Learning Macroscopic Brain Connectomes via Group-Sparse Factorization

摘要: Mapping structural brain connectomes for living human brains typically requires expert analysis and rule-based models on diffusion-weighted magnetic resonance imaging. A data-driven approach, however, could overcome limitations in such approaches improve precision mappings individuals. In this work, we explore a framework that facilitates applying learning algorithms to automatically extract connectomes. Using tensor encoding, design an objective with group-regularizer prefers biologically plausible fascicle structure. We show the is convex has unique solutions, ensuring identifiable individual. develop efficient optimization strategy extremely high-dimensional sparse problem, by reducing number of parameters using greedy algorithm designed specifically problem. significantly improves standard algorithm, called Orthogonal Matching Pursuit. conclude solutions found our method, showing can accurately reconstruct diffusion information while maintaining contiguous fascicles smooth direction changes.