Alzheimer's disease MRI image recognition method based on deep learning

Globally, the incidence of Alzheimer's disease is increasing. Although there are many diagnostic methods, traditional methods often rely on professional knowledge and consume much time. Therefore, there is an urgent need to introduce automated technology to improve diagnostic efficiency and accuracy. This study used three deep learning models, namely convolutional neural network (CNN), ResNet, and VGGNet. It optimized them according to data characteristics for Alzheimer's MRI image recognition, providing a reference for diagnosing Alzheimer's disease. This study used the dataset from Dataset Alzheimer, based on the data preprocessing of Alzheimer's Disease Neuroimaging Initiative (ADNI). The model was trained after the dataset was processed by data augmentation, pixel value scaling and MOTE oversampling. The experimental results indicated that the VGGNet model attained the highest overall accuracy at 76%, and the CNN and ResNet models reached accuracies of 75% and 74%, respectively. Nonetheless, the model's proficiency in reliably identifying categories with limited samples necessitates enhancement. Furthermore, further analysis is necessary to assess the model's generalisation potential and its relevance in clinical environments. Future research may explore more advanced model architectures, incorporating attention processes and generative adversarial networks (GANs), to enhance the model's recognition capabilities and its suitability for therapeutic applications.