Enhanced Alzheimer’s Disease Detection Using Deep Neural Networks with Spatial Feature Enhancement

Alzheimer’s Disease (AD) presents a pressing global health concern, significantly affecting the elderly. As the predominant form of dementia, AD progressively damages brain cells, resulting in significant neurological impairment. Developing an automated system for detecting AD is critical, prompting researchers to explore deep learning (DL) methods, which have demonstrated efficacy in diverse fields, including medical image analysis. Researchers are still facing challenges to achieving high-performance accuracy in AD work due to the intricacies involved in training hierarchical convolutional neural networks (CNNs). Initially, we applied an initial convolutional block comprising a convolutional layer and a spatial feature enhancement module, along with MaxPooling and Dropout layers. Subsequently, we implemented three intermediate convolutional blocks to capture complex relational features, followed by a final convolutional block. The final stage involves a DL-based classification block to generate probabilistic information, and we also experimented with machine learning (ML) algorithms such as Support Vector Machines (SVM) and Random Forests (RF). Our extensive experiments on the Kaggle binary class AD dataset demonstrated the efficacy of our model, achieving an impressive accuracy of 99.02%. The strengths of our proposed model lie in its ability to effectively capture and enhance spatial features through residual learning and multiple convolutional stages, leading to superior performance in AD detection compared to existing approaches.