Development and validation of a deep learning algorithm based on fundus photographs for estimating the CAIDE dementia...

Area of Science:

• Ophthalmology
• Neurology
• Artificial Intelligence

Background:

• The Cardiovascular Risk Factors, Aging, and Incidence of Dementia (CAIDE) dementia risk score is a validated tool for dementia risk stratification.
• Current limitations include the need for multidimensional data and fasting blood draws, highlighting the need for accessible screening methods.
• An effective, non-invasive tool is crucial for screening high dementia risk in large populations.

Purpose of the Study:

• To develop and validate a deep learning algorithm using fundus photographs to estimate the CAIDE dementia risk score.
• To assess the algorithm's performance in identifying individuals with high dementia risk (CAIDE score ≥ 10 points).

Main Methods:

• A deep learning algorithm was developed using fundus photographs.
• Internal validation involved a dataset of 271,864 participants from China.
• External validation utilized an independent dataset of 20,690 participants from Beijing.

Main Results:

• The algorithm achieved high performance with an AUC of 0.944 (95% CI: 0.939-0.950) in internal validation and 0.926 (95% CI: 0.913-0.939) in external validation.
• The estimated CAIDE dementia risk score correlated significantly with comprehensive cognitive function and specific cognitive domains.

Conclusions:

• The fundus photograph-based algorithm effectively identifies individuals at high risk for dementia in population settings.
• This approach offers a non-invasive and expedient method for dementia risk stratification.
• The algorithm has potential applications in dementia clinical trials for efficient participant selection.