Medical Image Lesion Analysis With Explainable Determination Basis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI systems, such as those described in Japanese Unexamined Patent Publication No. 2022-146822, lack user-friendly methods for radiologists to understand the analysis results of medical images, particularly in diagnosing lesion candidates, due to the complexity and opacity of AI processing.
Innovation Solution
An image processing apparatus and method that includes a hardware processor to analyze medical images, detect lesion candidate areas, select these areas, and output analysis information including the determination basis for the selected areas, enhancing user understanding and convenience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If AI processing is used to analyze medical images, then analysis accuracy is improved, but user understanding of the analysis process deteriorates
Solution Approach 1:
The patent introduces an intermediary visualization layer that translates AI processing results into visually interpretable formats. The system displays not only the final analysis results but also intermediate processing steps and determination bases in a graphical user interface, allowing radiologists to understand the AI's reasoning process without compromising analysis accuracy
Solution Approach 2:
The patent segments the AI analysis process into multiple interpretable components. Instead of presenting a single black-box result, the system divides the analysis into detectable lesion candidate areas, determination bases, and visualization results, making each component separately examinable and explainable to users
2Loss of information
If detailed analysis information is provided, then user understanding is improved, but system complexity increases
Solution Approach 1:
The patent applies local quality by providing detailed analysis information only where needed. The system automatically identifies and highlights specific lesion candidate areas and their determination bases in the medical image, rather than overwhelming the user with complete analysis data. This targeted approach reduces perceived system complexity while maintaining information accessibility
3Productivity
If AI analysis results are displayed directly, then productivity is improved, but ease of operation deteriorates
Solution Approach 1:
The patent implements feedback mechanisms that allow radiologists to interact with and verify AI analysis results. The system provides feedback in the form of visualization results and determination base information, enabling users to confirm or challenge AI findings. This interactive feedback loop maintains high diagnostic efficiency while improving ease of operation through user control
Data Source
AI summary
Disclosed is an image processing apparatus including a hardware processor that: performs first analysis of analyzing a medical image to detect a lesion candidate area; selects the lesion candidate area in the medical image; and outputs analysis information including a determination basis of analysis by the first analysis for the selected lesion candidate area.