Medical Image Reasoning With Proof-Based Decision Support
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing medical image analysis technologies provide results without grounds for image analysis, making it difficult for doctors to trust the analysis at the clinical stage.
Innovation Solution
A medical decision supporting apparatus and method that includes a visual and linguistic neural learner, cause miner, logical reasoner, and proof inference unit to generate answers with proof in response to user queries on medical images, utilizing visual and text embeddings, causal concepts, and background knowledge graphs to provide interpretable and reliable medical reports.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a medical image analysis model is used to provide analysis results, then the analysis speed and efficiency are improved, but the lack of grounds for analysis makes it difficult for doctors to trust the results
Solution Approach 1:
The patent introduces an intermediary explanation generation module that acts as a mediator between the image analysis model and the doctor. This module generates natural language explanations that bridge the gap by translating the model's internal reasoning into human-understandable grounds, thereby maintaining fast analysis while building trust through interpretability
Solution Approach 2:
The system implements feedback by generating explanations that reflect the model's analysis process back to the user. This feedback loop allows doctors to understand the grounds for analysis results, verify the reasoning, and thus trust the system more while maintaining efficient automated analysis
2Reliability
If detailed explanation and proof generation is added to the analysis system, then the trust and interpretability are improved, but the system complexity increases
Solution Approach 1:
The patent segments the complex explanation generation task into distinct functional modules: a reasoning process extraction module that identifies key analytical steps, and a natural language generation module that formulates explanations. This segmentation manages complexity by dividing the overall system into specialized, manageable components
Solution Approach 2:
The system performs preliminary action by pre-extracting reasoning processes and key analytical steps before generating final explanations. This preliminary structuring of information simplifies the subsequent explanation generation task and reduces the computational complexity of the overall system
Data Source
AI summary
A medical decision supporting apparatus according to an embodiment may include a visual and linguistic neural learner, which receives a medical image and extracts visual embedding, a cause miner, which extracts a predetermined index based on the visual embedding and pretrained embedding, and outputs a predetermined cause for the medical image based on a hash map including at least one index-concept pair and the extracted predetermined index, a logical reasoner, which generates a medical report based on the visual embedding, a background knowledge graph, and the predetermined cause, and a proof inference unit, which generates an answer including a proof for a user's question about the medical image based on the predetermined cause and the medical report.