Biomarkers.

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Dementia Research

Background:

• AI models for dementia prediction often use curated research datasets, limiting real-world generalizability.
• A gap exists in evaluating the performance of AI dementia prediction models on diverse, real-world clinical data.

Purpose of the Study:

• To assess the generalizability of AI models trained on large research datasets for dementia prediction using real-world UK National Health Service (NHS) data.
• To evaluate the effectiveness of AI-driven magnetic resonance imaging (MRI) analysis in predicting the time to dementia diagnosis.

Main Methods:

• 3D T1-weighted MRI scans and electronic health records from 1140 individuals at SLaM NHS Trust were analyzed.
• AI models (3D ResNet) trained on National Alzheimer's Coordinating Center (NACC) and Alzheimer's Disease Neuroimaging Initiative (ADNI) datasets were applied to the SLaM cohort.
• Probabilities of Alzheimer's disease dementia (p_AD) were integrated with clinical covariates (age, sex, MMSE) using Cox proportional hazards regression to estimate time-to-diagnosis.

Main Results:

• AI models achieved classification accuracies of 65.3%-68.6% in the SLaM cohort.
• The p_AD metric significantly predicted time-to-diagnosis, with hazard ratios ranging from 3.58 to 4.67 (p<0.0017).
• A 0.1 increase in p_AD correlated with a 13.6%-16.7% increased risk of dementia diagnosis within 8 years.

Conclusions:

• AI models trained on large research datasets demonstrate effectiveness in predicting dementia progression in real-world NHS patient populations.
• The study highlights the clinical utility and generalizability of AI models for dementia risk prediction in diverse clinical settings.
• AI-driven analysis of neuroimaging data shows promise for improving early detection and management of dementia.