Biosketch
Aswathi is a PhD student at the Institute for Artificial Intelligence in Medicine and the Institute of Diagnostic and Interventional Radiology at the Technical University of Munich, Germany. She obtained her Bachelor’s in Electronics and Communication Engineering from the University of Kerala, India. She pursued her Master’s in Control and Robotics: Signal and Image Processing at Ecole Centrale de Nantes, France. Her research interests lie in the domain of geometric deep-learning techniques for medical applications. Currently, she is working on deep learning for analysis of brain MRI.
relAI Research
Deep Learning for Longitudinal MR Image analysis of MS Patients
This research develops an automated deep learning pipeline for longitudinal MRI analysis in Multiple Sclerosis (MS). Monitoring disease progression currently relies on manual assessment, which is time-intensive and prone to variability. The proposed framework consists of three core stages: Segmentation, utilizing Random Convolutions to improve out-of-domain generalization across different scanners, and SegMaST, a Mamba-based spatio-temporal model, to capture global context between baseline and follow-up scans with linear complexity. Feature Extraction: Implementing VariViT, a Vision Transformer that handles variable-sized tumor crops without information loss via a “center-and-select” positional embedding. Relational Learning: Constructing Graph Neural Networks (GNNs) where lesions are nodes and edges encode spatial and temporal relationships. This integrated approach enables precise disease activity classification and clinical outcome prediction.