DISEASE CLASSIFICATION OF MACULAR OPTICAL COHERENCE TOMOGRAPHY SCANS USING DEEP LEARNING SOFTWARE: Validation on...

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence in Healthcare

Background:

• Optical coherence tomography (OCT) is crucial for diagnosing retinal diseases.
• Clinical decision support software aims to enhance diagnostic accuracy and efficiency.
• Evaluating software performance across diverse populations is essential for widespread adoption.

Purpose of the Study:

• To assess the performance of Pegasus optical coherence tomography (OCT) software.
• To evaluate its ability to identify retinal disease features in macula OCT scans.
• To test its efficacy across heterogeneous populations, varying demographics, devices, sites, and operators.

Main Methods:

• Processed 5,588 normal and anomalous macular OCT volumes (162,721 B-scans) from independent international centers.
• Utilized Pegasus-OCT software for feature identification.
• Evaluated software results against ground truth data.

Main Results:

• Achieved areas under the receiver operating characteristic curve (AUC) of at least 98% for general macular anomaly detection across all datasets.
• For high-quality scans, AUCs were at least 99% for age-related macular degeneration and 98% for diabetic macular edema.
• Demonstrated high performance in detecting specific retinal diseases.

Conclusions:

• Pegasus-OCT demonstrates robust performance in detecting age-related macular degeneration, diabetic macular edema, and general anomalies.
• Its ability to perform across diverse populations and acquisition sites is key for adoption.
• The software shows significant potential to support eye care services and manage the increasing demand for retinal disease diagnosis.