Ahmed M. El‐Gazzar

Ahmed ElGazzar, Rajat Thomas, Guido van Wingen

Graph Neural Networks (GNNs) have emerged as a powerful tool to learn from graph-structured data. A paramount example of such data is the brain, which operates as a network, from the micro-scale of neurons, to the macro-scale of regions. This organization deemed GNNs a natural tool of choice to model brain activity, and have consequently attracted a lot of attention in the neuroimaging community. Yet, the advantage of adopting these models over conventional methods has not yet been assessed in a systematic way to gauge if GNNs are capable of leveraging the underlying structure of the data to improve learning. In this work, we study and evaluate the performance of five popular GNN architectures in diagnosing major depression disorder and autism spectrum disorder in two multi-site clinical datasets, and sex classification on the UKBioBank, from functional brain scans under a general uniform framework. Our results show that GNNs fail to outperform kernel-based and structure-agnostic deep learning models, in which 1D CNNs outperform the other methods in all scenarios. We highlight that creating optimal graph structures for functional brain data is a major bottleneck hindering the performance of GNNs, where existing works use arbitrary measures to define the edges resulting in noisy graphs. We therefore propose to integrate graph diffusion into existing architectures and show that it can alleviate this problem and improve their performance. Our results call for increased moderation and rigorous validation when evaluating graph methods and advocate for more data-centeric approaches in developing GNNs for functional neuroimaging applications.

Siddharth Chaturvedi, Ahmed EL-Gazzar, Marcel van Gerven

Active particles are entities that sustain persistent out-of-equilibrium motion by consuming energy. Under certain conditions, they exhibit the tendency to self-organize through coordinated movements, such as swarming via aggregation. While performing non-cooperative foraging tasks, the emergence of such swarming behavior in foragers, exemplifying active particles, has been attributed to the partial observability of the environment, in which the presence of another forager can serve as a proxy signal to indicate the potential presence of a food source or a resource patch. In this paper, we validate this phenomenon by simulating multiple self-propelled foragers as they forage from multiple resource patches in a non-cooperative manner. These foragers operate in a continuous two-dimensional space with stochastic position updates and partial observability. We evolve a shared policy in the form of a continuous-time recurrent neural network that serves as a velocity controller for the foragers. To this end, we use an evolutionary strategy algorithm wherein the different samples of the policy-distribution are evaluated in the same rollout. Then we show that agents are able to learn to adaptively forage in the environment. Next, we show the emergence of swarming in the form of aggregation among the foragers when resource patches are absent. We observe that the strength of this swarming behavior appears to be inversely proportional to the amount of resource stored in the foragers, which supports the risk-sensitive foraging claims. Empirical analysis of the learned controller's hidden states in minimal test runs uncovers their sensitivity to the amount of resource stored in a forager. Clamping these hidden states to represent a lesser amount of resource hastens its learned aggregation behavior.

Ahmed El-Gazzar, Rajat Mani Thomas, Guido Van Wingen

Single-subject mapping of resting-state brain functional activity to non-imaging phenotypes is a major goal of neuroimaging. The large majority of learning approaches applied today rely either on static representations or on short-term temporal correlations. This is at odds with the nature of brain activity which is dynamic and exhibit both short- and long-range dependencies. Further, new sophisticated deep learning approaches have been developed and validated on single tasks/datasets. The application of these models for the study of a different targets typically require exhaustive hyperparameter search, model engineering and trial and error to obtain competitive results with simpler linear models. This in turn limit their adoption and hinder fair benchmarking in a rapidly developing area of research. To this end, we propose fMRI-S4; a versatile deep learning model for the classification of phenotypes and psychiatric disorders from the timecourses of resting-state functional magnetic resonance imaging scans. fMRI-S4 capture short- and long- range temporal dependencies in the signal using 1D convolutions and the recently introduced state-space models S4. The proposed architecture is lightweight, sample-efficient and robust across tasks/datasets. We validate fMRI-S4 on the tasks of diagnosing major depressive disorder (MDD), autism spectrum disorder (ASD) and sex classifcation on three multi-site rs-fMRI datasets. We show that fMRI-S4 can outperform existing methods on all three tasks and can be trained as a plug&play model without special hyperpararameter tuning for each setting

Ahmed El-Gazzar, Marcel van Gerven

In this work, we propose FlowTime, a generative model for probabilistic forecasting of multivariate timeseries data. Given historical measurements and optional future covariates, we formulate forecasting as sampling from a learned conditional distribution over future trajectories. Specifically, we decompose the joint distribution of future observations into a sequence of conditional densities, each modeled via a shared flow that transforms a simple base distribution into the next observation distribution, conditioned on observed covariates. To achieve this, we leverage the flow matching (FM) framework, enabling scalable and simulation-free learning of these transformations. By combining this factorization with the FM objective, FlowTime retains the benefits of autoregressive models -- including strong extrapolation performance, compact model size, and well-calibrated uncertainty estimates -- while also capturing complex multi-modal conditional distributions, as seen in modern transport-based generative models. We demonstrate the effectiveness of FlowTime on multiple dynamical systems and real-world forecasting tasks.

Ahmed El-Gazzar, Mirjam Quaak, Leonardo Cerliani et al.

Functional Magnetic Resonance Imaging (fMRI) captures the temporal dynamics of neural activity as a function of spatial location in the brain. Thus, fMRI scans are represented as 4-Dimensional (3-space + 1-time) tensors. And it is widely believed that the spatio-temporal patterns in fMRI manifests as behaviour and clinical symptoms. Because of the high dimensionality ($\sim$ 1 Million) of fMRI, and the added constraints of limited cardinality of data sets, extracting such patterns are challenging. A standard approach to overcome these hurdles is to reduce the dimensionality of the data by either summarizing activation over time or space at the expense of possible loss of useful information. Here, we introduce an end-to-end algorithm capable of extracting spatiotemporal features from the full 4-D data using 3-D CNNs and 3-D Convolutional LSTMs. We evaluate our proposed model on the publicly available ABIDE dataset to demonstrate the capability of our model to classify Autism Spectrum Disorder (ASD) from resting-state fMRI data. Our results show that the proposed model achieves state of the art results on single sites with F1-scores of 0.78 and 0.7 on NYU and UM sites, respectively.