Efficient Alzheimer’s Classification Leveraging Deep Learning Models on MRI Data with Layer Optimization and Early Stopping

Alzheimer’s disease is the predominant aetiology of dementia in the elderly population. Dementia is a collection of symptoms that impair memory, cognition, and social skills. It often entails cognitive impairment and heightened anxiety. A dataset including brain magnetic resonance imaging (MRI) is used to investigate this illness. This research categorizes the condition into two classifications: Demented and NonDemented. Three balanced data subsets were used for classification, including 1000, 2000, and 3000 samples from each class. Convolutional Neural Networks, namely MobileNet, VGG16, InceptionV3, and ResNet50, experimenting with various layer combinations throughout epochs of 25, 50, 75, and 100, incorporating early stopping; MobileNet exhibited superior performance were used. The suggested approach successfully classifies MRI scans of Alzheimer’s disease via trials done on the dataset. At epoch 50, MobileNet exhibits a test accuracy of 95.25%, a validation accuracy of 91.56%, a test loss of 0.184, and a validation loss of 0.218, with early halting occurring at 11 and an execution duration of 487.49 seconds.