Brain Decoding using Beta scores and Graph Neural Networks

For this poject, we have access to the brain activities, and its parcellations. Then the main task is to decode the brain activites into different classes, based on the input stimuli. Normally as the data per voxcel in the brain is a FMRI signal, as a preprocessing stage, this data is represented by different techniqes. Here for this project, \(\beta\) scores as a linear model fitted to the FMRI data has role to represent this data. Therefor after casting the data to the a space called, \(\beta\)-space, the data has more meaningful representaiton, and also the connectome as the underlying graph connection of the brain regions is feeded to the Grpah neural netwrok for message passign between different links of this nodes. Eventually, after the trainig, the netwrok is trying to do a classification task based on the training data. Here is the over all procedure for this project: