A Novel Deep Dense Block-Based Model for Detecting Alzheimer’s Disease

Abstract

Alzheimer’s disease (AD), the most common form of dementia and neurological disorder, affects a significant number of elderly people worldwide. The main objective of this study was to develop an effective method for quickly diagnosing healthy individuals (CN) before they progress to mild cognitive impairment (MCI). Moreover, this study presents a unique approach to decomposing AD into stages using machine-learning architectures with the help of tensor-based morphometric image analysis. The proposed model, which uses a neural network built on the Xception architecture, was thoroughly assessed by comparing it with the most recent convolutional neural network (CNN) models described in the literature. The proposed method outperformed the other models in terms of performance, achieving an impressive average classification accuracy of 95.81% using the dataset. It also had very high sensitivity, specificity, accuracy, and F1 scores, with average scores of 95.41%, 97.92%, 95.01%, and 95.21%, respectively. In addition, it showed a superior classification ability compared to alternative methods, especially for MCI estimation, as evidenced by a mean area under the ROC curve (AUC) of 0.97. Our study demonstrated the effectiveness of deep-learning-based morphometric analysis using brain images for early AD diagnosis.