Neural Networks models to predict individual stroke recovery from Multimodal MRI Data: from small to large datasets
Supervisors
Cassandra Sampaio Baptista, Institute of Neuroscience and Psychology, University of Glasgow
Tanaya Guha, School of Computing Science, University of Glasgow
Summary
Understanding and predicting individual stroke recovery and impairment is a key goal for clinicians and stroke survivors. Multimodal Magnetic Resonance Imaging (MRI) allows us to image functional and structural human brain properties in vivo and to relate them to functional impairment. Machine Learning (ML) methods have seen major breakthroughs in the last decade in the domain of natural image understanding, making its way to medical image analysis. Most ML-based MRI analysis use large data sets with hundreds or thousands of individuals, making it less than ideal for clinical populations that, with some exceptions, rely on small sample sizes. Recent advances in ML focus on learning from smaller datasets through approaches like self-supervised/unsupervised learning and data augmentation.
In this PhD project the student will leverage multimodal MRI using ML methods to perform data augmentation in small sample studies and to discover participant-specific data features to develop models to predict impairment in stroke survivors.