A Deep Learning System for Fully Automated Retinal Vessel Measurement in High Throughput Image Analysis

Area of Science:

• Ophthalmology
• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• The retinal microvasculature serves as a critical indicator for cardiovascular diseases.
• Current high-throughput tools for detailed retinal vessel analysis are insufficient.
• Automated analysis of retinal vasculature is needed for early disease detection and monitoring.

Purpose of the Study:

• To develop and validate an artificial intelligence system, the Retina-based Microvascular Health Assessment System (RMHAS).
• To achieve fully automated vessel segmentation and quantification of the retinal microvasculature.
• To establish a robust tool for predicting and monitoring cardiovascular diseases through retinal imaging.

Main Methods:

• Development of the Retina-based Microvascular Health Assessment System (RMHAS) utilizing deep learning.
• Validation of the system across diverse datasets including AV-WIDE, AVRDB, HRF, IOSTAR, LES-AV, and RITE.
• Performance evaluation through segmentation accuracy (AUC), agreement, and repeatability analyses.

Main Results:

• RMHAS demonstrated high segmentation accuracy for both arteries and veins across multiple datasets (AUCs ranging from 0.88 to 0.97).
• The system showed robust performance, supported by agreement and repeatability analyses.
• Quantitative vessel analysis was completed in under 2 seconds per case.

Conclusions:

• The developed RMHAS provides accurate and efficient automated segmentation and quantification of retinal microvasculature.
• This AI system holds significant potential for the early prediction and monitoring of cardiovascular diseases.
• RMHAS offers a high-throughput solution for in-depth retinal vessel analysis, addressing a current gap in clinical tools.