Fundus Blood Vessel Segmentation for Reliable Whole-Image Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional fundus image processing methods rely on blood vessels within limited regions around the optic papilla, leading to potential inaccuracies due to diseases, opacity, and image artifacts, affecting the reliability and accuracy of acquired information.
Innovation Solution
A fundus image processing device and non-transitory computer-readable storage medium using a machine learning algorithm to process fundus images, enabling the acquisition of blood vessel information from a wider region, including the optic papilla and macula, and adjusting image regions to improve accuracy and reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If information is acquired from a limited region around the optic papilla, then the processing is simple, but the reliability and accuracy of blood vessel information are inferior
Solution Approach 1:
The fundus image is segmented into multiple regions including a first region around the optic papilla and a second region away from it. Blood vessel information is extracted from both regions and integrated, allowing the system to maintain simplicity in individual region processing while achieving high reliability through comprehensive multi-region analysis
Solution Approach 2:
The approach transitions from analyzing a single two-dimensional region around the optic papilla to analyzing multiple two-dimensional regions across the entire fundus image. This dimensional expansion from one localized area to multiple distributed areas enables comprehensive blood vessel assessment without excessive complexity
2Measurement precision
If information is acquired from the whole fundus image including tissues other than blood vessels, then more comprehensive information is obtained, but diseases, opacity, and artifacts affect the accuracy
Solution Approach 1:
The system extracts only blood vessel information from the fundus image using specialized algorithms that isolate vascular structures from other tissues. By extracting specifically what is needed (blood vessels) and excluding other elements (retinal tissues, artifacts, opacity-affected areas), the system achieves high measurement precision while filtering out harmful factors
Solution Approach 2:
The patent employs an intermediary processing layer that acts as a mediator between the raw fundus image and the final blood vessel measurements. This intermediary layer includes algorithms that specifically target blood vessel structures and filter out interference from diseases, opacity, and artifacts, thereby protecting the measurement accuracy from harmful factors
3Loss of information
If only blood vessels within a specific region are analyzed, then the processing is straightforward, but the information acquired is insufficient for reliable diagnosis
Solution Approach 1:
The system merges blood vessel information from multiple regions including the first region around the optic papilla and the second region away from it. By combining data from these distributed regions into a unified analysis, the system achieves complete blood vessel information coverage without creating excessive processing complexity, as the merging is performed through systematic integration of standardized regional data
Data Source
AI summary
A processor of a fundus image processing device acquires a fundus image 50 photographed by a fundus image photographing device. The processor acquires blood vessel images 60A and 60B that indicate at least one of an arteriole and a venule in the fundus image 50 by inputting the fundus image 50 into a mathematical model trained by a machine learning algorithm. The processor acquires a blood vessel area that is an area of at least one of the arteriole and the venule in the whole of the blood vessel images 60A and 60B.