Medical Image Segmentation Threshold Adjustment for Overlapping Organs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing medical image processing technologies face challenges in accurately segmenting internal organs when their areas overlap, as current methods struggle to distinguish between overlapping regions, leading to inaccurate segmentation results.
Innovation Solution
A medical information processing apparatus that computes the probability of correspondence for each area on a medical image regarding specific tissues, acquires and adjusts thresholds for each tissue, and displays overlapping areas, allowing users to adjust thresholds to prevent overlap and improve segmentation accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple DL inferrers are used to segment different internal organs, then segmentation coverage is improved, but segmentation accuracy deteriorates when organ areas overlap
Solution Approach 1:
An overlapping area detection unit is introduced as an intermediary component that identifies regions where multiple organ mask images overlap. This mediator detects the conflict between multiple DL inferrer outputs and enables subsequent resolution through threshold adjustment or priority-based selection, thereby maintaining both comprehensive coverage and accurate segmentation in overlapping regions.
Solution Approach 2:
The system adjusts the threshold values of DL inferrers dynamically to resolve overlapping issues. By changing the parameter (threshold) of the inferrers, the system can suppress false positive detections in overlapping areas while maintaining sensitivity in non-overlapping regions, thus preserving both coverage and accuracy.
2Measurement precision
If threshold values are adjusted to prevent overlapping areas, then segmentation accuracy is improved, but operational complexity increases
Solution Approach 1:
The overlapping area detection unit automatically identifies regions where mask images overlap and provides this information to the threshold adjustment mechanism. This self-service approach enables the system to autonomously resolve overlapping issues without requiring manual intervention or complex operational procedures, thus improving accuracy while maintaining operational simplicity.
Solution Approach 2:
The system implements a feedback loop where the detection of overlapping areas informs subsequent threshold adjustments. The overlapping detection results are fed back to the threshold control mechanism, which automatically modifies thresholds to eliminate overlaps, creating a closed-loop system that improves accuracy without increasing operational complexity.
3Measurement precision
If overlapping areas are displayed for user review, then segmentation accuracy is improved, but processing time increases
Solution Approach 1:
The system applies partial action by displaying overlapping area information only in regions where conflicts actually occur, rather than processing or displaying the entire medical image. This selective approach allows users to review and correct only the problematic areas, improving segmentation accuracy while minimizing the time loss associated with user review.
Data Source
AI summary
A medical information processing apparatus includes processing circuitry. The processing circuitry being configured to: compute a probability of correspondence to a specific extraction target with respect to each of areas on a medical image, in regard to each of kinds of extraction targets; acquire a plurality of thresholds that are set for the respective kinds of the extraction targets, and determine whether the area corresponds to the extraction target, based on the threshold and the probability, in regard to each of the areas and in regard to each of the kinds of the extraction targets; and adjust the threshold in accordance with an input of the threshold.