Early Detection and Prediction of Alzheimer's Disease Using Machine Learning

Abstract

Alzheimer disease (AD) is a neurodegenerative disorder that is progressive and is seen to impact more than 32 million individuals globally, and it is the most prevalent type of dementia (1). The timely intervention and patient outcomes can be achieved by detecting and predicting AD early. This paper thoroughly evaluates machine learning strategies aimed at early AD detection and evaluates different algorithms, such as convolutional neural networks (CNNs), support vectors machines (SVMs), and ensemble classifiers using neuroimaging data. The study analyses data of the Alzheimer Disease Neuroimaging Initiative (ADNI) and the Open Access Series of Imaging Studies (OASIS) and compares performance measures of the various modalities such as structural MRI, functional MRI and positron emission tomography. Findings indicate that CNN architectures can reach up to 99.57 percent accuracy when performing multi-class classification tasks, whereas the conventional machine learning models such as SVM can also perform with accuracy of 85-96 percent in such tasks (2,3). Multimodal neuroimaging data has a great benefit in level of diagnostic accuracy when compared with single modality techniques. The major issues are lack of dataset diversity, ability to interpret models and the ethical aspect of patient privacy. The findings of this study can be used to advance the development of AI-based diagnostic devices in detecting early AD, and it has a high possibility of clinical use and better patient care.