Radiomics-Based Assessment of OCT Angiography Images for Diabetic Retinopathy Diagnosis

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Diabetic retinopathy (DR) is a leading cause of vision loss in diabetic patients.
• Early detection and diagnosis of DR are crucial for effective management and prevention of blindness.
• Machine learning (ML) offers potential for automated analysis of retinal imaging.

Purpose of the Study:

• To assess the diagnostic accuracy of ML models utilizing radiomic features from Optical Coherence Tomography (OCT) and OCT Angiography (OCTA) for diabetes mellitus (DM), DR, and referable DR (R-DR).
• To compare the performance of different ML algorithms and imaging modalities in diagnosing DM, DR, and R-DR.

Main Methods:

• Radiomic features were extracted from fundus retinographies, OCT, and OCTA images.
• Several ML models, including logistic regression, linear discriminant analysis, and support vector classifiers (SVC), were employed.
• Diagnostic accuracy was evaluated using the area under the receiver operating characteristic curve (AUC).

Main Results:

• OCT demonstrated the highest AUC for DM diagnosis (0.82).
• OCTA showed the best performance for DR (0.77) and R-DR (0.87) detection, particularly with deep capillary plexus scans.
• Incorporating clinical variables significantly improved AUC for most models.

Conclusions:

• Radiomics from OCT and OCTA, analyzed with ML, can effectively identify patients with DM, DR, and R-DR.
• OCT is optimal for DM diagnosis, while OCTA excels in DR and R-DR diagnosis.
• This study suggests radiomics-based ML on OCT/OCTA images as a potential tool for DR screening in type 1 DM patients.