Unified Tubular Structure Segmentation and Classification in Medical Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for extracting and separating tubular structures like blood vessels in medical images, especially those combining deep learning with traditional post-processing, often fail to achieve accurate segmentation and classification due to separate processing steps requiring parameter adjustments.
Innovation Solution
A method and apparatus that combines local and global features in a single deep learning process to perform simultaneous segmentation and classification of tubular structures, eliminating the need for traditional post-processing and parameter adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional methods (Hessian Matrix, graph cut, vascular structure matching) are used to extract tubular structures, then the processing steps are simple and well-defined, but the segmentation and classification accuracy is insufficient
Solution Approach 1:
The patent combines deep learning-based segmentation and classification into a single unified processing step, eliminating the need for separate traditional post-processing steps. The neural network simultaneously performs both segmentation and classification functions, achieving higher accuracy while simplifying the overall workflow compared to traditional multi-step approaches.
Solution Approach 2:
The deep learning model is designed to perform multiple functions (segmentation and classification) within a single framework. The neural network processes tubular structures by simultaneously extracting segmentation masks and classification labels, making the system more versatile and accurate compared to specialized traditional methods.
2Measurement precision
If deep learning methods are combined with traditional post-processing steps, then the processing can handle complex tubular structures, but parameter adjustments are required and accuracy remains insufficient
Solution Approach 1:
The patent merges deep learning-based segmentation and classification into a single integrated process, eliminating the need for separate post-processing steps that would require parameter adjustments. The unified neural network handles both tasks simultaneously with fixed parameters, improving ease of operation while maintaining high accuracy.
3Productivity
If multi-stage extraction is performed with separate steps for segmentation and classification, then the processing can be optimized for each step, but the overall processing time and computational resources increase
Solution Approach 1:
The patent combines segmentation and classification into a single parallel processing step within the deep learning framework. The neural network processes both tasks simultaneously on the same input data, eliminating sequential processing time and reducing computational overhead compared to multi-stage extraction methods.
Data Source
AI summary
A medical image processing apparatus of an embodiment includes processing circuitry. The processing circuitry receives a medical image of a target region. The processing circuitry generates an image pair including a local image having local features of the target region and a global image having global features of the target region on the basis of the received medical image. The processing circuitry performs segmentation and classification of the target region on the image pair by a neural network.