Fundus Image Processing for Accurate Non-Perfusion Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in accurately predicting non-perfusion areas in fundus images of the eye, which are regions with little or no blood flow due to retinal capillary vascular bed obstruction, limiting effective diagnosis and treatment of conditions like diabetic retinopathy and retinal vein occlusion.
Innovation Solution
An image processing method involving first and second image processing stages to identify non-perfusion areas, followed by a prediction process using enhanced vascular images, and displaying these areas on a combined fundus image for diagnosis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If single image processing method is used to identify non-perfusion areas, then processing time is reduced, but identification accuracy deteriorates
Solution Approach 1:
The image processing is divided into two distinct processing units: a first image processing unit that performs initial identification of non-perfusion areas, and a second image processing unit that performs additional identification. This segmentation allows each unit to specialize in different aspects of detection, improving overall accuracy while maintaining efficient processing through parallel or sequential operation.
Solution Approach 2:
The patent combines the results from the first image processing unit and the second image processing unit to generate a comprehensive non-perfusion area map. By merging multiple processing outcomes, the system achieves higher identification accuracy than any single processing method could provide alone, while the combined processing is optimized to avoid excessive time consumption.
2Measurement precision
If multiple image processing methods are combined to improve accuracy, then identification accuracy is improved, but device complexity increases
Solution Approach 1:
The processing system is segmented into two functional units with distinct processing characteristics. The first unit handles primary non-perfusion area identification, while the second unit handles supplementary identification. This segmentation manages complexity by creating modular, independently manageable processing components rather than a monolithic complex system.
Solution Approach 2:
The patent applies partial action by having the first image processing unit perform sufficient processing to identify major non-perfusion areas, and the second unit perform additional processing only where needed to improve accuracy. This approach achieves high identification accuracy without requiring both units to perform exhaustive processing, thereby controlling system complexity.
3Reliability
If detailed image processing is performed to accurately identify non-perfusion areas, then diagnostic reliability is improved, but processing time increases
Solution Approach 1:
The first image processing unit performs processing sufficient to identify the majority of non-perfusion areas with good accuracy, while the second unit performs additional processing only to refine and improve accuracy for difficult-to-detect areas. This partial/excessive action approach achieves high diagnostic reliability without requiring exhaustive detailed processing across the entire image, thus controlling processing time.
Solution Approach 2:
The first image processing unit performs preliminary identification of non-perfusion areas before the second unit performs additional processing. This preliminary action allows the system to establish a baseline detection that captures most cases, and then apply more detailed processing only where necessary, improving reliability without proportionally increasing processing time.
Data Source
AI summary
An enhancement image processing section performs enhancement image processing on a fundus image of a subject eye to enhance vascular portions (304). A prediction processing section predicts a non perfusion area in the fundus image that has been subjected to the enhancement image processing (306 to 312). A generation section generates a non perfusion area candidate image (314).