Presentation of a Segmentation Method for a Diabetic Retinopathy Patient's Fundus Region Detection Using a...

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Diabetic retinopathy (DR) is a common complication of diabetes, characterized by retinal blood flow abnormalities.
• Current DR diagnosis relies on manual fundus examination, which is time-consuming and costly.
• Early detection and monitoring are crucial for managing DR and preventing vision loss.

Purpose of the Study:

• To develop and evaluate a deep learning model for automated detection of diabetic retinopathy lesions.
• To assess the efficacy of a Convolutional Neural Network (CNN) architecture in identifying target regions in retinal fundus images.

Main Methods:

• Utilized the IDRiD dataset comprising fundus images of patients with and without diabetic retinopathy.
• Developed a CNN architecture for the automated detection of diabetic retinopathy lesions.
• Trained and validated the CNN model on a cohort of 80 patients' fundus imagery.

Main Results:

• The proposed CNN architecture achieved an accuracy of 83.84% in detecting target locations indicative of diabetic retinopathy.
• The model demonstrated proficiency in identifying both normal retinal structures and characteristic diabetic retinopathy lesions.
• The automated detection approach shows promise for efficient screening and monitoring.

Conclusions:

• The developed CNN model offers a potential automated solution for diabetic retinopathy detection.
• This technology can assist ophthalmologists in monitoring patients and managing the disease more effectively.
• Automated analysis of retinal images can help overcome the limitations of manual examination for widespread DR screening.