Classification of Alzheimer’s Disease via Deep Residual Network

Alzheimer’s Disease (AD) is a typical neuro-degenerative illness. Numerous studies have discovered that the development of AD could produce obvious changes in brain structure and cerebrospinal fluid (CSF). With obtained brain features, our study proposed an approach of Deep Residual Network (DRN) with brain imaging and CSF features for AD classification. To accommodate the input of one-dimensional data, we opted to substitute the convolutional layer in our model with the dense layer. Specifically, the accuracy rates for the AD/NC, AD/MCI, and MCI/NC groups stood at 92.50%, 80.0%, and 67.50% respectively. Moreover, the three-class classification achieved an accuracy of 63.33%. For performance evaluation, our model was superior to those of other machine learning (ML) models. The proposed model demonstrated robust capability in revealing the potential biomarkers for AD classification and early predication.