Fatemeh Ghayyem (MIND team at Inria Paris-Saclay)

Fatemeh Ghayyem

Title: Exploring brain function and structure: From sparse coding to multimodal metadata analysis

Abstract:In the rapidly advancing field of neuroimaging, the dual challenge of extracting clear insights from complex brain data and reliable knowledge from individual studies remains paramount. This presentation introduces innovative computational techniques that enhance our understanding of brain function and structure. First, I will explore proximal methods for sparse signal recovery, demonstrating the efficacy of these techniques in reconstructing signals from noisy and underdetermined measurements. Our approach incorporates nonconvex sparsity promotion, error constraints, and accelerated schemes to outperform existing algorithms. I will then delve into the extraction of interpretable patterns and discriminative features from brain functional network connectivity, employing a novel combination of independent component analysis and dictionary learning. This methodology effectively discriminates between healthy controls and schizophrenia patients, offering new insights into mental health disorders. Finally, I propose a new technique for brain metadata analysis, aiming to overcome the limitations of conventional 'bag of words' models and enhance the statistical power, reproducibility, and terminology consistency of neuroscientific studies. Our integrated approach provides a comprehensive framework for analyzing the intricate relationships between brain networks and behavior, marking a significant step forward in brain imaging analysis.

Biography: I am a postdoctoral researcher at the MIND team in Inria-Saclay starting March 2023, working under the supervision of Bertrand Thirion and Demian Wassermann. My current research focuses on the intersection of machine learning, signal processing, and statistics to analyze brain multimodal metadata using Large Language Models (LLMs) in the fields of neuroscience and neuroimaging. Before that, I worked as a postdoctoral researcher for a year at the MLSP-lab at the University of Maryland, Baltimore County (UMBC) under the supervision of Prof. Tulay Adali. During this period, my research was focused on identifying new patterns and discriminant features of brain functional network connectivity, subgroup identification, brain graph neural network, and reproducibility and replicability assessment. I obtained my Ph.D. in biomedical engineering from the GIPSA-lab, University Grenoble Alpes, under the supervision of Prof. Christian Jutten and Dr. Bertrand Rivet, where I worked on optimal sensor placement for source extraction. Before pursuing my Ph.D., I worked as a research assistant at the DSP-lab at the Sharif University of Technology under the supervision of Prof. Massoud Babaie-Zadeh, where I developed accelerated dictionary learning and sparse representation algorithms. I also obtained an M.Sc. degree in biomedical engineering from the Sharif University of Technology, where I explored the reconstruction of magnetic resonance images using compressed sensing techniques.

Website: https://ghayem.github.io 

In Room M7.101 of Monod campus, ENSL.