AI Medical Image Segmentation Auditing via Error Heatmaps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current artificial intelligence-based medical image segmentation technologies lack a reliable method to audit the accuracy of segmentation results, which is crucial for ensuring the stability and reliability of image segmentation processes.
Innovation Solution
A method and apparatus that preprocess medical images and segmentation outputs, generate segmentation error heatmap images using a deep learning model, calculate segmentation error risk, and provide auditing information to users, ensuring accurate output validation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If artificial intelligence-based medical image segmentation is performed, then segmentation efficiency is improved, but segmentation accuracy cannot be guaranteed without auditing
Solution Approach 1:
The patent implements a feedback mechanism by generating segmentation error heatmaps that provide visual feedback on potential errors in the AI segmentation results. The heatmap is generated by comparing the segmented image with the original medical image, and this feedback information is used to identify and correct segmentation errors, thereby improving reliability while maintaining efficiency
Solution Approach 2:
The patent introduces an intermediary auditing system that acts as a mediator between the AI segmentation process and the final diagnosis. The auditing apparatus generates segmentation error heatmaps that highlight potential errors without directly modifying the segmentation process, allowing radiologists to review and verify results efficiently
2Measurement precision
If segmentation error auditing is performed using deep learning models, then segmentation accuracy is improved, but processing time increases
Solution Approach 1:
The patent applies preliminary action by pre-processing the segmented image to generate a pre-processed segmented image that is optimized for error detection. The deep learning model is pre-trained on medical image data to recognize segmentation errors efficiently. This preliminary preparation enables faster and more accurate error detection during the auditing process
Solution Approach 2:
The patent uses local quality by generating segmentation error heatmaps that highlight only the specific regions with potential errors rather than analyzing the entire image uniformly. The heatmap focuses computational resources on areas with high error probability, reducing overall processing time while maintaining high detection accuracy
3Adaptability or versatility
If preprocessing is performed to convert data formats, then image compatibility is improved, but processing complexity increases
Solution Approach 1:
The patent implements universality by designing a preprocessing module that can handle multiple data formats (CT, MRI, PET) and convert them to a unified format suitable for segmentation and auditing. The deep learning model is trained to process various medical image types, making the system versatile and adaptable to different imaging modalities without requiring separate processing pipelines
Data Source
AI summary
Disclosed is a method of auditing of artificial intelligence-based medical image segmentation, including: performing preprocessing to generate a preprocessed segmentation image by receiving an input medical image and an output segmentation image provided from a medical image segmentation device and preprocessing the output segmentation image based on the input medical image; generating a heatmap image to generate a segmentation error heatmap image, which includes a segmentation error region in the preprocessed segmentation image, by inputting the preprocessed segmentation image to a deep learning model trained in advance; calculating an error risk to calculate a segmentation error risk for the segmentation error region based on pixel values of the segmentation error heatmap image; and providing auditing information to provide the auditing information for auditing accuracy of the output segmentation image based on the calculated segmentation error risk to a user.