Code for the Masterthesis
"A Deep Vision Approach for Feature Extraction of 4D task-based fMRI Sequences"
This thesis concentrates on extracting features of brain activity from 4D task-based functional magnetic resonance images (tfMRI) by using convolutional neural networks and gradient-weighted class-activation- mapping for feature visualization. In contrast to the common statistical evaluation method for tfMRI sequences, the feature extraction method presented here uses the whole topology and temporal structure of the tfMRI sequences and is therefore capable to also take into account inter-voxel and temporal correlations.