A novel contrast enhancement technique for diabetic retinal image pre-processing and classification.

Area of Science:

• Ophthalmology
• Medical Imaging
• Machine Learning

Background:

• Diabetic Retinopathy (DR) affects 35-60% of diabetic individuals aged 18-65, posing a significant risk of blindness.
• Early DR diagnosis is crucial for vision preservation, but challenges exist in machine learning analysis of fundus images, especially microaneurysm detection.
• Microaneurysms are key indicators for DR diagnosis, yet their accurate identification in raw fundus images remains difficult.

Purpose of the Study:

• To develop and evaluate a novel pre-processing technique for enhanced diabetic retinopathy classification.
• To address the challenge of accurately detecting microaneurysms and reducing false positives in fundus images.
• To improve the performance of machine learning models in identifying early signs of diabetic retinopathy.

Main Methods:

• A novel pre-processing technique combining Modified Fuzzy C-means Clustering and Support Vector Machine classifier.
• Image processing steps include RGB to HSI conversion, median filtering, Intensity Histogram Equalization, and connected component analysis.
• Morphological operations were used to remove the optic disc, minimizing confusion with microaneurysms.

Main Results:

• The proposed method achieved a high accuracy rate of 99.31% on publicly available datasets.
• Demonstrated superior performance compared to existing state-of-the-art algorithms for microaneurysm detection.
• Significantly improved the detection of microaneurysms while reducing false detections.

Conclusions:

• The proposed pre-processing technique effectively enhances diabetic retinopathy classification by overcoming false microaneurysm detection challenges.
• Comparative analysis confirms the method's effectiveness against state-of-the-art algorithms.
• The study highlights the importance of advanced pre-processing for accurate DR diagnosis.