Biology subjects

Plis, S.

Publications and source records attributed to Plis, S..

12 recordsLinked to original sources

bioRxiv2026

Recovering directional brain networks under temporal undersampling: an application to schizophrenia

Two decades of functional connectivity research have established schizophrenia as a disorder of distributed dysconnectivity, with a robust thalamocortical signature: reduced prefrontal and increased sensory coupling. A fundamental issue is that functional connectivity is undirected, operates at a single slow timescale, and cannot reveal causal direction. Moreover, the mismatch between BOLD sampling speed and neural dynamics can hide edges, fabricate spurious ones, and reverse the apparent orientation of causal relationships. To overcome these limitations, we introduce a general framework for estimating directed causal graphs from fMRI that explicitly accounts for temporal undersampling. We apply RnR, a causal discovery method built on the rate-agnostic RASL framework, to resting-state fMRI from the multi-site FBIRN cohort. Rather than returning a single directed graph, RnR recovers an equivalence class of directed graphs consistent with the observed data, each annotated with the sampling rate that would produce it and classifies each estimated orientation by its stability across inferred rates. This provides a principled basis for distinguishing directed interpretations that are safe to trust from those that are timescale-contingent. Benchmarking against five single-timescale estimators on schizophrenia data, RnR recovered substantially more group-differentiating directed edges, reproducing the fields most replicated finding in directed form: a sensory-to-visual hyperconnectivity hub oriented from the post-central gyrus component to primary visual cortex. Through simulation, we show that coarse spatial resolution has shielded single-timescale methods from the full effects of undersampling, whereas finer parcellations will require explicit undersampling modeling. This work reframes fMRI effective connectivity estimation by treating undersampling as a fundamental property of the measurement, enabling directed interpretations that are grounded in the data-generating process.

neuroscience↗

bioRxiv2024

st-DenseViT: A Weakly Supervised Spatiotemporal Vision Transformer for Dense Prediction of Dynamic Brain Networks

ObjectiveModeling dynamic neuronal activity within brain networks enables the precise tracking of rapid temporal fluctuations across different brain regions. However, current approaches in computational neuroscience fall short of capturing and representing the spatiotemporal dynamics within each brain network. We developed a novel weakly supervised spatiotemporal dense prediction model capable of generating personalized 4D dynamic brain networks from fMRI data, providing a more granular representation of brain activity over time. MethodsWe developed a model that leverages the vision transformer (ViT) as its backbone, jointly encoding spatial and temporal information from fMRI inputs using two different configurations: space-time and sequential encoders. The model generates 4D brain network maps that evolve over time, capturing dynamic changes in both spatial and temporal dimensions. In the absence of ground-truth data, we used spatially constrained windowed independent component analysis (ICA) components derived from fMRI data as weak supervision to guide the training process. The model was evaluated using large-scale resting-state fMRI datasets, and statistical analyses were conducted to assess the effectiveness of the generated dynamic maps using various metrics. ResultsOur model effectively produced 4D brain maps that captured both inter-subject and temporal variations, offering a dynamic representation of evolving brain networks. Notably, the model demonstrated the ability to produce smooth maps from noisy priors, effectively denoising the resulting brain dynamics. Additionally, statistically significant differences were observed in the temporally averaged brain maps, as well as in the summation of absolute temporal gradient maps, between patients with schizophrenia and healthy controls. For example, within the Default Mode Network (DMN), significant differences emerged in the temporally averaged space-time configurations, particularly in the thalamus, where healthy controls exhibited higher activity levels compared to subjects with schizophrenia. These findings highlight the models potential for differentiating between clinical populations. ConclusionThe proposed spatiotemporal dense prediction model offers an effective approach for generating dynamic brain maps by capturing significant spatiotemporal variations in brain activity. Leveraging weak supervision through ICA components enables the model to learn dynamic patterns without direct ground-truth data, making it a robust and efficient tool for brain mapping. SignificanceThis work presents an important new approach for dynamic brain mapping, potentially opening up new opportunities for studying brain dynamics within specific networks. By framing the problem as a spatiotemporal dense prediction task in computer vision, we leverage the spatiotemporal ViT architecture combined with weakly supervised learning techniques to efficiently and effectively estimate these maps.

neuroscience↗

bioRxiv2024

Privacy-Preserving Visualization of Brain Functional Connectivity

Data visualizations are an integral part of neuroimging research, supporting activities ranging from exploratory data analysis to the interpretation and communication of findings. While essential, visualizations can also reveal private information about individual participants. In this paper, we discuss how visualizations may inadvertently lead to privacy leakage and explore methods to mitigate such risks. Our work investigates ways to securely share visualizations that faithfully preserve the patterns supporting the derived insights from data analysis, rather than deriving conclusions from the visualizations themselves. We address the problem of privacy-preserving visualization under the framework of differential privacy, focusing on commonly used visualization methods for functional network connectivity. Several perturbation-based strategies are investigated for protecting correlationrelated measures, with analyses of their privacy costs and the effects of pre- and post-processing. To achieve a better balance between privacy and visual utility, we propose workflows for connectogram and seed-based connectivity visualizations that preserve the qualitative structure of non-private results. Overall, this work illustrates how differential privacy can be effectively applied to neuroimaging visualization, highlighting its potential as a principled approach for safeguarding sensitive information.

neuroscience↗

bioRxiv2024

Efficient Federated Learning for distributed NeuroImaging Data

Recent advancements in neuroimaging have led to greater data sharing among the scientific community. However, institutions frequently maintain control over their data, citing concerns related to research culture, privacy, and accountability. This creates a demand for innovative tools capable of analyzing amalgamated datasets without the need to transfer actual data between entities. To address this challenge, we propose a decentralized sparse federated learning (FL) strategy. This approach emphasizes local training of sparse models to facilitate efficient communication within such frameworks. By capitalizing on model sparsity and selectively sharing parameters between client sites during the training phase, our method significantly lowers communication overheads. This advantage becomes increasingly pronounced when dealing with larger models and accommodating the diverse resource capabilities of various sites. We demonstrate the effectiveness of our approach through the application to the Adolescent Brain Cognitive Development (ABCD) dataset.

neuroscience↗

bioRxiv2023

Using an ODE model to separate Rest and Task signals in fMRI

Cortical activity results from the interplay between network-connected regions that integrate information and stimulus-driven processes originating from sensory motor networks responding to specific tasks. Separating the information due to each of these components has been challenging, and the relationship as measured by fMRI in each of these cases Rest (network) and Task (stimulus-driven) remains a significant open question in the study of large-scale brain dynamics. In this study, we develop a network ordinary differential equation (ODE) model using advanced system identification tools to analyze fMRI data from both rest and task conditions. We demonstrate that task-specific ODEs are essentially a subset of rest-specific ODEs across four different tasks from the Human Connectome Project. By assuming that task activity is a relative complement of rest activity, our model significantly improves predictions of reaction times on a trial-by-trial basis, leading to a 9 % increase in explanatory power (R2) across the 14 sub-tasks tested. We have additionally shown that these results hold for predicting missing trials and accuracy on a per individual basis as well as classifying Tasks trajectories or resulting dynamic Task functional connectivity. Our findings establish the principle of the Active Cortex Model, which posits that the cortex is always active and that Rest State encompasses all processes, while certain subsets of processes get elevated to perform specific task computations. Thus, this study is an important milestone in the development of the fMRI equation - to causally link large-scale brain activity, brain structural connectivity, and behavioral variables within a single framework.

neuroscience↗

bioRxiv2023

Enhancing Collaborative Neuroimaging Research: Introducing COINSTAC Vaults for Federated Analysis and Reproducibility

Collaborative neuroimaging research is often hindered by technological, policy, administrative, and methodological barriers, despite the abundance of available data. COINSTAC is a platform that successfully tackles these challenges through federated analysis, allowing researchers to analyze datasets without publicly sharing their data. This paper presents a significant enhancement to the COINSTAC platform: COINSTAC Vaults (CVs). CVs are designed to further reduce barriers by hosting standardized, persistent, and highly-available datasets, while seamlessly integrating with COINSTACs federated analysis capabilities. CVs offer a user-friendly interface for self-service analysis, streamlining collaboration and eliminating the need for manual coordination with data owners. Importantly, CVs can also be used in conjunction with open data as well, by simply creating a CV hosting the open data one would like to include in the analysis, thus filling an important gap in the data sharing ecosystem. We demonstrate the impact of CVs through several functional and structural neuroimaging studies utilizing federated analysis showcasing their potential to improve the reproducibility of research and increase sample sizes in neuroimaging studies.

neuroscience↗

bioRxiv2021

A DEEP LEARNING APPROACH TO ESTIMATING INITIAL CONDITIONS OF BRAIN NETWORK MODELS IN REFERENCE TO MEASURED FMRI DATA

1.Brain Network Models (BNMs) are a family of dynamical systems that simulate whole brain activity using neural mass models to represent local activity in different brain regions that influence each other via a global structural network. Research has been interested in using these network models to explain measured whole brain activity measured via resting state functional magnetic resonance imaging (rs-fMRI). Properties computed over longer periods of simulated and measured data such as average functional connectivity (FC), have shown to be comparable with similar properties estimated from measured rs-fMRI data. While this shows that these network models have similar properties over the dynamical landscape, it is unclear how well simulated trajectories compare with empirical trajectories on a timepoint-by-timepoint basis. Previous studies have shown that BNMs are able to produce relevant features at shorter timescales, but analysis of short-term trajectories or transient dynamics as defined by synchronized predictions from BNM made at the same timescale as the collected data has not yet been conducted. Relevant neural processes exist in the time frame of measurements and are often used in task fMRI studies to understand neural responses to behavioral cues. Therefore, it is important to investigate how much of these dynamics are captured by our current brain simulations. To test the nature of BNMs short term trajectories against observed data, we utilize a deep learning technique known as Neural ODE that based on an observed sequence of fMRI measurements, estimates the initial conditions such that the BNMs simulation is synchronized to produce the closest trajectory relative to the observed data. We test to see if the parameterization of a specific well studied BNM, the Firing Rate Model, calculated by maximizing its accuracy in reproducing observed short term trajectories matches with the parameterized model that produces the best average long-term metrics. Our results show that such an agreement between parameterization using long and short simulation analysis exists if also considering other factors such as the sensitivity in accuracy with relative to changes in structural connectivity. Therefore, we conclude that there is evidence that by solving for initial conditions, BNMs can be simulated in a meaningful way when comparing against measured data trajectories, although future studies are necessary to establish how BNM activity relate to behavioral variables or to faster neural processes during this time period.

neuroscience↗

bioRxiv2020

A classification-based approach to estimate the number of resting fMRI dynamic functional connectivity states

Recent work has focused on the study of dynamic (vs static) brain connectivity in resting fMRI data. In this work, we focus on temporal correlation between time courses extracted from coherent networks or components called functional network connectivity (FNC). Dynamic functional network connectivity (dFNC) is most commonly estimated using a sliding window-based approach to capture short periods of FNC change. These data are then clustered to estimate transient connectivity patterns or states. Determining the number of states is a challenging problem. The elbow criterion is a widely used approach to determine the optimal number of states. In our work, we present an alternative approach that evaluates classification (e.g. healthy controls versus patients) as a measure to select the optimal number of states (clusters). We apply different classification strategies to perform classification between healthy controls (HC) and patients with schizophrenia (SZ) for different numbers of states (i.e. varying the model order in the clustering algorithm). We compute cross-validated accuracy for different model orders to evaluate the classification performance. Our results are consistent with our earlier work which shows that overall accuracy improves when dynamic connectivity measures are used separately or in combination with static connectivity measures. Results also show that the optimal model order for classification is different from that using the standard k-means model selection method and that such optimization improves resulting in cross-validated accuracy. The optimal model order obtained from the proposed approach also gives significantly improved classification performance over the traditional model selection method. In sum, the observed results suggest that if ones goal is to perform classification, using the proposed approach as a criterion for selecting the optimal number of states in dynamic connectivity analysis leads to improved accuracy in hold-out data.

neuroscience↗

bioRxiv2020

Addressing Inaccurate Nosology in Mental Health: A Multi Label Data Cleansing Approach for Detecting Label Noise from Structural Magnetic Resonance Imaging Data in Mood and Psychosis Disorders

BackgroundMental health diagnostic approaches are seeking to identify biological markers to work alongside advanced machine learning approaches. It is difficult to identify a biological marker of disease when the traditional diagnostic labels themselves are not necessarily valid. MethodsWe worked with T1 structural magnetic resonance imaging data collected from individuals with mood and psychosis disorders from over 1400 individuals comprising healthy controls, psychosis patients and their unaffected first-degree relatives including 176 bipolar probands, 134 schizoaffective probands, 240 schizophrenia proband, 581 patients relatives and 362 controls. We assumed there might be noise in the diagnostic labeling process. We detected label noise by classifying the data multiple times using a support vector machine classifier, and then we flagged those individuals in which all classifiers unanimously mislabeled those subjects. Next, we assigned a new diagnostic label to these individuals, based on the biological data (MRI), using iterative data cleansing approach. ResultsSimulation results showed our method was highly accurate in identifying label noise. Both diagnostic and Biotype categories showed about 65% and 63% respectively of noisy labels with the largest amount of relabeling occurring between the healthy control and bipolar and schizophrenia disorder individuals as well as in the unaffected close relatives. The extraction of imaging features highlighted regional brain changes associated with each group. ConclusionsThis approach represents an initial step towards developing strategies that need not assume existing mental health diagnostic categories are always valid, but rather allows us to leverage this information while also acknowledging that there are misassignments.

neuroscience↗

bioRxiv2020

Hype versus hope: Deep learning encodes more predictive and robust brain imaging representations than standard machine learning

Previous successes of deep learning (DL) approaches on several complex tasks have hugely inflated expectations of their power to learn subtle properties of complex brain imaging data, and scale to large datasets. Perhaps as a reaction to this inflation, recent critical commentaries unfavorably compare DL with standard machine learning (SML) approaches for the analysis of brain imaging data. Yet, their conclusions are based on pre-engineered features which deprives DL of its main advantage: representation learning. Here we evaluate this and show the importance of representation learning for DL performance on brain imaging data. We report our findings from a large-scale systematic comparison of SML approaches versus DL profiled in a ten-way age and gender-based classification task on 12,314 structural MRI images. Results show that DL methods, if implemented and trained following the prevalent DL practices, have the potential to substantially improve compared to SML approaches. We also show that DL approaches scale particularly well presenting a lower asymptotic complexity in relative computational time, despite being more complex. Our analysis reveals that the performance improvement saturates as the training sample size grows, but shows significantly higher performance throughout. We also show evidence that the superior performance of DL is primarily due to the excellent representation learning capabilities and that SML methods can perform equally well when operating on representations produced by the trained DL models. Finally, we demonstrate that DL embeddings span a comprehensible projection spectrum and that DL consistently localizes discriminative brain biomarkers, providing an example of the robustness of prediction relevance estimates. Our findings highlight the presence of non-linearities in brain imaging data that DL frameworks can exploit to generate superior predictive representations for characterizing the human brain, even with currently available data sizes.

neuroscience↗

bioRxiv2019

Decentralized Distribution-sampled Classification Models with Application to Brain Imaging

0.1 backgroundIn this age of big data, large data stores allow researchers to compose robust models that are accurate and informative. In many cases, the data are stored in separate locations requiring data transfer between local sites, which can cause various practical hurdles, such as privacy concerns or heavy network load. This is especially true for medical imaging data, which can be constrained due to the health insurance portability and accountability act (HIPAA). Medical imaging datasets can also contain many thousands or millions of features, requiring heavy network load.\n\n0.2 New MethodOur research expands upon current decentralized classification research by implementing a new singleshot method for both neural networks and support vector machines. Our approach is to estimate the statistical distribution of the data at each local site and pass this information to the other local sites where each site resamples from the individual distributions and trains a model on both locally available data and the resampled data.\n\n0.3 ResultsWe show applications of our approach to handwritten digit classification as well as to multi-subject classification of brain imaging data collected from patients with schizophrenia and healthy controls. Overall, the results showed comparable classification accuracy to the centralized model with lower network load than multishot methods.\n\n0.4 Comparison with Existing MethodsMany decentralized classifiers are multishot, requiring heavy network traffic. Our model attempts to alleviate this load while preserving prediction accuracy.\n\n0.5 ConclusionsWe show that our proposed approach performs comparably to a centralized approach while minimizing network traffic compared to multishot methods.\n\n0.6 HighlightsO_LIA novel yet simple approach to decentralized classification\nC_LIO_LIReduces total network load compared to current multishot algorithms\nC_LIO_LIMaintains a prediction accuracy comparable to the centralized approach\nC_LI

bioinformatics↗

bioRxiv2018

Deep Residual Learning for Neuroimaging: An application to Predict Progression to Alzheimer’s Disease

This work investigates the suitability of deep residual neural networks (ResNets) for studying neuroimaging data in the specific application of predicting progression from mild cognitive impairment (MCI) to Alzheimers disease (AD). We focus on predicting the subset of MCI individuals that would progress to AD within three years (progressive MCI) and the subset of MCI individuals that do not progress to AD within this period (stable MCI). This prediction was conducted first as a standard binary classification task by training a ResNet architecture using MCI individuals only, followed by a modified domain transfer learning version that additionally trained on the AD and cognitively normal (CN) individuals. For this modified inter-MCI classification task, the ResNet architecture achieved better than state-of-the-art performance in predicting progression to AD using structural MRI data alone (> 7% than the second best performing method), within 1% of the state-of-the-art performance considering learning using multiple structural modalities as well, and a significant performance improvement over the classical support vector machine learning framework (p = 2.5762 x 10-8). The learnt surrogate (i.e. cross-validation-fold-specific) models and the final predictive model (trained on all data) in this modified classification task showed highly similar peak activations, significant correspondence of which in the medial temporal lobe and other areas could be established with previous reports in AD literature, thus further validating our findings. Our results highlight the possibility of early identification of modifiable risk factors for understanding progression to AD using similar advanced deep learning architectures.

neuroscience↗