Carpal Segmentation in Pediatric Wrist X-rays
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for carpal bone segmentation and recognition in young children are inaccurate and difficult due to uncertainties in the number and shape of carpal bones, obscure boundaries with soft tissues, and uneven density, leading to challenges in effective segmentation and recognition.
Innovation Solution
A carpal segmentation and recognition method using adaptive threshold segmentation with variable threshold windows and edge detection, combined with a carpal anatomy priori model for boundary optimization, to accurately segment and recognize carpal bones in child orthotopic wrist X-ray images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional threshold segmentation is used on carpal region images, then the segmentation process is simple, but the segmentation accuracy is low due to obscure boundaries and uneven density
Solution Approach 1:
The patent divides the carpal region image into multiple local windows and applies threshold segmentation to each window separately. This local segmentation approach allows the algorithm to adapt to varying density conditions in different regions, improving overall segmentation accuracy while managing complexity through systematic processing
Solution Approach 2:
The patent employs adaptive threshold values for different local windows based on their specific density characteristics. Each window receives customized threshold parameters according to its local image properties, enabling accurate segmentation of regions with obscure boundaries and uneven density without requiring complex global processing
2Reliability
If single-method segmentation is used, then the process is straightforward, but it cannot accurately segment carpal bones with obscure boundaries or extract special carpal bones
Solution Approach 1:
The patent combines multiple segmentation methods (threshold segmentation, edge detection, and region growing) into a composite segmentation system. Each method complements the others: threshold segmentation handles uniform regions, edge detection captures obscure boundaries, and region growing connects fragmented segments, achieving reliable segmentation of all carpal bone types including special cases
Solution Approach 2:
The patent merges results from multiple segmentation methods by integrating their outputs. The combination of threshold-based segmentation, edge detection, and region growing approaches creates a robust segmentation system that overcomes the limitations of any single method, accurately identifying all carpal bones including those with obscure boundaries or unusual characteristics
3Adaptability or versatility
If fixed threshold segmentation is used, then the algorithm is simple, but it fails to adapt to varying density and contrast in different carpal regions
Solution Approach 1:
The patent transforms the static fixed threshold approach into a dynamic adaptive threshold system. Threshold values are automatically adjusted for each local window based on its specific image characteristics such as density and contrast. This dynamic adaptation allows the segmentation algorithm to effectively handle varying conditions across different carpal regions without requiring manual intervention
Solution Approach 2:
The patent changes the threshold parameter dynamically according to local image properties. Instead of using a constant threshold value, the system calculates adaptive thresholds for each window based on local density and contrast measurements. This parameter adaptation enables the segmentation to respond to varying conditions in different carpal regions, improving versatility while maintaining algorithmic simplicity through automated parameter adjustment
Data Source
AI summary
The present application relates to the technical field of image recognition, and provides a carpal segmentation and recognition method, including: performing threshold segmentation on a carpal region of interest on a child orthotopic wrist X-ray image based on an adaptive threshold segmentation manner of variable threshold segmentation windows, and extracting edge information of the carpal region of interest based on an edge detection manner; combining a binarized image obtained by performing the threshold segmentation with the extracted edge information to obtain an initial segmentation image; performing carpal recognition on the initial segmentation image by using a carpal anatomy priori model to obtain an initial recognition image including information of each carpal bone; and performing boundary optimization on the initial recognition image, and outputting a carpal recognition image obtained after the boundary optimization is performed.