DISCERN: Generative Framework for Vessel Segmentation using Convolutional Neural Network and Visual Codebook.

Area of Science:

• Medical Imaging
• Computer Vision
• Biomedical Engineering

Background:

• Accurate segmentation of retinal blood vessels is crucial for diagnosing various eye diseases.
• Existing methods often struggle with accuracy, robustness, and efficiency.

Purpose of the Study:

• To develop and evaluate a novel two-stage framework for automated retinal vessel segmentation.
• To improve the accuracy and efficiency of vessel segmentation in retinal fundus images.

Main Methods:

• A two-stage framework utilizing convolutional neural networks (CNNs) and Totally Random Trees Embedding.
• Stage one: CNN for patch correlation with groundtruth reduction.
• Stage two: Visual codebook generation and a generative nearest neighbor search space for unseen patches.

Main Results:

• The framework successfully generates segmentation patches not seen during training.
• Achieved superior performance compared to state-of-the-art methods on DRIVE and STARE datasets.
• Demonstrated high accuracy, robustness, speed, and simplicity.

Conclusions:

• The proposed framework is highly suitable for automated retinal vessel segmentation.
• Offers a significant advancement in the field of medical image analysis for ophthalmology.
• Potential for clinical application in early disease detection and monitoring.