Deep Learning-Based System for Disease Screening and Pathologic Region Detection From Optical Coherence Tomography...

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Medical Imaging

Background:

• Retinal diseases require timely screening and diagnosis.
• Optical coherence tomography (OCT) is a key imaging modality for retinal analysis.
• Accurate detection of retinal lesions is crucial for effective treatment.

Purpose of the Study:

• To apply deep learning models for retinal disease screening.
• To develop models for lesion detection in OCT images.
• To evaluate the performance of AI in classifying and locating retinal pathologies.

Main Methods:

• Collected and labelled 37,138 OCT images from 775 patients.
• Developed and trained deep learning models (ResNet50, YOLOv3).
• Evaluated model performance on an independent holdout set using accuracy, sensitivity, specificity, and F1 score.

Main Results:

• Binary classification of OCT images achieved 98.5% accuracy.
• Multiclass disease classification exceeded 99% accuracy for specific conditions.
• Lesion location detection recall ranged from 87.0% to 98.2%.

Conclusions:

• Deep learning models demonstrate high performance in retinal disease screening and lesion detection.
• These AI tools can potentially assist ophthalmologists in diagnosis.
• The models show promise for clinical application in improving retinal disease management.