Ensemble-of-classifiers-based approach for early Alzheimer’s Disease detection

Alzheimer’s disease (AD) is a deadly neurological condition. Deep learning approaches (DL) techniques have just been utilized to track the evolution of Alzheimer’s disease. These studies only employed baseline neuro imaging data. Because of the high cost of neuro imaging data, it is constantly restricted or unavailable. As a result, this research developed a novel, four-phase early Alzheimer’s disease detection approach: “(a) pre-processing, (b) feature extraction, (c) feature selection, and (d) classification”. Data cleaning and normalization is used in pre-processing. Consequently, features like “Weighted Geometric Mean Principle Component Analysis (WGM-PCA), Statistical Features, higher-order statistical features, and Weighted modified correlation-based features” are retrieved from the pre-processed data. Employing the Improved Attribute Ranker (IAR), the most relevant characteristics are chosen. Furthermore, the disease classification phase is represented by a deep learning model based on an ensemble of classifiers, containing optimized “Bi-GRU, Multi-Layer Perceptron (MLP), and Quantum Neural Network (QDNN)”, respectively. The ultimate decision is obtained via optimal Bi-GRU, which is trained using MLP and QDNN outcomes. Both the MLP and the QDNN would be trained using the chosen IAR-based features. Interestingly, to improve the network’s detection accuracy, the weight of the QDNN model is adjusted using the recently proposed Enhanced Math Optimizer Accelerated Arithmetic Optimization (EMOAOA) technique. Particularly, the proposed EMOAOA+EC achieved detecting accuracies of 95% at the 60th LR, 95.5% at the 70th LR, 98% at the 80th LR, and 98.7% at the 90th LR. The development of the optimized ensemble classifier is responsible for this improvement.