Alzheimer’s disease Prediction by Hybrid CNN and SVM Classifier with Metaheuristic Approach

Abstract

The most common type of dementia is Alzheimer’s disease (AD). It is critical to identify the AD at the stage of Mild Cognitive Impairment (MCI) in early. If it is possible to early identification, then it has more chance to cure the disease. This paper implements a novel predictive approach for early detection of AD utilizing Magnetic Resonance Imaging (MRI) images. The developed model involves Feature Extraction, Optimal Feature selection, Classification. At first, the Gray Level Co-Occurrence Matrix (GLCM), Haralick features, and geometric Haralick feature techniques are used to extract the geometric correlation and variances features. This work carries out optimal feature selection using the Combined Grey Wolf -Dragon Updating (CG-DU) hybrid model. This optimization model has been used in Convolutional Neural Network (CNN) for the optimized weights and activation function. Optimally chosen features by CNN are subjected to the Classifier Support Vector Machine (SVM) for AD classification. The final output is obtained from both CG-DU+CNN and SVM outcomes. In the end, the performance of the implemented approach is computed to the existing approaches based on various metrics.