Retinal Vessel Segmentation in OCTA Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for segmenting retinal blood vessels from optical coherence tomography (OCT) images face challenges due to low contrast, variability in vessel sizes, and the presence of pathologies, leading to inefficient and inaccurate results, especially in distinguishing blood vessels from the background.
Innovation Solution
An automated segmentation system using a preprocessing stage with Generalized Gauss-Markov random field (GGMRF) and regional dynamic histogram equalization (RDHE) models, followed by an initial segmentation stage integrating intensity and spatial higher-order joint Markov-Gibbs random field models, and a refinement stage with a 2D connectivity filter to extract connected regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If threshold-based segmentation techniques are used, then the segmentation process is simple and fast, but significant errors occur due to similarities in gray levels between blood vessels and background
Solution Approach 1:
The patent divides the segmentation process into multiple stages: preprocessing to enhance contrast, initial segmentation using region growing with multiple seed points, and refinement steps. This multi-stage segmentation approach overcomes the limitations of simple threshold-based methods by progressively improving accuracy while maintaining computational efficiency.
Solution Approach 2:
The patent introduces an intensity map as an intermediary representation that enhances the contrast between blood vessels and background before segmentation. This intermediate processing step transforms the low-contrast original image into a high-contrast intensity map, enabling more accurate segmentation without requiring complex direct segmentation algorithms.
2Reliability
If region growing segmentation techniques are used, then segmentation can capture vessel continuity, but results are sensitive to the location of the seed point
Solution Approach 1:
The patent makes the region growing algorithm universal by implementing automatic seed point selection that works across different image types and conditions. The system automatically identifies multiple seed points based on intensity characteristics, eliminating the need for manual seed point placement and ensuring consistent performance regardless of image variations.
Solution Approach 2:
The segmentation system performs self-service by automatically selecting seed points without user intervention. The algorithm identifies potential seed points based on intensity map characteristics and automatically initiates region growing from multiple locations, making the process robust to variations in vessel appearance and background conditions.
3Ease of manufacture
If fundus images are used for segmentation, then the imaging process is simple, but the resolution is low and depth information is lacking
Solution Approach 1:
The patent transforms the imaging parameters by converting standard fundus images into intensity maps with enhanced contrast characteristics. This parameter transformation emphasizes the optical properties of blood vessels while suppressing background variations, effectively increasing the resolution and detectability of vessel details without changing the fundamental imaging modality.
4Measurement precision
If manual observation and measurement of blood vessel changes are performed, then diagnostic accuracy can be high, but the process is very complex and requires highly trained persons
Solution Approach 1:
The segmentation system performs self-service by automatically executing the entire segmentation and measurement process without human intervention. The algorithm automatically identifies vessels, measures their characteristics, and generates diagnostic information, replacing the need for manual observation and measurement by trained professionals.
Solution Approach 2:
The patent replaces the mechanical process of manual observation and measurement with an automated computational system. The segmentation algorithm performs all measurements and analyses that would otherwise require manual intervention, substituting human expertise with automated image processing and analysis techniques.
Data Source
AI summary
Methods for automated segmentation system for retinal blood vessels from optical coherence tomography angiography images include a preprocessing stage, an initial segmentation stage, and a refining stage. Application of machine-learning techniques to segmented images allow for automated diagnosis of retinovascular diseases, such as diabetic retinopathy.