Alzheimer's Disease Classification using the Fusion of Improved 3D-VGG-16 and Machine Learning Classifiers

Alzheimer's disease is a significant cause of mortality in older adults. It is characterised by the accumulation of amyloid-β (Aβ) and hyperphosphorylated tau in the brain. Structural magnetic resonance imaging (MRI) can detect these progressive brain changes in the early stages. This work presented an improved 3D-VGG-16 deep neural network model fused with three machine learning classifiers: GPC, SVM, and RF. The improved 3D-VGG-16 incorporated the dropout technique, removed fully connected layers, and replaced them with new ones. To train and test T1-weighted 3D MRI images were utilised, a total of 445 images where 239 (AD-69, CN-170) from ADNI and 206 (AD-136, HC-70) from MIRIAD. The models achieved accuracies of 98.75%, 97.5%, and 93.75% (ADNI), and 96.25%, 93.75%, and 91.25% (MIRIAD) for 3D-VGG-16+GPC, 3D-VGG-16+SVM, and 3D-VGG-16+RF. The proposed model with dropout, 3D data adaption, and simultaneous data evaluation demonstrated accurate AD classification, especially with the GPC classifier.