Automatic Detection of AMD and DME Retinal Pathologies Using Deep Learning

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Diabetic macular edema (DME) and age-related macular degeneration (AMD) are leading causes of irreversible vision loss.
• Late diagnosis of DME and AMD significantly increases the risk of permanent vision impairment.
• Early detection and treatment are vital for preserving sight and improving patient quality of life.

Purpose of the Study:

• To develop and validate automatic deep learning-based methods for early detection of DME and AMD.
• To leverage optical coherence tomography (OCT) imaging for enhanced diagnostic capabilities.
• To improve patient outcomes by enabling timely intervention for common retinal pathologies.

Main Methods:

• Development of a convolutional neural network (CNN) model from scratch for image analysis.
• Utilized spectral-domain optical coherence tomography (SD-OCT) scans for pathology detection.
• Trained and evaluated the CNN model on the Duke dataset of OCT images.

Main Results:

• The developed CNN model achieved a classification accuracy exceeding 99% on the Duke dataset.
• The deep learning approach demonstrated high efficacy in identifying DME and AMD from SD-OCT scans.
• The method provides a robust and accurate tool for automated eye disease detection.

Conclusions:

• Automatic deep learning methods can accurately detect DME and AMD using SD-OCT scans.
• High-accuracy detection facilitates early diagnosis and timely treatment, preventing vision loss.
• This technology holds significant potential for improving ophthalmic care and patient outcomes.