Microaneurysms detection in color fundus images using machine learning based on directional local contrast
Shengchun Long1, Jiali Chen2, Ante Hu1
1College of Computer Science and Technology, Zhejiang University of Technology, Hangzhou, 310023, China.
Biomedical Engineering Online
|April 17, 2020
Summary
This study introduces a machine learning method using directional local contrast for detecting microaneurysms, the earliest sign of diabetic retinopathy (DR). The approach offers an effective basis for early DR diagnosis from fundus images.
Area of Science:
- Ophthalmology
- Medical Imaging
- Machine Learning
Background:
- Diabetic retinopathy (DR) is a leading cause of vision loss due to delayed diagnosis.
- Microaneurysms are the earliest clinical sign of DR.
- Accurate detection of microaneurysms in fundus images is crucial for DR screening.
Purpose of the Study:
- To propose a novel machine learning-based method for detecting microaneurysms in color fundus images.
- To utilize directional local contrast (DLC) for improved microaneurysm classification.
- To evaluate the performance of different machine learning techniques for DR diagnosis.
Main Methods:
- Blood vessels enhanced and segmented using Hessian matrix eigenvalue analysis.
- Microaneurysm candidates identified via shape characteristics and connected components analysis, excluding blood vessels.
- Image patches classified as microaneurysm or non-microaneurysm using extracted features and machine learning.
Main Results:
- The proposed DLC method achieved superior accuracy and computation time compared to existing algorithms.
- Area Under the Curve (AUC) values of 0.87 (e-ophtha MA) and 0.86 (DIARETDB1) were obtained.
- Free-response ROC (FROC) scores of 0.374 and 0.210 were recorded on the respective databases.
Conclusions:
- Machine learning utilizing directional local contrast effectively detects microaneurysms in fundus images.
- The method provides a strong scientific foundation for early clinical diagnosis of diabetic retinopathy.
- This technique aids in timely intervention, potentially preventing vision impairment.
