Analysis of Functional Areas of Human Brain Based on Reconstructed Images of DMFG-generated Countermeasure Network

Abstract

The structure of human brain is complex, and fMRI data can be used to reveal the working mechanism of human brain. We construct a generative confrontation deep learning network based on DMFG-loss function. Using this network, we can not only reconstruct the simple scene images perceived and imagined by human brain with high precision, but also achieve good results for the restoration and reconstruction of complex natural images. In addition, we propose to set the detection threshold based on the constant false alarm algorithm. Further, we explore the distribution of brain sensitive areas, and make a deep analysis of the impact of different regions on image reconstruction. The contribution ratio of specific brain regions to the image reconstruction of human brain is gived. This will help to explore the unknown areas of human brain and reveal the mechanism of human brain operation. It has broad application prospects in brain computer interaction and human brain decoding.