Organ Segmentation Quality Assessment Using Reference Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deep learning-based medical image segmentation techniques face challenges in accurately segmenting organs due to inter-patient variability and inconsistencies, which can lead to inaccurate results in clinical applications like IMRT, necessitating a method to assess the quality of segmentation output.
Innovation Solution
An automated post-processing tool that compares current segmentation features with reference values to determine the quality of organ segmentation, generating reports and alerts for clinicians, and providing recommendations to ensure accurate results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If deep learning-based automated segmentation is used, then productivity is improved by reducing manual workload, but manufacturing precision deteriorates due to insufficient accuracy for clinical applications
Solution Approach 1:
The system implements a feedback mechanism by comparing automated segmentation features against reference values derived from manual segmentations. Quality scores are calculated based on this comparison, and segmentation results are flagged for review when quality thresholds are not met, creating a closed-loop quality control system that maintains high accuracy while using automated methods
Solution Approach 2:
The patent introduces an intermediary quality assessment system that acts as a mediator between automated segmentation and clinical use. This intermediary layer compares segmentation features, calculates quality scores, and determines whether manual review is needed, allowing automated segmentation to be used efficiently while maintaining clinical accuracy standards
2Manufacturing precision
If manual segmentation is performed, then manufacturing precision is improved through expert delineation, but productivity deteriorates due to tedious and time-consuming process
Solution Approach 1:
The system applies partial manual action by performing manual segmentation only on cases where automated quality assessment falls below thresholds. Most cases are processed automatically with high accuracy, while only a subset requiring manual review receives expert attention, optimizing the balance between productivity and precision
3Ease of operation
If automated segmentation is used, then ease of operation is improved by eliminating manual delineation, but reliability deteriorates due to inter-patient variability and inconsistent results
Solution Approach 1:
The system performs preliminary action by pre-calculating reference values from manual segmentations and establishing quality thresholds before automated segmentation is applied. This preliminary preparation enables the automated system to reliably assess quality and consistently identify cases needing review, improving reliability while maintaining ease of operation
Data Source
AI summary
An automated post-processing tool to assess the quality of deep learning-based organ segmentation in medical images is described. According to an example, a method comprises determining, by a system comprising a processor, current values of defined features of respective segmentation masks generated for different anatomical structures included in medical image data via auto-segmentation of the medical image data. The method further comprises determining, by the system, respective measures of correspondence between the current values and corresponding reference values determined for the defined features, determining one or more measures of quality of the auto-segmentation based on the respective measures of correspondence, generating quality assessment report data for the auto-segmentation comprising the one or more measures of quality in standard format that can be displayed by standard clinical software.