Medical Image Segmentation With Threshold-Based Anchor Points
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image analysis techniques in medical imaging lack the desired accuracy for disease screening, diagnosis, and treatment management, necessitating human intervention to correct imprecise data generated by machine learning models, which is not fully embraced by the medical and regulatory communities.
Innovation Solution
An image analysis system that generates anchor points and indicators on a graphical user interface, allowing users to efficiently adjust anatomical region boundaries, using a threshold-based selection process to identify critical points along boundaries, reducing the need for manual adjustment of every pixel and enhancing the accuracy of machine learning model training.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If machine learning models are used for image segmentation, then processing speed and initial classification accuracy are improved, but boundary precision and region definition accuracy deteriorate
Solution Approach 1:
The patent segments the boundary into discrete anchor points that can be independently adjusted. Instead of requiring manual pixel-by-pixel adjustment, the system identifies and adjusts only at these anchor points, maintaining the speed advantage while improving boundary precision through targeted human intervention at critical locations.
Solution Approach 2:
The machine learning model performs self-correction through an iterative process where it generates initial segmentations, the system identifies anchor points for potential adjustment, and the model re-trains using corrected data. This allows the model to improve its own boundary precision without requiring complete manual re-segmentation of entire images.
2Manufacturing precision
If manual adjustment of every pixel is performed, then boundary precision is improved, but time consumption and labor requirements worsen
Solution Approach 1:
The patent extracts only the critical anchor points from the continuous boundary for manual adjustment. By identifying and isolating these specific points where boundaries need correction, the system eliminates the need for time-consuming pixel-by-pixel manual adjustment while maintaining high boundary precision through focused intervention at key locations.
Solution Approach 2:
Instead of requiring complete manual adjustment of all boundary pixels (excessive action), the system applies partial action by adjusting only the necessary anchor points. This partial adjustment approach achieves sufficient boundary precision without the time cost of full manual pixel adjustment.
3Manufacturing precision
If anchor points are densely distributed, then boundary accuracy is improved, but computational complexity and processing time worsen
Solution Approach 1:
The patent dynamically adjusts the density and distribution of anchor points based on the specific image characteristics and segmentation requirements. By changing the parameter of anchor point density adaptively rather than using a fixed high density, the system achieves high boundary accuracy when needed while reducing computational complexity in regions where fine adjustment is not required.
Data Source
AI summary
A method that is implemented by one or more computer devices and that includes receiving an image associated with a portion of anatomy of a subject. Boundary points are extracted for an initial boundary associated with a corresponding region of pixels in the image. The boundary points are evaluated according to a sequential order to select anchor points. The evaluating includes determining that a current boundary point being evaluated is a next anchor point when at least one perpendicular distance computed for a portion of the initial boundary located between a previous anchor point and the current boundary point with respect to a line extending between the previous anchor point and the current boundary point is greater than a selected threshold. An anchor point image is generated for display in a graphical user interface on a display device. The anchor point image includes anchor indicators that represent the anchor points.