DRRisk: A Web-based tool to Assess the Risk of Diabetic Retinopathy through Machine Learning on Electronic Health...

Area of Science:

• Ophthalmology
• Medical Informatics
• Computer Science

Background:

• Diabetic retinopathy (DR) is a leading cause of vision loss in individuals with diabetes.
• Early detection and intervention are crucial for preventing blindness.
• Existing risk assessment methods may not be readily accessible, especially in underserved populations.

Purpose of the Study:

• To develop and evaluate DRRisk, a machine learning-based web tool for assessing diabetic retinopathy risk.
• To leverage electronic health record (EHR) data for automated risk stratification.
• To support healthcare providers in prioritizing patients for DR screening and management.

Main Methods:

• Development of a web-based application (DRRisk) using Python (Flask framework) and standard web technologies (HTML, CSS, JavaScript).
• Utilization of machine learning algorithms trained on EHR data to predict DR risk.
• Consultation with clinical experts for tool design and validation.

Main Results:

• DRRisk successfully assesses current diabetic retinopathy risk.
• The tool categorizes patient risk into low, moderate, or high levels based on machine learning model output.
• The system provides a quantifiable risk percentage for DR.

Conclusions:

• The DRRisk tool effectively calculates diabetic retinopathy risk using EHR data.
• It can aid in identifying diabetic patients who require screening for undiagnosed DR.
• The tool has the potential to improve DR management and prevent vision loss, particularly in underserved communities.