Figure 2:

Schematic representation of a sample bi-modal framework (sMRI-fMRI-NN) followed by the respective cGAN (sMRI-fMRI-cGAN). (a) The sMRI-fMRI-NN is composed of two different modules: feature reduction module, equal to the respective module for each channel in the full framework, and data fusion & classification module where the latent features are fused together and then classified. (b) After training, the latent features of both modalities obtained from the feature reduction module were extracted for each subject and given as input for the training of the associated cGAN, whose loss is described in (c) and (d).