Fundus Image Segmentation for Retinal Vessel Quality and Measurement

Solution Overview

Problem

Existing fundus image analysis systems face challenges in providing fast, reliable, and detailed retinal vessel quantification due to complexity in image processing, variations in image quality and resolution, and difficulties in segmenting broken vessels and classifying arteries and veins, especially at branching or crossing points.

Innovation Solution

A fundus image analysis system utilizing a pre-segmentation quality assessment, a segmentation module with a four-stacked lightweight U-Net architecture, and a measurement module to compute region-specific and global geometric measures, including a retinal vessel segmentation root and branches for arteries, veins, and optic discs, with post-segmentation quality assessment to exclude poor-quality images.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If deep learning methods are used for retinal vessel segmentation, then processing speed and accuracy are improved, but system complexity and training difficulty increase

Engineering Contradiction:
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The deep learning system is divided into multiple independent U-Net modules, each responsible for specific segmentation tasks (vessel segmentation, artery segmentation, vein segmentation, optic disc segmentation). This modular architecture improves processing efficiency while managing system complexity through functional decomposition.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system processes images through multiple dimensional transformations including image pyramids (multi-scale analysis) and multi-channel feature extraction. This enables comprehensive vessel segmentation while maintaining manageable computational complexity at each processing stage.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Measurement precision

If deep learning methods are used for retinal vessel segmentation, then segmentation accuracy is improved, but training time and computational resources increase

Engineering Contradiction:
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The training process is segmented into multiple stages, with each U-Net module trained independently on specific segmentation tasks. This staged training approach achieves high overall accuracy while reducing total training time compared to training a single comprehensive model.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system performs pre-segmentation quality assessment and image preprocessing before main segmentation. This preliminary action prepares data optimally for each module, improving training efficiency and reducing the computational burden during main processing.

Inventive Principle:
Principle #10Preliminary action

3Reliability

If comprehensive image quality assessment is performed, then measurement reliability is improved, but processing time increases

Engineering Contradiction:
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

Image quality assessment is performed as a preliminary step before main segmentation. This early quality filtering ensures that only suitable images proceed to detailed processing, improving measurement reliability while minimizing time expenditure on poor-quality images.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The quality assessment process is divided into multiple independent evaluation modules that assess different image quality aspects separately. This segmented approach improves assessment comprehensiveness and reliability while maintaining efficient processing through parallel evaluation.

Inventive Principle:
Principle #1Segmentation

4Measurement precision

If multiple segmentation modules are used for detailed vessel classification, then measurement precision is improved, but system complexity increases

Engineering Contradiction:
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The segmentation system is divided into specialized modules: vessel segmentation module, artery segmentation module, vein segmentation module, and optic disc segmentation module. Each module focuses on specific classification tasks, achieving high precision while managing complexity through functional specialization.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The U-Net architecture used in each module is a universal framework that can be adapted to different segmentation tasks. This multi-functional base architecture reduces overall system complexity by using a proven, reusable design pattern across all segmentation modules.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentUS20260013719A1Fundus image analysis system
Publication Date: 2026.01.15
  • US20260013719A1 patent drawing
  • US20260013719A1 patent drawing
  • US20260013719A1 patent drawing

AI summary

Described herein is a fundus image analysis system including a pre-segmentation image quality assessment module for receiving a fundus input image, and performing overall retinal image quality assessment and measurement quality assessment on the fundus input image; segmentation module for segmenting retinal vessel, artery, vein and optic disc to produce segmentation maps from the fundus input image; and a measurement module for computing region specific measurements within a standard zone within the fundus input image, and global physical or geometric measures of the whole fundus input image.