Spinal Image Graph Labeling for Accurate Vertebra Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods for identifying and labeling vertebrae in tomographic images, such as those using cascaded convolutional neural networks, are prone to errors that propagate and hinder accurate labeling of individual vertebrae.
Innovation Solution
An image processing apparatus and method that extracts key points from an image, constructs a graph structure using these points, and associates them with labels of different layers to accurately label vertebrae and ribs by analyzing a graph structure using a graph cut method.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If individual vertebrae are recursively extracted one by one using cascaded convolutional neural networks, then the labeling process can be performed, but errors in extraction processing propagate and affect subsequent extraction processing, making it difficult to accurately label vertebrae
Solution Approach 1:
The patent segments the labeling task into two independent stages: first extracting key points with hierarchical labels (cervical spine, thoracic spine, lumbar spine), then separately extracting individual vertebrae labels. This segmentation prevents error propagation by isolating the hierarchical labeling from the individual vertebrae labeling, allowing each stage to operate independently without affecting the other.
Solution Approach 2:
The patent introduces key points as intermediary elements that carry hierarchical labels. These key points serve as mediators between the image and the final vertebrae labels. By using key points as intermediaries, the system can first establish the hierarchical structure and then use this structure to guide the extraction of individual vertebrae labels, preventing direct error propagation in the recursive extraction process.
2Measurement precision
If a graph structure is constructed and analyzed to associate nodes with labels of the second layer, then accurate labeling of individual vertebrae can be achieved, but the processing complexity increases
Solution Approach 1:
The patent segments the labeling process into two distinct stages: first extracting key points with hierarchical labels, then using these key points to extract individual vertebrae labels. This segmentation reduces processing complexity by breaking down the complex task into manageable steps, where each stage has a specific function and can be optimized independently.
Solution Approach 2:
The patent transitions from a two-dimensional image processing approach to a three-dimensional hierarchical labeling approach by introducing key points as intermediaries. This dimensional change allows the system to first establish hierarchical relationships and then use these relationships to guide the extraction of individual vertebrae labels, making the complex task more manageable through an additional organizational dimension.
Data Source
AI summary
A processor extracts, from an image including a plurality of structures that spatially continuously present and whose corresponding labels have a hierarchy, respective key points of the plurality of structures in association with labels of a first layer; uses the key points as nodes to derive a graph structure in which the labels of the first layer are associated with the nodes; and associates the nodes with labels of a second layer lower than the first layer by analyzing the graph structure.