ELEMENT: Multi-Modal Retinal Vessel Segmentation Based on a Coupled Region Growing and Machine Learning Approach.

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Retinal vascular structures are crucial for diagnosing ocular diseases like AMD, diabetic retinopathy, and glaucoma.
• Current retinal vessel segmentation methods are manual or interactive, leading to time consumption and human errors.

Purpose of the Study:

• To propose a new multi-modal framework, ELEMENT, for automated retinal vessel segmentation.
• To improve the accuracy and efficiency of retinal vessel segmentation for enhanced ocular disease diagnosis.

Main Methods:

• Developed ELEMENT (vEsseL sEgmentation using Machine lEarning and coNnecTivity), a framework combining feature extraction and pixel-based classification.
• Utilized region growing and machine learning with features capturing grey level and vessel connectivity properties.
• Propagated connectivity information during classification to reduce inconsistencies.

Main Results:

• ELEMENT achieved high overall accuracy (97.40% on DRIVE dataset), outperforming 25 of 26 state-of-the-art methods.
• The framework surpassed all evaluated state-of-the-art methods on STARE (98.27%), CHASE-DB (97.78%), VAMPIRE FA (98.34%), IOSTAR SLO (98.04%), and RC-SLO (98.35%) datasets.
• ELEMENT demonstrated reduced inconsistencies and increased segmentation throughput.

Conclusions:

• ELEMENT offers a superior approach to retinal vessel segmentation compared to existing methods, including deep learning techniques.
• The framework's high accuracy and efficiency support its potential for improving the diagnosis of various ocular diseases.
• Automated segmentation using ELEMENT can significantly reduce diagnostic time and human error in ophthalmology.