Low-cost predictive models of dementia risk using machine learning and exposome predictors

• DOI: 10.1007/s12553-024-00937-5
• Publication Date: 2024-12-21T07:03:07Z
• Authors: Marina Camacho; Angélica Atehortúa; Tim Wilkinson; Polyxeni Gkontra; Karim Lekadir
• Keywords: Exposome

Abstract Purpose Diagnosing dementia, affecting over 55 million people globally, is challenging and costly, often leading to late-stage diagnoses. This study aims develop early, accurate, cost-effective dementia screening methods using exposome predictors machine learning. We investigate whether low-cost combined with learning models can reliably identify individuals at risk of dementia. Methods analyzed data from 500,000 UK Biobank participants, selecting 1523 diagnosed an equal number healthy controls, matched by age sex. A total 3046 participants were included: 2740 for internal validation 306 external validation. used 128 factors baseline visits, imputed missing data, assessed two predictive models: a classical logistic regression ensemble classifier (XGBoost). Feature importance was estimated within the models. Results The XGBoost model outperformed model, achieving mean AUC 0.88 in identified novel that might be as potential markers such facial aging, frequency use sun/ultraviolet light protection, length mobile phone use. Conclusions Machine utilizing showing superior performance. approach highlights low-cost, readily available Future studies should validate these findings diverse populations explore integration additional enhance prediction accuracy.