Supervised Machine Learning Based Multi-Task Artificial Intelligence Classification of Retinopathies.

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Artificial intelligence (AI) shows potential as an accessible screening tool for ocular diseases, especially in underserved regions lacking ophthalmologists.
• Quantitative optical coherence tomography angiography (OCTA) effectively detects vascular changes crucial for diagnosing retinovascular diseases.
• Current AI applications for classifying multiple ocular diseases are not well-established.

Purpose of the Study:

• To develop and validate a supervised machine learning (ML) model for multi-task classification of ocular diseases using OCTA.
• To differentiate normal from diseased eyes, distinguish between different ocular diseases, and stage disease severity.
• To identify optimal quantitative OCTA features for accurate multi-task classification.

Main Methods:

• Supervised ML-based multi-task classification of OCTA images.
• Automatic extraction of quantitative OCTA features: blood vessel tortuosity (BVT), vascular caliber (BVC), vessel perimeter index (VPI), blood vessel density (BVD), foveal avascular zone (FAZ) area (FAZ-A), and FAZ contour irregularity (FAZ-CI).
• Stepwise backward elimination to select sensitive OCTA features and optimal combinations for classification, validated using diabetic retinopathy (DR) and sickle cell retinopathy (SCR).

Main Results:

• Demonstrated a supervised ML approach for multi-task OCTA classification.
• Identified key OCTA features for differentiating and staging ocular diseases.
• Successfully validated the classifier using diabetic retinopathy and sickle cell retinopathy as proof-of-concept.

Conclusions:

• The developed AI classification methodology is effective for multi-task OCTA analysis.
• This approach can be extended to various ocular diseases, paving the way for mass-screening platforms.
• The technology holds significant promise for clinical deployment and telemedicine, enhancing eye care accessibility.