A New Design of 3D-CNN model for Alzheimer detection using Enhanced Hybrid-Genetic Algorithm based SVM Classifier

Abstract

Advances in healthcare technologies have gained considerable attention for their potential to promote longer and healthier lives. Among neurodegenerative conditions, Alzheimer’s disease (AD) remains the leading cause of dementia, and the financial impact of caring for individuals with AD is expected to escalate substantially in the coming years. As a result, the development of dependable computer-aided diagnostic tools for the early and accurate detection of AD has become a critical priority. Deep learning techniques, particularly those designed for medical image analysis, provide strong advantages over traditional machine-learning methods. Recent research has demonstrated that convolutional neural networks (CNNs) applied to brain MRI scans can deliver highly effective diagnostic performance. Building on these advancements, this study proposes a fully integrated 3D-CNN architecture for classifying Alzheimer’s disease. Using data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), the model achieved classification accuracies of 93.6%, 94.8%, and 95.8% for different AD-related prediction tasks. For binary classification between AD and cognitively normal (CN) subjects, the ADNI dataset was used, and for multi-class experiments the model achieved an accuracy of 97.5%. To further demonstrate its effectiveness, the 3D-CNN model was compared against a CNN based SVM classifier, showing superior performance.