Nonlocal operator learning for fMRI encoding and decoding tasks
Researchers have developed a new framework using neural integral operators to analyze functional MRI (fMRI) data, focusing on capturing nonlocal spatiotemporal context. This approach aims to improve both the encoding of brain activity from stimuli and the decoding of stimuli from brain activity. Experiments on open-source fMRI datasets demonstrated that larger temporal windows generally enhance performance and lead to more structured learned representations, suggesting that architectures designed to exploit distributed nonlocal brain dynamics are beneficial. AI
IMPACT Introduces a novel neural operator framework for analyzing complex spatiotemporal brain data, potentially advancing neuroscience research.