Learning Glioblastoma Tumor Heterogeneity Using Brain Inspired Topological Neural Networks

ABSTRACT

Glioblastoma (GBM) is characterized by high intra-tumoral heterogeneity, which poses a significant challenge for diagnosis and treatment planning. We propose a brain-inspired Topological Neural Network (TNN) that leverages the structural connectivity patterns of the human brain to model the growth and infiltration of GBM cells. By integrating multi-modal MRI data with a topological prior based on white matter architecture, our TNN can more accurately predict tumor invasion pathways compared to standard convolutional networks. This work demonstrates the potential of incorporating biological structural priors into deep learning models for personalized oncology, offering a more nuanced understanding of the relationship between tumor dynamics and brain connectivity.