Toward an Interpretable Alzheimer's Disease Diagnostic Model with Regional Abnormality Representation via Deep Learning

• We proposed a network that provides deep learning-based diagnosis and interpretability for Alzheimer's disease.
• We used a 1.5-T T1-weighted MRI dataset with images from 801 subjects from the ADNI (Alzheimer's Disease Neuroimaging Initiative).
• We constructed a neural network to represent the abnormality of each subspace for each region.
• We pretrained our neural network with greedy layer-wise pretraining by constructing stacked denoising auto-encoders (SDAEs) (Vincent et al., 2010).
• The represented abnormality is finally classified by the SVM.

• We showed an accuracy of 89.22% for mild cognitive impairment (MCI) vs. cognitive normal (CN) and an accuracy of 88.52% for progressive MCI (pMCI) vs. stable MCI (sMCI) classification tasks.

• The closer to red, the greater the degree of abnormality (from a model perspective).

Implemented by MATLAB(R2017a)

DeepLearnToolbox (https://github.com/rasmusbergpalm/DeepLearnToolbox)
LIBSVM v3.21(https://www.csie.ntu.edu.tw/~cjlin/libsvm/)

Used converted MRI (from .nii or .img to .mat)
Used template Kabani's atlas (Kabani, N. J., 1998. 3D anatomical atlas of the human brain. NeuroImage 7, P–0717.)