Punctured window based multiscale line detector for efficient segmentation of retinal blood vessels

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Retinal vasculature changes are key indicators for diagnosing diseases like diabetes and hypertension.
• Accurate segmentation of retinal vessels is crucial for efficient diagnosis and cost-effective treatment.

Purpose of the Study:

• To develop an automated algorithm for retinal vasculature segmentation.
• To improve upon classic line detector methods by addressing background intensity estimation flaws.

Main Methods:

• A novel 'punctured window' approach is introduced to enhance line detection in low-contrast regions.
• Image denoising using a fractional filter, followed by multi-scale line detection with punctured windows.
• Final segmentation achieved through hysteresis thresholding after combining multi-scale responses and underlying image intensity.

Main Results:

• The algorithm demonstrates enhanced ability to identify minor vessels in low-contrast areas.
• Tested on four public datasets (RC-SLO, STARE, CHASE_DB1, DRIVE) using class imbalance-unaffected metrics.
• The proposed technique shows comparable performance to state-of-the-art methods, outperforming many.

Conclusions:

• The enhanced line detector with punctured windows significantly improves retinal vessel segmentation accuracy.
• The algorithm provides a robust and consistent performance across multiple datasets and metrics.
• This automated method facilitates efficient diagnosis of retinal vascular diseases.