A Pilot Study of Diabetes Mellitus Classification from rs-fMRI Data Using Convolutional Neural Networks

Figure 1

The abstraction of the proposed algorithmic pipeline and visualization process. (a) The proposed Inception-v4-Residual network with seven parameter layers (about 0.09 million parameters). (b) Grad-CAM overview: Given an image and a class of interest as input, we forward-propagate the ALFF image through the CNN part of the model and then through the task-specific computations to obtain a raw score for the category. Then, the gradients are set to 0 for all classes except the desired class, which is set to 1. Finally, this signal is back-propagated to the parametric rectified convolutional feature maps of interest, which we combine to compute the coarse Grad-CAM localization (blue heatmap), which represents where the model has to look to make the particular decision.

(a)

(b)