A systematic survey of machine learning and deep learning techniques for Alzheimer’s disease diagnosis

The Increasing number of demented people, particularly those living with Alzheimer’s disease (AD) is considered fundamentally challenging nowadays and puts great pressure on the health-care systems. This problem has led researchers to seek suitable methods for detecting the disease in its early stages. The widespread adoption of technology and the advances in artificial intelligence have contributed to finding solutions to such problems. The accurate and cost-effective early detection of AD is one of the primary goals of researchers, physicians, and other stakeholders. One promising approach is the application of machine learning to historical data from individuals with and without dementia. Therefore, identifying algorithms that show promising results in the early detection of the disease is essential. The aim of this study is to survey and critically analyze 62 research studies published between 2017 and 2026 that investigate the prediction and early detection of AD using data-driven techniques, particularly ML and deep learning techniques. Unlike previous reviews, the goal of this survey is to compare machine learning and deep learning techniques applied to different AD datasets in order to present the most favorable approaches for early detection across different stages of the disease. This has been done by highlighting strengths, weaknesses, limitations, and evaluation metrics to give researchers recommendations of the most favorable approaches for early detection. A systematic review following the PRISMA guidelines was done to ensure methodological rigor and reproducibility. The novelty of this survey lies in bringing together various studies to determine the most appropriate approaches used for early detection across AD stages. These findings can guide future research studies by identifying promising research trends and challenges.