Graph Neural Networks applied to classification of autism and symptoms in fMRI data

• Grant number: 24/03245-9
• Support Opportunities: Scholarships abroad - Research Internship - Master's degree
• Start date: July 29, 2024
• End date: January 28, 2025
• Field of knowledge: Physical Sciences and Mathematics - Computer Science
• Principal Investigator: André Carlos Ponce de Leon Ferreira de Carvalho
• Grantee: Ana Letícia Garcez Vicente
• Supervisor: Leila Wehbe
• Host Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil
• Institution abroad: Carnegie Mellon University (CMU), United States
• Associated to the scholarship: 23/06737-7 - Graph Neural Networks applied to classification of autism symptoms in fMRI data, BP.MS

Abstract

Science has been actively seeking tools to aid in the early diagnosis of diseases, including psychiatric disorders. Functional magnetic resonance imaging (fMRI) has emerged as a valuable tool for studying and diagnosing these disorders, allowing for the observation of brain activation through the BOLD signal. However, fMRI data are inherently complex, and leveraging machine learning techniques, particularly deep learning, holds promise in facilitating these studies. Among these techniques, Graph Neural Networks (GNNs) stand out for their ability to handle complex data structures, such as brain images, by treating the brain as a graph, with regions as nodes and edges representing connections between regions. The objective of this master's project is to investigate various GNN models for classifying fMRIs. Initially, we preprocessed a dataset containing data from both healthy control individuals and those with autism spectrum disorder, transforming the fMRI data into graph representations. The primary aim of this project is to explore and develop different GNN architectures for classifying neurotypical individuals and those with autism. Furthermore, we aim to utilize the GNN model to identify the severity of autism-related symptoms using categorizations from psychometric questionnaires, such as the Social Responsiveness Scale questionnaire, as class labels. This research seeks to contribute to the advancement of diagnostic methodologies for psychiatric disorders, particularly in the realm of autism spectrum disorder, through the integration of advanced machine learning techniques with neuroimaging data.