Investigation of Alzheimer’s Disease Biomarkers using Deep Learning on MR Images: A Systematic Review

Abstract

Alzheimer’s Disease (AD) is a progressive neurodegenerative disorder and the leading cause of dementia, particularly among the elderly. Although no cure currently exists, early diagnosis and intervention can slow disease progression and improve quality of life. Magnetic Resonance Imaging (MRI) plays a crucial role in the early detection of AD by capturing characteristic brain changes such as ventricular enlargement, entorhinal cortex degeneration, hippocampal atrophy, and cortical volume loss. In recent years, Deep Learning (DL) has shown exceptional potential in medical image analysis, including disease detection and classification. This systematic review investigates recent studies that combine DL techniques with brain MRI for AD diagnosis and classification. A total of 94 peer-reviewed articles published between 2020 and 2024 were initially identified from five major scientific databases, of which 46 met the inclusion criteria. Several studies also utilized publicly available datasets, particularly from the Alzheimer’s Disease Neuroimaging Initiative (ADNI). The reviewed literature reports encouraging results, with some DL models achieving classification accuracies as high as 98.73%. However, issues such as inconsistent data splitting protocols—especially the lack of patient-wise separation in some studies—raise concerns about potential data leakage, which could lead to overestimated model performance.