Automatic Aortic Segmentation Using Confidence Scores
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image processing techniques require user input for segmenting the aorta, which can be time-consuming and prone to errors, and there is a need for improved speed, accuracy, and ease of use in diagnostic imaging.
Innovation Solution
A system and method that automatically detect and segment the aorta using a processor to search axial image slices, calculate confidence scores for identifying seed disks, and grow the segmentation of the aorta without requiring user-specified seed locations, utilizing sub-sampled data for efficient processing and refining the segmentation with higher resolution data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If user seed input is required for aorta segmentation, then segmentation accuracy can be controlled by user expertise, but processing time increases and ease of operation decreases
Solution Approach 1:
The system performs automatic aorta detection and segmentation without requiring user seed input. The processor automatically searches axial image slices, calculates confidence scores, identifies seed disks, and grows the segmentation using sub-sampled data, making the system self-sufficient and eliminating the need for user intervention in the segmentation process
Solution Approach 2:
The system pre-processes image data by creating sub-sampled versions of axial image slices before segmentation. This preliminary action enables efficient automatic detection and seed disk identification without requiring user input during the actual segmentation process
2Ease of operation
If automatic segmentation without user input is implemented, then ease of operation improves, but segmentation accuracy may be compromised
Solution Approach 1:
The system calculates confidence scores for each axial image slice during the search process to determine whether it contains a cross-section image of the aorta. This feedback mechanism allows the system to automatically verify and adjust its segmentation process, ensuring accuracy without user intervention
Solution Approach 2:
The patent replaces the mechanical user interaction system (requiring users to manually select seed points) with an automated computational system that uses image processing algorithms, confidence score calculation, and automatic seed disk identification to achieve segmentation without human input
3Productivity
If sub-sampled data is used for processing, then processing speed increases, but image resolution decreases
Solution Approach 1:
The system divides the original high-resolution image data into sub-sampled versions, processing different levels of resolution at different stages. This segmentation of the data allows efficient automatic detection using lower-resolution sub-sampled data while preserving the ability to achieve accurate segmentation through multi-scale processing
Data Source
AI summary
A method including searching image data corresponding to a series of axial image slices with a processor, searching axial image slices from a starting image slice and calculating a confidence score that an image slice includes a cross-section image of an aorta, identifying an image slice containing at least one seed disk, including an ascending aorta seed disk, from candidate image slices identified according to the confidence score, and growing a 3D segmentation of the ascending aorta by stacking ascending aorta image disks included in consecutive image slices beginning from the ascending aorta seed disk.