Detection of neovascularization in retinal images using multivariate m-Mediods based classifier

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Science

Background:

• Diabetic retinopathy is a major global cause of blindness.
• Proliferative diabetic retinopathy involves abnormal new blood vessel growth.
• Early detection is critical for preserving patient vision.

Purpose of the Study:

• To develop an automated system for detecting abnormal blood vessels in proliferative diabetic retinopathy.
• To grade proliferative diabetic retinopathy using digital retinal images.
• To improve early and accurate diagnosis of this vision-threatening condition.

Main Methods:

• A novel multivariate m-Mediods based classifier was proposed.
• Vascular patterns and optic disc were extracted using multilayered thresholding and Hough transform.
• Fundus images were graded based on classification and optic disc coordinates.

Main Results:

• The proposed method demonstrated high accuracy in detecting abnormal blood vessels.
• The system effectively graded proliferative diabetic retinopathy.
• Evaluation on public retinal image databases confirmed the system's performance.

Conclusions:

• The developed automated system shows significant promise for early and accurate detection and grading of proliferative diabetic retinopathy.
• This method can aid ophthalmologists in timely diagnosis and treatment planning.
• Further research can explore integration into clinical workflows for improved patient outcomes.