Figure 4:

The overall workflow of our model. N is number of nodes, M is number of convolutional feature maps, H₁ is number of neurons in the first FC layer and H₂ is the number of neurons in the second FC layer. Our model uses specialized E2E and E2N filters as well as employs multi-task learning on a variety of available t-fMRI paradigms. Each grey module represents a separate 3-class segmentation task. The variables L, M₁, M₂ and M₃ represent the language, finger, tongue, and foot networks respectively, as shown by the segmentation maps where red, blue, and white refer to the eloquent, neither, and tumor classes respectively.