Modular machine learning for Alzheimer's disease classification from retinal vasculature

Area of Science:

• Ophthalmology
• Neurology
• Artificial Intelligence

Background:

• Alzheimer's disease (AD) is the primary cause of dementia, necessitating early detection for effective treatment.
• Current diagnostic imaging for AD is expensive and not widely accessible, limiting routine screening.
• The retina's vasculature presents a potential, accessible alternative for AD screening.

Purpose of the Study:

• To evaluate the feasibility of using retinal vasculature analysis for early Alzheimer's disease screening.
• To develop and validate a machine learning pipeline for detecting AD from retinal images.

Main Methods:

• A highly modular machine learning pipeline was designed for image analysis.
• Data from the UK Biobank was utilized for training and testing the pipeline.
• Saliency analysis was incorporated to enhance model interpretability.

Main Results:

• The machine learning pipeline achieved an average classification accuracy of 82.44% in identifying Alzheimer's disease.
• Saliency analysis revealed that small retinal vessels are significant indicators for AD diagnosis.
• Findings align with existing research on the link between retinal changes and neurodegenerative diseases.

Conclusions:

• Retinal vasculature analysis using machine learning is a promising, accurate, and potentially cost-effective method for Alzheimer's disease screening.
• The developed pipeline offers a novel approach to complement existing diagnostic tools.
• Further research can refine this technique for clinical application in dementia screening.