Medical Image Training Data Selection From Test Reports
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning systems in the medical field face challenges in accurately acquiring large amounts of necessary data for lesion detection and classification due to the inclusion of unsuitable or irrelevant data, leading to a decrease in learning accuracy.
Innovation Solution
A learning device and method that extracts standard images from medical test reports, sorts training data based on these images, and uses machine learning to identify appropriate data for learning, ensuring high accuracy and relevance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If all acquired data is used in learning, then the quantity of training data increases, but the learning accuracy decreases due to inclusion of wrong type of data and unsuitable data
Solution Approach 1:
The patent extracts only the necessary and suitable data from the acquired medical data by using test reports as a reference. The test report-based extraction method identifies and extracts only the data that is appropriate for learning, separating it from the large volume of acquired data that includes unsuitable portions.
Solution Approach 2:
The patent introduces test reports as an intermediary to bridge the gap between acquired data and suitable training data. The test reports serve as a mediator that enables accurate identification and extraction of appropriate data, allowing the system to filter out unsuitable data while maintaining high learning accuracy.
2Reliability
If test reports are used to extract necessary data with high accuracy, then the learning accuracy improves, but the efficiency of acquiring large amounts of images decreases
Solution Approach 1:
The patent performs preliminary extraction of standard images from test reports before extracting the actual training data. By pre-identifying relevant standard images based on test report information, the system prepares the necessary reference material in advance, which streamlines the subsequent data extraction process and improves overall efficiency.
Solution Approach 2:
The patent segments the data extraction process into distinct stages: first extracting standard images from test reports, then using those standard images as references to extract actual training data. This segmentation allows each stage to be optimized independently, improving the overall efficiency of acquiring large amounts of training images.
3Measurement precision
If manual extraction of images from test reports is performed, then the accuracy of data extraction improves, but the time and labor required increases significantly
Solution Approach 1:
The patent replaces the mechanical manual extraction process with an automated image processing system. The system uses computer vision and pattern recognition algorithms to automatically extract standard images from test reports and subsequently extract training data, eliminating the need for manual intervention while maintaining high extraction accuracy.
Solution Approach 2:
The patent enables the system to perform self-service extraction by automatically identifying and extracting images based on test report information without human intervention. The automated system processes the extraction task independently, significantly reducing the time and labor required while maintaining consistent accuracy.
Data Source
AI summary
A standard image is extracted from a still image group or a moving image associated with medical test report data stored in a database, using the medical test report data. A frame image group is created from the still image group or the moving image that configures the standard image, a learning candidate image data set is extracted from the frame image group based on the standard image, and training data is sorted out from the learning candidate data set. Learning is performed using the training data which is sorted out.