OCT-based deep-learning models for the identification of retinal key signs

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Medical Imaging

Background:

• Accurate detection of retinal abnormalities from Optical Coherence Tomography (OCT) images is critical for clinical practice and patient diagnosis.
• Current diagnostic methods can be time-consuming and require specialized expertise.

Purpose of the Study:

• To develop and evaluate a novel deep learning (DL) system for recognizing specific retinal abnormality signs in OCT images.
• To assess the system's accuracy in distinguishing healthy retinas from pathological conditions and identifying various retinal diseases.

Main Methods:

• Utilized binary convolutional neural networks with the Visual Geometry Group 16 architecture.
• Developed nine models trained on 10,770 central fovea cross-section OCT images retrospectively collected from 2017-2022.
• Images were labeled by two retinal specialists to identify healthy retinas and eight distinct retinal abnormality signs.

Main Results:

• The DL system achieved high accuracy rates, ranging from 93% to 99%, in identifying healthy retinas and specific pathological signs.
• The models demonstrated significant potential in classifying retinal conditions with high precision.
• The approach reduced dataset creation time, highlighting the efficiency of DL in medical image analysis.

Conclusions:

• The developed deep learning system serves as a valuable diagnostic aid for ophthalmologists.
• This AI-powered tool can improve the accuracy and efficiency of ocular pathology diagnosis and clinical decision-making.
• The study underscores the potential of DL models to enhance patient care in ophthalmology.