A Foundational Model for Brain Imaging

Faculty: Prof. R. Venkatesh Babu (CDS) and Prof. Suresh Sundaram (Aerospace/CPS)

A comprehensive understanding of brain function and structure requires integrating data from diverse neuroimaging modalities, such as functional MRI (fMRI), diffusion tensor imaging (DTI), and structural MRI (sMRI). This project aims to develop a foundational brain model using advanced deep learning techniques that leverage multimodal neuroimaging data. By employing state-of-the-art representation learning and/or contrastive learning methods and training on large-scale, heterogeneous neuroimaging datasets, we seek to generate a robust and generalizable model that captures intricate relationships within the brain.

This foundational model will serve as a benchmark for various downstream tasks, including disease classification, progression modelling, and cognitive assessment, thereby advancing both computational neuroscience and clinical research. A key focus of this project is ensuring adaptability across different imaging protocols and patient populations through domain adaptation techniques, which is critical for translational impact in healthcare. Additionally, integrating explainable AI methods will enhance transparency by providing insights into the model’s decision-making processes.