A Review of Supervised Learning Techniques for Dementia Detection from Cognitive Data

Abstract

Accurate prediction of dementia is important for early access to healthcare resources. Conventional diagnostic tools often depend on resource-intensive clinical assessments, including neuropathological evaluations and specialised medical personnel. In response, artificial intelligence (AI), particularly supervised machine learning (ML) techniques, has provided promising diagnostic models derived from historical data related to dementia stages. This study presents a comprehensive review of recent literature focussed on the application of supervised learning methods in dementia prediction, with an emphasis on cognitive assessment data. We critically analyse and compare common supervised learning algorithms applied to the Alzheimer’s Disease Neuroimaging Initiative (ADNI) and other publicly available repositories. The review outlines each method’s operational mechanisms, advantages, and limitations. Additionally, we identify and discuss crucial factors, such as cognitive test batteries, diagnostic frameworks, and performance metrics, which are impactful on the future of dementia diagnosis research.