Medical Image Labeling via Learned Network Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The time and effort required for manual labeling of medical images for supervised learning in artificial intelligence are excessive, necessitating a more efficient method for lesion region labeling.
Innovation Solution
A labeling method and device that utilize a learned network function to detect regions of interest in medical images, allowing user input for correction and labeling of lesion regions, with the ability to increase resolution for precise selection and return to original resolution for accurate labeling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual labeling is performed directly by a person, then labeling accuracy can be ensured, but the time needed for labeling increases exponentially
Solution Approach 1:
The system performs preliminary automated detection of regions of interest and lesion estimation regions using a learned network function before the user performs manual labeling. This pre-processing step narrows down the areas that require manual verification, significantly reducing the overall labeling time while maintaining accuracy through user correction of the pre-detected regions.
Solution Approach 2:
The patent introduces an automated detection system as an intermediary between complete manual labeling and fully automated labeling. The system detects regions of interest and generates lesion estimation regions that serve as intermediate results, which are then refined by user input. This intermediary approach balances automation efficiency with human oversight for accuracy.
2Manufacturing precision
If the resolution of the region of interest is increased for precise selection, then labeling precision is improved, but processing time and computational resources increase
Solution Approach 1:
The system segments the medical image into regions of interest based on detected anatomical structures or pathological features. By focusing computational resources and resolution enhancement only on these segmented regions rather than the entire image, the system achieves high labeling precision while minimizing processing time and computational overhead.
Solution Approach 2:
The patent applies local quality enhancement by increasing resolution only for specific regions of interest that require precise labeling, rather than processing the entire image at high resolution. This localized approach maintains labeling precision where needed while reducing overall processing time and computational resource consumption.
Data Source
AI summary
Provided is a labeling method and a computing device therefor. The method includes the steps of: acquiring a medical image, receiving the medical image, detecting regions of interest corresponding to at least one subject disease through a learned network function, and displaying the regions of interest on the medical image, and labeling a lesion region corresponding to the subject disease on the medical image according to the regions of interest and user inputs to the regions of interest.