Arthur Mensch, Julien Mairal, Bertrand Thirion, Gaël Varoquaux, PLoS Computational Biology. https://hal.archives-ouvertes.fr/hal-01874713v2

Altered Central Multisensory Processing in Post-concussion Vestibular Dysfunction

Vestibular symptoms are one of the most frequent complaints after concussion and are associated with prolonged recovery. However, the underlying structural and functional alterations underlying post-concussion vestibular dysfunction are not well understood. Furthermore, what constitutes the vestibular network in humans has not been fully elucidated. In this talk, I will review the clinical findings of vestibular impairment following concussion and the current models of central vestibular processing. I will then discuss the data supporting central multisensory processing as the primary driver of subacute and chronic post-concussion vestibular symptoms. Defining the alterations in vestibular processing after head injury will allow better prognostication and more targeted, patient-centric neurorehabilitation therapies.

• Hurtado JE, Heusel-Gillig L, Risk BB, Trofimova A, Abidi SA, Allen JW, Gore RK. “Technology-enhanced visual desensitization home exercise program for post-concussive visually induced dizziness: a case series.” Physiother Theory Pract. 2020. https://doi.org/10.1080/09593985.2020.1815259
• Indovina I, Bosco G, Riccelli R, Maffei V, Lacquaniti F, Passamonti L, Toschi N. “Structural connectome and connectivity lateralization of the multimodal vestibular cortical network.” Neuroimage. 2020. https://doi.org/10.1016/j.neuroimage.2020.117247
• Trofimova A, Smith JL, Ahluwalia V, Hurtado J, Gore RK, Allen JW. Alterations in Resting-State Functional Brain Connectivity and Correlations with Vestibular/Ocular-Motor Screening Measures in Postconcussion Vestibular Dysfunction. Journal of Neuroimaging. 2021. https://doi.org/10.1111/jon.12834
• Allen JW, Trofimova A, Ahluwalia V, Smith JL, Abidi SA, Peters MAK, Rajananda S, Hurtado JE, Gore RK. “Altered Processing of Complex Visual Stimuli in Patients with Postconcussive Visual Motion Sensitivity.” AJNR American Journal of Neuroradiology. 2021. https://doi.org/10.3174/ajnr.A7007

Age-related brain changes and risk factors for cognitive decline

Functional differences in the default mode network and related memory systems are observed in typical aging. Our research, as an example, shows that particularly functional connectivity within the posterior memory system appears affected by aging. Furthermore, we found that negative subsequent memory effects may differentially support memory performance across the lifespan, suggesting a developmental maturation and age-related decline.

The posterior default mode regions also appear most vulnerable to early Alzheimer’s disease related brain changes. Subjective cognitive decline, the perceived decline in cognitive abilities in the absence of deficits on clinical assessments, is a known risk factor for Alzheimer’s disease. Our work shows that participants with subjective cognitive decline have lower default mode network functional connectivity compared to older adults without, particularly in posterior regions, suggesting possible early functional brain changes in older adults with subjective cognitive decline in similar brain regions as observed in early Alzheimer’s disease. However, as these findings are based on cross-sectional data, they don’t provide insight in actual brain changes in those with subjective cognitive decline. One of our current lines of research aims to address this gap in the literature by examining changes in functional connectivity over a three-year period. In line with our cross-sectional finding, our longitudinal results reveal a steeper decline in default mode network functional connectivity in older adults with subjective cognitive decline than those without.

Precision dynamical mapping using topological data analysis reveals a unique hub-like transition state at rest

Even in the absence of external stimuli, neural activity is both highly dynamic and organized across multiple spatiotemporal scales. The continuous evolution of brain activity patterns during rest is believed to help maintain a rich repertoire of possible functional configurations that relate to typical and atypical cognitive phenomena. Whether these transitions or “explorations” follow some underlying arrangement or instead lack a predictable ordered plan remains to be determined. Here, using a precision dynamics approach, we aimed at revealing the rules that govern transitions in brain activity at rest at the single participant level. We hypothesized that by revealing and characterizing the overall landscape of whole brain configurations (or states) we could interpret the rules (if any) that govern transitions in brain activity at rest. To generate the landscape of whole-brain configurations we used Topological Data Analysis based Mapper approach. Across all participants, we consistently observed a rich topographic landscape in which the transition of activity from one state to the next involved a central hub-like “transition state.” The hub topography was characterized as a shared attractor-like basin where all canonical resting-state networks were represented equally. The surrounding periphery of the landscape had distinct network configurations. The intermediate transition state and traversal through it via a topographic gradient seemed to provide the underlying structure for the continuous evolution of brain activity patterns at rest. In addition, differences in the landscape architecture were more consistent within than between subjects, providing evidence of idiosyncratic dynamics and potential utility in precision medicine.

Recommended Articles:

• “Precision dynamical mapping using topological data analysis reveals a unique hub-like transition state at rest.” Manish Saggar, James M. Shine, Raphaël Liégeois, Nico U. F. Dosenbach, and Damien Fair, bioRxiv, 2021. https://www.biorxiv.org/content/10.1101/2021.08.05.455149v2
• “Towards a New Approach to Reveal Dynamical Organization of the Brain Using Topological Data Analysis.” Manish Saggar, Olaf Sporns, Javier Gonzalez-Castillo, Peter A. Bandettini, Gunnar Carlsson, Gary Glover, and Allan L. Reiss, Nature Communications, 2018. https://www.nature.com/articles/s41467-018-03664-4

Cross-Frequency Coupling in the Brain

Cognitive processing and goal-directed behavior are hypothesized to be the result of frequency specific interactions of specialized but widely distributed cortical regions. Cross-frequency coupling (CFC) has been proposed to coordinate the neural dynamics between these distinct frequency bands across spatial and temporal scales. One particular form of CFC known as phase amplitude coupling (PAC) quantifies the interplay between the phase of a slower oscillation and the envelope of a faster oscillation. The existing methods for assessing PAC have some limitations including limited frequency resolution, sensitivity to noise, data length and sampling rate due to the inherent dependence on bandpass filtering. Moreover, most of the current applications of PAC to neuronal data focuses on average coupling within a single channel. In this talk, I will present some recent developments in the computation of PAC. First, I will introduce a new time-frequency based phase-amplitude coupling measure that addresses the biases encountered by Hilbert transform based PAC measures. Next, I will introduce an extension of PAC analysis from the bivariate to the multivariate case allowing us to look at whole brain cross-frequency coupling networks. Finally, I will present applications of these new PAC computation tools on EEG data collected during error monitoring.

• T. T. K. Munia and S. Aviyente, “Time-frequency based phase-amplitude coupling measure for neuronal oscillations,” Scientific Reports, 2019. https://doi.org/10.1038/s41598-019-48870-2
• T. T. K. Munia and S. Aviyente, “Comparison of wavelet and RID-Rihaczek based methods for phase-amplitude coupling,” IEEE Signal Processing Letters, vol. 26, pp. 1897-1901, 2019. https://doi.org/10.1109/LSP.2019.2952025
• T. T. K. Munia and S. Aviyente, “Multivariate Analysis of Bivariate Phase-Amplitude Coupling in EEG Data Using Tensor Robust PCA,” IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol. 29, pp. 1268-1278, 2021. https://doi.org/10.1109/TNSRE.2021.3092890

The lure of misleading causal statements in functional connectivity research

As neuroscientists we want to understand how causal interactions or mechanisms within the brain give rise to perception, cognition, and behavior. It is typical to estimate interaction effects from measured activity using statistical techniques such as functional connectivity, Granger Causality, or information flow, whose outcomes are often falsely treated as revealing mechanistic insight. Since these statistical techniques fit models to low-dimensional measurements from brains, they ignore the fact that brain activity is high-dimensional. Here we focus on the obvious confound of common inputs: the countless unobserved variables likely have more influence than the few observed ones. Any given observed correlation can be explained by an infinite set of causal models that take into account the unobserved variables. Therefore, correlations within massively undersampled measurements tell us little about mechanisms. We argue that these mis-inferences of causality from correlation are augmented by an implicit redefinition of words that suggest mechanisms, such as connectivity, causality, and flow.

• Mehler, D. M. A., & Kording, K. P. (2018). The lure of misleading causal statements in functional connectivity research. arXiv preprint arXiv:1812.03363. https://arxiv.org/abs/1812.03363
• Reid, A. T., Headley, D. B., Mill, R. D., Sanchez-Romero, R., Uddin, L. Q., Marinazzo, D., ... & Cole, M. W. (2019). Advancing functional connectivity research from association to causation. Nature neuroscience, 22(11), 1751-1760. https://www.nature.com/articles/s41593-019-0510-4

Turbulence in Human Brain Dynamics

Turbulence is a special dynamical state driving many physical systems by way of its ability to facilitate fast energy/information transfer across scales. These qualities are important for brain function, but it is currently unknown if the brain also exhibits turbulence as a fundamental organisational principle. Using large-scale neuroimaging empirical data from 1003 healthy participants, we demonstrate amplitude turbulence in human brain dynamics. Furthermore, we build a whole-brain model with coupled oscillators to demonstrate that the best fit of our model to the data corresponds to a region of maximally developed amplitude turbulence, which also corresponds to maximal sensitivity to the processing of external stimulations (information capability). The model shows the economy of anatomy by following the Exponential Distance Rule of anatomical connections as a cost-of-wiring principle. This establishes a firm link between turbulence and optimal brain function. Overall, our results reveal a novel way of analysing and modelling whole-brain dynamics that for the first time ever establishes turbulence as a fundamental basic principle of brain organisation.

• Deco, G. and Kringelbach, M.L., 2020. Turbulent-like dynamics in the human brain. Cell reports, 33(10), p.108471. https://doi.org/10.1016/j.celrep.2020.108471
• Deco, G., Kemp, M. and Kringelbach, M.L., 2021. Leonardo da Vinci and the search for order in neuroscience. Current Biology, 31(11), pp.R704-R709. https://doi.org/10.1016/j.cub.2021.03.098
• Deco, G., Perl, Y.S., Vuust, P., Tagliazucchi, E., Kennedy, H. and Kringelbach, M.L., 2021. Rare long-range cortical connections enhance human information processing. Current Biology. https://doi.org/10.1016/j.cub.2021.07.064

Toward an open ecosystem for neuroimaging data analysis and sharing

Significant concerns have been raised regarding the reproducibility of current scientific practices. I will provide an overview of the problems with reproducibility (focusing in particular on neuroimaging research) and outline a set of tools that provide the ability to build fully reproducible analysis workflows. I will also discuss the need for transparency through data sharing, and the importance of community standards for data organization to make data sharing effective.

• Botvinik-Nezer, Rotem, et al. Variability in the analysis of a single neuroimaging dataset by many teams. Nature 582.7810 (2020): 84-88. https://www.nature.com/articles/s41586-020-2314-9

Toward causality in the brain: from lesions to psychedelics

The recent explosion of neuroimaging studies in large-scale populations of humans has begin to reveal complex mappings between brain and behavior. Cross-sectional studies do not allow, however, for exploring causality in the brain. Lesion studies have historically allowed direct linking between an area of damage and a specific behavioral change, and more recent virtual lesion (TMS), ECoG and pharmacological experiments have allowed exploration of the effect of manipulating brain function on behavior. These type of studies take steps in the direction of understanding causality in brain-behavior relationships. In her talk, Amy will discuss recent work in this area and present some of her own work studying the effect of damage on both brain structure/function and behavior, how the brain recovers from damage and how psychedelics may impact whole-brain function.

• Kuceyeski, A., Maruta, J., Relkin, N., Raj, A., 2013. The Network Modification (NeMo) Tool: elucidating the effect of white matter integrity changes on cortical and subcortical structural connectivity. Brain Connect 3, 451-463. https://doi.org/10.1089/brain.2013.0147
• Kuceyeski, A. , Jamison, K.W., Owen, J.P., Raj, A., Mukherjee, P., 2019. Longitudinal increases in structural connectome segregation and functional connectome integration are associated with better recovery after mild TBI. Hum Brain Mapp 40, 4441-4456. https://doi.org/10.1002%2Fhbm.24713
• Tozlu, C., Jamison, K., Gu, Z., Gauthier, S.A., Kuceyeski, A., 2021. Estimated connectivity networks outperform observed connectivity networks when classifying people with multiple sclerosis into disability groups. Neuroimage Clin 32, 102827. https://doi.org/10.1016%2Fj.nicl.2021.102827
• Olafson, E.R., Jamison, K.W., Sweeney, E.M., Liu, H., Wang, D., Bruss, J.E., Boes, A.D., Kuceyeski, A., 2021. Functional connectome reorganization relates to post-stroke motor recovery and structural and functional disconnection. Neuroimage 245, 118642. https://doi.org/10.1016/j.neuroimage.2021.118642
• Singleton P, Luppi A, Carhart-Harris RL, Cruzat J, Roseman L, Deco G, Kringelbach M, Stamatakis EA, Kuceyeski A. Psychedelics flatten the brain’s energy landscape: evidence from receptor-informed network control theory. Nature Communications (in press).
• Olafson E, Russello G, Jamison K, Liu H, Wang D, Bruss J, Boes A, Kuceyeski A. Increased prevalence of a frontoparietal brain state is associated with better motor recovery after stroke affecting dominant-hand cortico-spinal tract. Communications Biology (in press).

Dynamics of threat and reward processing in the human brain

Research on the emotional and motivational brain often employs relatively static paradigms, such as the presentation of emotion-laden faces. Because natural behaviors evolve temporally, advancing understanding of dynamic processes holds promise in opening new research avenues. In this presentation, I will discuss recent attempts to develop paradigms to study the dynamics of threat- and reward-related processes, as well as their interactions. I will also describe work exploring how recurrent neural networks can be used to characterize spatio-temporal dynamics as measured by functional MRI data, and how recurrent neural networks can be used to study more naturalistic/dynamic paradigms.

Exploring the brain basis of population mental health in the UK Biobank

The UK Biobank is a longitudinal population neuroimaging dataset with extensive neuroimaging, genomic, and phenotypic measures for up to N=100,000 older-age participants. The statistical power, deep phenotyping, and lack of exclusion criteria in this cohort provide an important opportunity to investigate ‘real-world’ mental health. I will present several studies performed in our lab over the past three years to validate UK Biobank mental health measures, identify reproducible multivariate neural correlates of mental health, parse clinical and biological heterogeneity in mental health, and identify lifestyle factors that promote resilience to mental health.

• Dutt, R. K., Hannon, K., Easley, T. O., Griffis, J. C., Zhang, W., & Bijsterbosch, J. D. (2021). “Mental health in the UK Biobank: A roadmap to self‐report measures and neuroimaging correlates.” Human Brain Mapping. https://doi.org/10.1002/hbm.25690
• Zhang, W., Paul, S. E., Winkler, A., Bogdan, R., & Bijsterbosch, J. D. (2022). “Shared brain and genetic architectures between mental health and physical activity.” In bioRxiv. https://doi.org/10.1101/2021.12.28.21268486
• Easley, T. O., Chen, R., Hannon, K., Dutt, R. K., & Bijsterbosch, J. D. (2022). “Population modeling with machine learning can enhance measures of mental health - Open-Data Replication.” In bioRxiv (p. 2022.04.04.487069). https://doi.org/10.1101/2022.04.04.487069

Postponed

Use-Inspired Music Cognition: Music-Based Interventions for Cognitive and Brain Health

As the field of music cognition is rapidly burgeoning, researchers are beginning to consider how the unique amalgam of scientific and humanistic study of music may translate towards large-scale interventions that may improve cognition for many, including but not limited to people from neurodiverse populations. In this talk I will examine novel ways in which music cognition research may help improve cognition, in ways that move away from overused tropes (e.g. the Mozart Effect) towards future directions of use-inspired music cognition research. As use cases, I will describe some recent studies in my lab that capitalize on new musical technology, developed from first principles from music cognition research, to help those with attention deficits, memory disorders, and Parkinson's Disease. Our results show how music cognition can help refine and target music-based interventions for multiple special populations, by pinpointing ways in which music capitalizes on fundamental operating characteristics of the brain.

• Tichko, P., Kim, J. C., Large, E., & Loui, P. (2020). Integrating music‐based interventions with Gamma frequency stimulation: Implications for healthy ageing. European Journal of Neuroscience. https://doi.org/10.1111/ejn.15059
• Krotinger, A., & Loui, P. (2021). Rhythm and groove as cognitive mechanisms of dance intervention in Parkinson’s disease. PLoS ONE, 16(5), e0249933. https://doi.org/10.1371/journal.pone.0249933
• Savage, P. E., Loui, P., Tarr, B., Schachner, A., Glowacki, L., Mithen, S., & Fitch, W. T. (2020). Music as a coevolved system for social bonding. Behavioral and Brain Sciences, 44, e59. https://doi.org/10.1017/S0140525X20000333

Data Integration in Neuroimaging Studies

As advances in technology allow the acquisition of complementary information, it is common for scientific studies to collect multiple datasets. In this talk, we will examine two approaches to data integration. In the first part, we will analyze the joint and individual structure in cognitive assessments and brain morphometry from the Alzheimer’s Disease Neuroimaging Initiative. We introduce probabilistic joint and individual variation (ProJIVE), which extends probabilistic PCA to multiple datasets. ProJIVE reveals links between brain regions and cognitive performance. In the second part of this talk, we introduce Simultaneous Non-Gaussian component analysis (SING) in which dimension reduction and feature extraction are achieved simultaneously. Instead of maximizing variance using PCA, SING maximizes non-Gaussianity to extract a lower-dimensional subspace and reveal new insights. We apply our method to a working memory task and resting-state correlations from the Human Connectome Project. We find joint structure as evident from learned spatial correspondence. Moreover, some of the subject scores are related to fluid intelligence.

• Benjamin B. Risk, Irina Gaynanova (2021) Simultaneous non-Gaussian component analysis (SING) for data integration in neuroimaging. Ann. Appl. Stat. 15(3): 1431-1454. https://doi.org/10.1214/21-AOAS1466

TBA

Edge-centric Connectomics

Comparison and integration of neuroimaging data from different brains and populations is fundamental in neuroscience in that it underlies countless statistically meaningful conclusions about the human brain. However, the uniqueness of each human brain imposes fundamental challenges to existing approaches that aim to compare and integrate brain science data across individuals and populations. This longstanding challenge has escalated and become more urgent with the recent dramatic growth of publicly available brain science data, particularly neuroimaging data. Despite numerous efforts in the brain science field over the past few decades, there is still a fundamental lack of basic understanding and concrete representation of the regularity and variability of the human brain. In this talk, I will share the research experience on representation of human brain commonality and individuality using neuroimaging data in the Cortical Architecture Imaging and Discovery Lab at the University of Georgia in the past decade. I will introduce the opportunities and challenges in creating a universal and individualized brain reference system that encodes functional localizations of brain structures by fiber connection patterns and topographic folding patterns, which possess finer granularity, better functional homogeneity, more accurate functional localization, and intrinsically established correspondence across different brains.

Disentangle Human Brain Commonality and Individuality

Comparison and integration of neuroimaging data from different brains and populations is fundamental in neuroscience in that it underlies countless statistically meaningful conclusions about the human brain. However, the uniqueness of each human brain imposes fundamental challenges to existing approaches that aim to compare and integrate brain science data across individuals and populations. This longstanding challenge has escalated and become more urgent with the recent dramatic growth of publicly available brain science data, particularly neuroimaging data. Despite numerous efforts in the brain science field over the past few decades, there is still a fundamental lack of basic understanding and concrete representation of the regularity and variability of the human brain. In this talk, I will share the research experience on representation of human brain commonality and individuality using neuroimaging data in the Cortical Architecture Imaging and Discovery Lab at the University of Georgia in the past decade. I will introduce the opportunities and challenges in creating a universal and individualized brain reference system that encodes functional localizations of brain structures by fiber connection patterns and topographic folding patterns, which possess finer granularity, better functional homogeneity, more accurate functional localization, and intrinsically established correspondence across different brains.

• Xiao Li, Tao Liu, Yujie Li, Qing Li, Xianqiao Wang, Xintao Hu, Lei Guo, Tuo Zhang, Tianming Liu, Marmoset Brain ISH Data Revealed Molecular Difference Between Cortical Folding Patterns, Cerebral Cortex, 2020. https://doi.org/10.1093/cercor/bhaa317
• Huan Liu, Shu Zhang, Xi Jiang, Tuo Zhang, Heng Huang, Fangfei Ge, Lin Zhao, Xiao Li, Xintao Hu, Junwei Han, Lei Guo, Tianming Liu, The Cerebral Cortex is Bisectionally Segregated into Two Fundamentally Different Functional Units of Gyri and Sulci, Cerebral Cortex, 2018. https://doi.org/10.1093/cercor/bhy305
• Fangfei Ge, Xiao Li, Mir Jalil Razavi, Hanbo Chen, Tuo Zhang, Shu Zhang, Lei Guo, Xiaoping Hu, Xianqiao Wang, Tianming Liu, Denser Growing Fiber Connections Induce 3-hinge Gyral Folding, Cerebral Cortex, 2017. https://doi.org/10.1093/cercor/bhx227
• Hanbo Chen, Yujie Li, Fangfei Ge, Gang Li, Dinggang Shen, Tianming Liu, Gyral Net: A New Representation of Cortical Folding Organization, Medical Image Analysis, 2017. https://doi.org/10.1016/j.media.2017.07.001
• Xiao Li, Hanbo Chen, Tuo Zhang, Xiang Yu, Xi Jiang, Kaiming Li, Longchuan Li, Mir Jalil Razavi, Xianqiao Wang, Xintao Hu, Junwei Han, Lei Guo, Xiaoping Hu, Tianming Liu, Commonly-Preserved and Species-Specific Gyral Folding Patterns across Primate Brains, Brain Structure and Function, 2016. https://doi.org/10.1007/s00429-016-1329-3
• Fan Deng, Xi Jiang, Dajiang Zhu, Tuo Zhang, Kaiming Li, Lei Guo, Tianming Liu, A functional model of cortical gyri and sulci, Brain Structure and Function, 2013. https://dx.doi.org/10.1007%2Fs00429-013-0581-z
• Hanbo Chen, Tuo Zhang, Lei Guo, Kaiming Li, Xiang Yu, Longchuan Li, Xintao Hu, Junwei Han, Xiaoping Hu, Tianming Liu, Coevolution of Gyral Folding and Structural Connection Patterns in Primate Brains, Cerebral Cortex, 2012. https://doi.org/10.1093/cercor/bhs113
• Jingxin Nie, Lei Guo, Kaiming Li, Yonghua Wang, Guojun Chen, Longchuan Li, Hanbo Chen, Fan Deng, Xi Jiang, Tuo Zhang, Ling Huang, Carlos Faraco, Degang Zhang, Cong Guo, Pew-Thian Yap, Xintao Hu, Gang Li, Jinglei Lv, Yixuan Yuan, Dajiang Zhu, Junwei Han, Dean Sabatinelli, Qun Zhao, L Stephen Miller, Bingqian Xu, Ping Shen, Simon Platt, Dinggang Shen, Xiaoping Hu, Tianming Liu, Axonal Fiber Terminations Concentrate on Gyri, Cerebral Cortex, 2011.https://doi.org/10.1093/cercor/bhaa317

Functional MRI of dynamic neural and physiological states

Spontaneous fMRI signals provide a valuable window into human brain functional organization. Recent work has also demonstrated the potential for extracting information about dynamic internal states from fMRI. In this talk, I will discuss studies in which we use multimodal functional imaging to investigate the dynamics of spontaneous brain activity and to track ongoing changes in alertness and autonomic physiology.

• Goodale, S.E., Ahmed, N., Zhao, C., de Zwart, J.A., Özbay, P.S., Picchioni, D., Duyn, J., Englot, D.J., Morgan, V.L., Chang, C., 2021. fMRI-based detection of alertness predicts behavioral response variability. Elife 10. https://doi.org/10.7554%2FeLife.62376
• Bayrak, R.G., Hansen, C.B., Salas, J.A., Ahmed, N., Lyu, I., Huo, Y., Chang, C., 2021. From Brain to Body: Learning Low-Frequency Respiration and Cardiac Signals from fMRI Dynamics. Medical Image Computing and Computer Assisted Intervention – MICCAI 2021. https://doi.org/10.1007/978-3-030-87234-2_52
• Martin, C.G., He, B.J., Chang, C., 2021. State-related neural influences on fMRI connectivity estimation. Neuroimage 244, 118590. https://doi.org/10.1016/j.neuroimage.2021.118590

Deep learning radiological evaluation of epilepsy

Conventional neuroradiological diagnosis of epilepsy still relies on qualitative inspection of clinical MRI. Epilepsy is associated with subtle quantitative abnormalities that have not been leveraged to improve diagnostic accuracy. In this talk, we will discuss applications of deep learning to address this scientific and clinical gap.

• Bonilha, L., Rorden, C., Castellano, G., Pereira, F., Rio, P. A., Cendes, F., & Li, L. M. (2004). Voxel-based morphometry reveals gray matter network atrophy in refractory medial temporal lobe epilepsy. Archives of neurology, 61(9), 1379-1384. https://doi.org/10.1001/archneur.61.9.1379
• Gleichgerrcht, E., Munsell, B., Bhatia, S., Vandergrift III, W. A., Rorden, C., McDonald, C., ... & Bonilha, L. (2018). Deep learning applied to whole‐brain connectome to determine seizure control after epilepsy surgery. Epilepsia, 59(9), 1643-1654. https://doi.org/10.1111/epi.14528
• Gleichgerrcht, E., Munsell, B., Keller, S. S., Drane, D. L., Jensen, J. H., Spampinato, M. V., ... & Bonilha, L. (2022). Radiological identification of temporal lobe epilepsy using artificial intelligence: a feasibility study. Brain communications, 4(2), fcab284. https://doi.org/10.1093/braincomms/fcab284
• Gleichgerrcht, E., Munsell, B. C., Alhusaini, S., Alvim, M. K., Bargalló, N., Bender, B., ... & Wiest, R. (2021). Artificial intelligence for classification of temporal lobe epilepsy with ROI-level MRI data: A worldwide ENIGMA-Epilepsy study. NeuroImage: Clinical, 31, 102765. https://doi.org/10.1016/j.nicl.2021.102765

Gamer in the scanner: imitation of human video gameplay and fMRI brain activity using artificial neural networks

Artificial neural networks can successfully play video games, yet these AI agents have difficulty adapting to changes in the game environment, or transfer knowledge across different games. As human players can efficiently transfer skills across environments, the Courtois NeuroMod team is working to align the representations of artificial neural networks with human players [1]. We first designed and validated a fully MRI-compatible video game controller [2]. The data collected for this project are part of an extremely deep individual fMRI sample currently featuring up to 140 hours of fMRI per subject (N=6), made available for the community as part of the Courtois NeuroMod data bank (https://cneuromod.ca). We successfully trained artificial agents to imitate the actions of humans playing the game “Shinobi III: revenge of the ninja master” and found that the internal representations of the agents could be used to effectively predict individual brain activity measured with functional magnetic resonance imaging [3]. This work could open new avenues to train robust AI video game characters, and gain new insights in brain representations for active and complex stimuli.

Quantitative, Multidimensional MR Spectroscopic Imaging by Integrating Spin Physics and Machine Learning

While neuroimaging has transformed neuroscience by allowing us to map detailed in vivo structural and functional organizations of human brain, technologies to probe the rich molecular complexity and underpinnings of brain functions in vivo are lacking. Magnetic resonance spectroscopic imaging (MRSI) allows for multiplexed molecular imaging and metabolic profiling of the brain in vivo, but its applications have been limited by low sensitivity, poor spatial resolution, slow imaging speed and challenges in separating molecular signals of interest. In this talk, I will discuss our efforts in addressing these challenges. Specifically, I will present our progresses on achieving simultaneous, high-resolution mapping of metabolites, neurotransmitters, and their biophysical parameters using a quantitative multidimensional MRSI approach that builds on and expanding a subspace imaging framework. I will also discuss how we integrate physics-based modeling and machine learning to address the limitations of subspace modeling. Finally, I will discuss our collaborative efforts on clinical translations of the new imaging technology. We expect these developments to create new tools to help better understand the molecular basis of brain function and diseases, improve diagnosis and treatment assessment.

• F. Lam, C. Ma, B. Clifford, C. L. Johnson, Z.-P. Liang. High-resolution 1H-MRSI of the brain using SPICE: Data Acquisition and Image Reconstruction, Magn. Reson. Med., 76:1059-1070, 2017. Link to Paper
• F. Lam, Y. Li, R. Guo, Y. Zhao, B. Clifford, Z.-P. Liang. Ultrafast magnetic resonance spectroscopic imaging using SPICE with learned subspaces, Magn. Reson. Med., 83:377-390, 2020. Link to Paper
• F. Lam, Y. Li, X. Peng. Constrained magnetic resonance spectroscopic imaging by learning nonlinear low-dimensional models, IEEE Trans. Med. Imaging, 39:545-555, 2020. Link to Paper
• Z. Wang, Y. Li, F. Lam, High-resolution, 3D multi-TE 1H-MRSI using fast spatiospectral encoding and subspace imaging, Magn. Reson. Med., 87:1103-1118, 2022Link to Paper

Topological inference for brain networks in temporal lobe epilepsy using the Wasserstein distance

Persistent homology summarizes the changes of topological structures in data through over multiple scales called filtrations. Doing so detect hidden topological signals that persist over different scales. However, a key obstacle of applying persistent homology to brain networks has been the lack of robust statistical inference framework. To address this problem, we present a new topological inference procedure based on the Wasserstein distance. Our approach has no explicit models and statistical distributional assumptions. The inference is performed in a completely data driven fashion. Our metric-based inference significantly differs from traditional feature-based topological data analysis (TDA). The method is applied to the resting-state functional magnetic resonance images (rs-fMRI) of the temporal lobe epilepsy patients and able to localize brain regions that contribute the most to topological differences. We made computer code available at Github . The talk is based on Anand and Chung 2023, IEEE Transactions on Medical Imaging (arXiv:2110.14599) and Songdechakraiwut and Chung 2023 Annals of Applied Statistics (arXiv:2012.00675).

• D. V. Anand and M. K. Chung, Hodge Laplacian of Brain Networks, IEEE Transactions on Medical Imaging, 2023 Link to Paper
• T. Songdechakraiwut and M. K. Chung, Topological learning for brain networks, The Annals of Applied Statistics, vol. 17, no. 1, pp. 403-433, 2023. Link to Paper

From searching for lesions to exploring distributed circuitopathy in psychosis.

Researchers have sought for more than 100 years for a lesion (or lesions) that might lead to schizophrenia, the major psychotic disorder. Early imaging modalities revealed enlarged ventricles, which imply reduced brain volumes, but no characteristic cortical pathology was identified. High-resolution structural magnetic resonance imaging (MRI), finally provided a tool to identify cortical volume loss in vivo. Volumetric studies revealed not only initial gray matter loss in temporal and frontal cortices, but progressive loss after the emergence of psychosis. Advances in 3D microscopy in post-mortem studies indicated increased packing density of cortical pyramidal cells and reduced dendritic arborization in schizophrenia, leading to the idea that the disease reflected dendrotoxicity. Although specific areas (such as frontal and temporal cortex) showed more robust volume loss, the idea that schizophrenia reflected a dysconnectivity among brain regions emerged. Volumetric studies suggested that the areas with greatest loss were areas with the most extensive connectivity – heteromodal cortices or areas that served as hubs. Recently, thought has focused on structural and functional dysconnectivity in schizophrenia, with the emerging concept of the disorder being one of circuitopathy. Decomposing source-activity from high-temporal resolution methods such as EEG and MEG into spectral components allows for functional and effective connectivity measures between areas. Methods for spectral effective connectivity still need development and validation. Issues include reliable and valid source localization, methods for data reduction (particularly parcellation), identification of critical frequency bands underlying inter-areal communication, and derivation of networks from whole brain data. Data and approaches from my laboratory will be used to illustrate our approaches, and how spectral connectivity indicates alpha-band and theta-band dysconnectivity between cortical areas are central deficits in psychosis, even early in disease course at the emergence of psychosis.

• Sklar AL, Coffman BA, Longenecker JM, Curtis M, Salisbury DF. Load-dependent functional connectivity deficits during visual working memory in first-episode psychosis. J Psychiatr Res. 2022. Link to Paper
• Coffman BA, Haas G, Olson C, Cho R, Ghuman AS, Salisbury DF. Reduced Dorsal Visual Oscillatory Activity During Working Memory Maintenance in the First-Episode Schizophrenia Spectrum. Front Psychiatry. 2020. Link to Paper
• Curtis MT, Sklar AL, Coffman BA, Salisbury DF. Functional connectivity and gray matter deficits within the auditory attention circuit in first-episode psychosis. Front Psychiatry. 2023. Link to Paper
• Curtis MT, Ren X, Coffman BA, Salisbury DF. Attentional M100 gain modulation localizes to auditory sensory cortex and is deficient in first-episode psychosis. Hum Brain Mapp. 2023. Link to Paper
• Phalen H, Coffman BA, Ghuman A, Sejdić E, Salisbury DF. Non-negative Matrix Factorization Reveals Resting-State Cortical Alpha Network Abnormalities in the First-Episode Schizophrenia Spectrum. Biol Psychiatry Cogn Neurosci Neuroimaging. 2020. Link to Paper

Neuroimaging of affect and motivation: Pathways, models, and interventions.

Affect and motivation — pleasure and pain, desire and threat — are central to human life. Their experience defines our wellbeing, and the brain processes that underlie them drive behavior and learning. Developing models of the brain circuits that underlie them, and how they interact, could transform how we understand and measure them, and provide biological targets for interventions ranging from drugs to psychotherapy. Neuroimaging, including Functional Magnetic Resonance Imaging, is playing a transformative role in our ability to model the brain bases of affective and motivational processes. However, developing such models will require computational advances, particularly in our ability to model how emergent properties like pain arise from complex interactions among brain systems. In this talk, I describe a series of studies that combine fMRI imaging with statistical and machine learning to develop measures that are sensitive and specific for particular types of pain and affect and generalizable across diverse populations. These studies provide a brain-based picture of the organization of pain and affect, revealing both distinctions and similarities that are not predicted by folk psychological theories. We find that on one hand, pain and other affective states are distributed, relying on interactions across multiple brain systems. At the same time, however, new techniques provide ways of decomposing these systems into particular pathways that can be referenced to animal models and targeted by interventions.

Neural population coding.

In this talk I will present our work, based on the development and applications of mathematical methods from Information Theory and Machine Learning, to study how the functions of neural population codes emerge form the interaction of different neurons.

• Runyan, C. A., Piasini, E., Panzeri, S. & Harvey, C. D. Distinct timescales of population coding across cortex. Nature 548 (2017). Link to Paper
• Valente, M. et al. Correlations enhance the behavioral readout of neural population activity in association cortex. Nat Neurosci 24, 975-986 (2021) Link to Paper
• Panzeri, S., Moroni, M., Safaai, H. & Harvey, C. D. The structures and functions of correlations in neural population codes. Nat Rev Neurosci 23, 551-567,(2022). Link to Paper

Understand the Brain Using Interpretable Machine Learning Models

Computational neuroscience is a burgeoning field embracing exciting scientific questions, a deluge of data, an imperative demand for quantitative models, and a close affinity with artificial intelligence. These opportunities promote the advancement of data-driven machine learning methods to help neuroscientists deeply understand our brains. In particular, my work lies in such an interdisciplinary field and spans the development of scientifically-motivated probabilistic modeling approaches for neural and behavior analyses. In this talk, I will first present my work on developing Bayesian methods to identify latent manifold structures with applications to neural recordings in multiple cortical areas. The models are able to reveal the underlying signals of neural populations as well as uncover interesting topography of neurons where there is a lack of knowledge and understanding about the brain. Discovering such low-dimensional signals or structures can help shed light on how information is encoded at the population level, and provide significant scientific insight into the brain. Next, I will talk about probabilistic priors that encourage region-sparse activation for brain decoding. The proposed model provides spatial decoding weights for brain imaging data that are both more interpretable and achieve higher decoding performance. Finally, I will introduce a series of works on semi-supervised learning for animal behavior analysis and understanding. I will show that when we have a very limited amount of human-labeled data, the semi-supervised learning frameworks can well resolve the scarce data issue by leveraging both labeled and unlabeled data in the context of pose tracking, video understanding, and behavioral segmentation. By actively working on both neural and behavioral studies, I hope to develop interpretable machine learning and Bayesian statistical approaches to understanding neural systems integrating extensive and complex behaviors, thus providing a systematic understanding of neural mechanisms and biological functions.

• Wu, Anqi, et al. Gaussian process based nonlinear latent structure discovery in multivariate spike train data. Advances in neural information processing systems 30 (2017). Link to Paper
• Wu, Anqi, et al. Learning a latent manifold of odor representations from neural responses in piriform cortex. Advances in Neural Information Processing Systems 31 (2018). Link to Paper
• Wu, Anqi, Oluwasanmi Koyejo, and Jonathan W. Pillow. Dependent relevance determination for smooth and structured sparse regression. J. Mach. Learn. Res. 20.89 (2019). Link to Paper
• Wu, Anqi, et al. Sparse Bayesian structure learning with dependent relevance determination priors. Advances in Neural Information Processing Systems 27 (2014). Link to Paper
• Wu, Anqi, et al. Deep Graph Pose : a semi-supervised deep graphical model for improved animal pose tracking. Advances in Neural Information Processing Systems 33 (2020) Link to Paper
• Blau, Ari, et al. SemiMultiPose : A Semi-supervised Multi-animal Pose Estimation Framework. arXiv preprint arXiv:2204.07072 (2022). Link to Paper
• Whiteway, Matthew R., et al. Partitioning variability in animal behavioral videos using semi-supervised variational autoencoders. PLoS computational biology 17.9 (2021) Link to Paper
• Whiteway, Matthew R., et al. Semi-supervised sequence modeling for improved behavioral segmentation. bioRxiv (2021) Link to Paper

Quantifying the idiosyncrasy of brain activations

Research in neuroscience often assumes universal neural mechanisms, but increasing evidence points towards sizeable individual differences in brain activations. What remains unclear is the extent of the idiosyncrasy and whether different types of analyses are associated with different levels of idiosyncrasy. Here we develop a new method for addressing these questions. The method consists of computing the within-subject reliability and subject-to-group similarity of brain activations and submitting these values to a computational model that quantifies the relative strength of group- and subject-level factors. We apply this method to a perceptual decision-making task and find that activations related to trial-level task, reaction time (RT), and confidence are influenced equally strongly by group- and subject-level factors. However, for activations related to average RT or confidence in a block of trials, the subject-level factors can be up to 6 times more important than group-level factors. In all cases, group- and subject-level factors are dwarfed by a noise factor. Overall, our method allows for the quantification of group- and subject-level factors of brain activations and thus provides a more detailed understanding of the idiosyncrasy levels in brain activations.

• J. Nakuci, J. Yeon, K. Xue, J.H. Kim, S.P. Kim, and D. Rahnev, Quantifying the contribution of subject and group factors in brain activation, bioRxiv, p. 2022.08.01.502338, 2023. Link to Paper

A Regularized Blind Source Separation Framework for Unveiling Hidden Sources of Brain Functional Connectome

Brain connectomics has become increasingly important in neuroimaging studies to advance understanding of neural circuits and their association with neurodevelopment, mental illnesses, and aging. These analyses often face major challenges, including the high dimensionality of brain networks, unknown latent sources underlying the observed connectivity, and the large number of brain connections leading to spurious findings. In this talk, we will introduce a novel regularized blind source separation (BSS) framework for reliable mapping of neural circuits underlying static and dynamic brain functional connectome. The proposed LOCUS methods achieve more efficient and reliable source separation for connectivity matrices using low-rank factorization, a novel angle-based sparsity regularization, and a temporal smoothness regularization. We develop a highly efficient iterative Node-Rotation algorithm that solves the non-convex optimization problem for learning LOCUS models. Simulation studies demonstrate that the proposed methods have consistently improved accuracy in retrieving latent connectivity traits. Application of LOCUS methods to the Philadelphia Neurodevelopmental Cohort (PNC) neuroimaging study generates considerably more reproducible findings in revealing underlying neural circuits and their association with demographic and clinical phenotypes, uncovers dynamic expression profiles of the circuits and the synchronization between them, and generates insights on gender differences in the neurodevelopment of brain circuits.

• Wang, Y. and Guo, Y. (2023). LOCUS: A regularized blind source separation method with low-rank structure for investigating brain connectivity. Annals of Applied Statistics, 17(2): 1307-1332. Link to Paper

Pediatric Population Neuroimaging and the Exploration of Stochastic Processes in Brain Development

The prenatal period of life is a time of considerable brain development as the brain emerges from a single cell to a brain that very much resembles an adult brain by the time of birth. This period of extensive growth is also a critical period where environmental factors, such as nutrition, cannabis, cigarette smoking, medication, and other factors can influence brain development. Following birth, the brain undergoes continued brain development and postnatal factors can also have an influence on optimal brain development. This talk will describe the effects of environmental factors during the prenatal and early postnatal period on the developing brain, including the likely role of stochastic processes. During prenatal life the effects of nutrition, substance use, medications, and other environmental exposures on brain structure and function will be discussed within the context of a large, population-based study of child development. An overview of how stochastic events can influence brain development as well as an important hypothesis of how optimizing neurodevelopment during fetal life may, within a population-context, prevent the emergence of psychopathology.

• White TJH. Brain Development and Stochastic Processes During Prenatal and Early Life: You Can't Lose It if You've Never Had It; But It's Better to Have It and Lose It, Than Never to Have Had It at All. J Am Acad Child Adolesc Psychiatry. 2019;58(11):1042-1050. Link to Paper
• White, T., Muetzel, R.L., El Marroun, H. et al. Paediatric population neuroimaging and the Generation R Study: the second wave. Eur J Epidemiol 33, 99–125 (2018). Link to Paper
• Gärtner K. A third component causing random variability beside environment and genotype. A reason for the limited success of a 30 year long effort to standardize laboratory animals? Laboratory Animals. 1990;24(1):71-77. Link to Paper
• Gerit Arne Linneweber et al. ,A neurodevelopmental origin of behavioral individuality in the Drosophila visual system.Science367,1112-1119(2020). Link to Paper
• White, Tonya. Behavioral Phenotypes, Stochastic Processes, Entropy, Evolution, and Individual Variability: Toward a Unified Field Theory for Neurodevelopment and Psychopathology. Aperture Neuro, vol. BWAS Editorials, Dec. 2022 Link to Paper

Understanding disease heterogeneity via weakly-supervised machine learning: Focus on neuropsychiatric and neurodegenerative diseases

Neuroimaging has significantly expanded our understanding of brain changes in neuropsychiatric disorders as well as in aging and neurodegenerative diseases. However, it wasn’t until the advent of machine learning tools that imaging signatures that can be detected in individuals, rather than groups, were constructed. This talk will present work on deriving imaging signatures of diagnostic and predictive value. It will then focus on weakly-supervised machine learning methods for analysis of the heterogeneity of brain imaging phenotypes, arriving at a dimensional representation reflecting the heterogeneity of brain aging and of various brain diseases. Finally, international consortia pooling and harmonizing large numbers of brain MRIs from many studies are presented as means for creating sufficiently large datasets for robust machine learning training and heterogeneity analysis, but also pose new challenges, including that or harmonization across studies.

• Yang, Z., Wen, J., Abdulkadir, A. et al. Gene-SGAN: discovering disease subtypes with imaging and genetic signatures via multi-view weakly-supervised deep clustering. Nat Commun 15, 354 (2024). Link to Paper

Machine learning paradigms towards single subject prediction

The nature of mental illness remains a conundrum. Traditional disease categories are increasingly suspected to misrepresent the causes underlying mental disturbance. Yet psychiatrists and investigators now have an unprecedented opportunity to benefit from complex patterns in brain, behavior, and genes using methods from machine learning (e.g., support vector machines, modern neural-network algorithms, cross-validation procedures). Combining these analysis techniques with a wealth of data from consortia and repositories has the potential to advance a biologically grounded redefinition of major psychiatric disorders. Increasing evidence suggests that data-derived subgroups of psychiatric patients can better predict treatment outcomes than DSM/ICD diagnoses can. In a new era of evidence-based psychiatry tailored to single patients, objectively measurable endophenotypes could allow for early disease detection, individualized treatment selection, and dosage adjustment to reduce the burden of disease. This primer aims to introduce clinicians and researchers to the opportunities and challenges in bringing machine intelligence into psychiatric practice.

• Bzdok D, Thieme A, Levkovskyy O, Wren P, Ray T, Reddy S. Data science opportunities of large language models for neuroscience and biomedicine. Neuron. 2024. Link to Paper
• Schulz MA, Bzdok D, Haufe S, Haynes JD, Ritter K. Performance reserves in brain-imaging-based phenotype prediction. Cell Rep. 2024. Link to Paper
• Bzdok D, Meyer-Lindenberg A. Machine Learning for Precision Psychiatry: Opportunities and Challenges. Biol Psychiatry Cogn Neurosci Neuroimaging. 2018. Link to Paper

Can the Human Connectome guide Brain Stimulation?

Invasive and noninvasive brain stimulation methods are applied to focal points in the depth or on the surface of the brain. However, their focal application leads to network effects that are distributed across the entire brain. We can study network effects of focal brain stimulation by pairing them with the human connectome. By doing so, we may investigate which networks need to be stimulated to observe a specific effect. Moreover, we can use brain stimulation sites to segregate the human connectome into functional networks, each tied to specific behaviors, clinical signs or symptoms. One particularly useful method is deep brain stimulation, an invasive neurosurgical procedure that applies highly localized but strong stimulation signals onto specific subcortical areas. In this talk, I will review connectomic effects of deep brain stimulation and other brain stimulation methods. We will cover results in diseases ranging from the movement disorders spectrum (Parkinson’s Disease, Dystonia, Essential Tremor) to neuropsychiatric (Tourette’s & Alzheimer’s Disease) and psychiatric (Obsessive Compulsive Disorder, Depression) diseases. I will also demonstrate how findings in seemingly different diseases (such as Parkinson’s Disease and Depression) could be transferred to cross-inform one another and how the same method can be used to study neurocognitive effects, such as risk-taking behavior or impulsivity.

• Hollunder, B., Ostrem, J.L., Sahin, I.A. et al. Mapping dysfunctional circuits in the frontal cortex using deep brain stimulation. Nat Neurosci 27, 573-586 (2024). Link to Paper
• Li, N., Baldermann, J.C., Kibleur, A. et al. A unified connectomic target for deep brain stimulation in obsessive-compulsive disorder. Nat Commun 11, 3364 (2020). Link to Paper
• Ríos, A.S., Oxenford, S., Neudorfer, C. et al. Optimal deep brain stimulation sites and networks for stimulation of the fornix in Alzheimer's disease. Nat Commun 13, 7707 (2022). Link to Paper

Biologically Inspired Deep Learning as a New Window into Brain Dysfunction

Deep learning has disrupted nearly every major field of study from computer vision to genomics. The unparalleled success of these models has, in many cases, been fueled by an explosion of data. Millions of labeled images, thousands of annotated ICU admissions, and hundreds of hours of transcribed speech are common standards in the literature. Clinical neuroscience is a notable holdout to this trend. It is a field of unavoidably small datasets, massive patient variability, and complex (largely unknown) phenomena. My lab tackles these challenges across a spectrum of projects, from answering foundational neuroscientific questions to translational applications of neuroimaging data to exploratory directions for probing neural circuitry. One of our key strategies is to integrate a priori information about the brain and biology into the model design. This talk will highlight two ongoing projects that epitomize this strategy. First, I will showcase an end-to-end deep learning framework that fuses neuroimaging, genetic, and phenotypic data, while maintaining interpretability of the extracted biomarkers. We use a learnable dropout layer to extract a sparse subset of predictive imaging features and a biologically informed deep network architecture for whole-genome analysis. Specifically, the network uses hierarchical graph convolution that mimic the organization of a well-established gene ontology to track the convergence of genetic risk across biological pathways. Second, I will present a deep-generative hybrid model for epileptic seizure detection from scalp EEG. The latent variables in this model capture the spatiotemporal spread of a seizure; they are complemented by a nonparametric likelihood based on convolutional neural networks. I will also highlight our current end-to-end extensions of this work focused on seizure onset localization.

• Ghosal, Sayan, Qiang Chen, Giulio Pergola, Aaron L. Goldman, William Ulrich, Daniel R. Weinberger, and Archana Venkataraman. A biologically interpretable graph convolutional network to link genetic risk pathways and neuroimaging markers of disease. Link to Paper
• Ghosal, Sayan, Qiang Chen, Giulio Pergola, Aaron L. Goldman, William Ulrich, Karen F. Berman, Giuseppe Blasi et al. G-MIND: an end-to-end multimodal imaging-genetics framework for biomarker identification and disease classification. In Medical Imaging 2021, Image Processing, vol. 11596, pp. 63-72. SPIE, 2021. Link to Paper
• Craley, Jeff, Emily Johnson, and Archana Venkataraman. A spatio-temporal model of seizure propagation in focal epilepsy. IEEE transactions on medical imaging 39, no. 5 (2019): 1404-1418. Link to Paper
• Craley, Jeff, Emily Johnson, Christophe C. Jouny, David Hsu, Raheel Ahmed, and Archana Venkataraman. SZLoc: A multi-resolution architecture for automated epileptic seizure localization from scalp EEG. Medical Imaging with Deep Learning (2022). Link to Paper