Supervised retinal vessel segmentation from color fundus images based on matched filtering and AdaBoost classifier

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Retinal blood vessel analysis is crucial for diagnosing eye diseases such as diabetes and hypertension.
• Accurate segmentation of the retinal vasculature is a challenging but essential task in medical image analysis.

Purpose of the Study:

• To propose an automated method for retinal vessel segmentation using matched filter techniques and an AdaBoost classifier.
• To enhance the accuracy and efficiency of retinal blood vessel extraction from fundus images.

Main Methods:

• Image preprocessing included morphological operations, contrast limited adaptive histogram equalization (CLAHE), and Retinex for inhomogeneity correction.
• Blood vessels were enhanced using a combination of B-COSFIRE and Frangi matched filters.
• An AdaBoost classifier, utilizing pixel-wise statistical features, was employed for vessel network extraction, followed by postprocessing.

Main Results:

• The proposed method demonstrated high accuracy on publicly available datasets: DRIVE (0.972), STARE (0.951), and CHASE_DB1 (0.948).
• Segmentation results were comparable to state-of-the-art methods and closely matched manual segmentations.
• The method effectively extracts the retinal blood vessel network.

Conclusions:

• The developed automated retinal vessel segmentation method is accurate and reliable.
• This technique shows significant potential for clinical applications in diagnosing eye diseases.
• The combination of matched filters and AdaBoost provides a robust approach for retinal image analysis.