Lesion Area Extraction Using Lesion-Specific Evaluation Functions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lesion area extraction methods in medical images do not fully utilize the unique characteristics of different lesion types, leading to degraded extraction performance, as they use the same evaluation function for all types of lesions.
Innovation Solution
A lesion area extraction apparatus and method that records and utilizes specific lesion area extraction processing data tailored to each type of lesion, using machine learning-based evaluation functions and position information from radiology reports to accurately extract lesion areas, reducing user workload and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the same evaluation function is used for all types of lesions, then the device complexity is reduced and ease of operation is improved, but the extraction precision and reliability deteriorate
Solution Approach 1:
The patent segments the extraction process by creating separate evaluation functions for different lesion types (e.g., lung nodules, liver lesions, brain tumors). Each lesion type has its own specialized evaluation function that captures its unique characteristics, thereby improving extraction precision without requiring a completely separate system for each type.
Solution Approach 2:
The patent applies local quality by tailoring the evaluation function parameters and characteristics to match specific lesion types. Each lesion type receives a customized evaluation approach rather than a uniform method, allowing the system to optimize extraction precision for each local context while maintaining overall system manageability.
2Measurement precision
If lesion-specific processing data is used for each lesion type, then extraction precision is improved, but the quantity of data and device complexity increase
Solution Approach 1:
The patent creates a universal framework that handles multiple lesion types through a common architecture. The system uses a standardized interface and data structure that can accommodate different lesion types, allowing the same system to process diverse lesion data without requiring completely separate processing pipelines for each type.
Solution Approach 2:
The patent manages data quantity by dynamically adjusting parameters based on lesion type. Rather than storing completely separate datasets for each lesion type, the system modifies evaluation function parameters and characteristics according to the specific lesion being analyzed, reducing redundant data storage while maintaining precision.
3Ease of operation
If a single evaluation function is used for all lesions, then ease of operation is improved, but extraction precision for specific lesion types deteriorates
Solution Approach 1:
The patent introduces dynamics by allowing the system to automatically adapt the evaluation function based on the detected lesion type. The system dynamically selects and adjusts the appropriate evaluation parameters without requiring manual reconfiguration, maintaining ease of operation while achieving lesion-specific precision through automatic adaptation.
4Measurement precision
If multiple lesion-specific evaluation functions are created, then extraction precision is improved, but the complexity of selecting the appropriate function increases
Solution Approach 1:
The patent implements self-service by enabling the system to automatically identify the lesion type and select the appropriate evaluation function without user intervention. The system autonomously determines which evaluation function to apply based on image analysis and lesion characteristics, eliminating the complexity of manual function selection while maintaining high extraction precision.
Data Source
AI summary
Recording a plurality of lesion area extraction processing data generated in advance according to a plurality of types of lesion areas, recording a radiology which includes a character string having a lesion description character and being related to position information of a lesion area in the medical image, determining lesion area extraction processing data used for the extraction from the plurality of lesion area extraction processing data based on the lesion description character provided in the radiology report, and performing the extraction using the determined lesion area extraction processing data and the position information of the lesion area related to the character string.