Multi-method Analysis for Early Diagnosis of Alzheimer's Disease on Magnetic Resonance Imaging (MRI) Using Deep Learning and Hybrid Methods

Alzheimer's disease (AD) is a progressive and irreversible brain disorder that leads to cognitive impairment and an inability to perform daily tasks. It is the primary cause of dementia, accounting for the majority of cases. Abnormal protein deposits in the brain are responsible for AD, resulting in the death of brain cells. While there is currently no cure, Computer Assisted Diagnostics (CAD) with Magnetic Resonance Imaging (MRI) has facilitated disease detection. In this paper, we aimed to improve Alzheimer's disease identification by employing ensemble-modified transfer learning techniques and hybrid approaches that combine deep learning with machine learning. We utilized four pre-trained models (AlexNet, ResNet-50, VGG16, and InceptionV3) and applied batch normalization and regularization techniques. Additionally, various machine learning algorithms (Support Vector Machine, Logistic Regression, Random Forest, and K Nearest Neighbors) were used, achieving high performance in diagnosing dementia. Our results indicate that the hybrid methods combining deep learning and machine learning outperformed the modified deep learning models. Particularly, the AlexNet-M + SVM hybrid model demonstrated an accuracy, precision, sensitivity, specificity and F-measure of 91.41%, 91.41%, 91.41%, 100.00% and 91.39%, respectively. In summary, our study highlights the importance of hybrid approaches for AD diagnosis, showcasing the superior performance of the hybrid model AlexNet-M + SVM. The findings contribute to the advancement of Computer Assisted Diagnostics for the early detection and management of this debilitating neurological condition.