Coronary Image Evaluation With Explainable Rule-Based AI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing medical imaging evaluation systems, particularly for coronary artery disease, lack transparency and interactivity, with artificial intelligence algorithms providing unexplainable results and requiring significant user time for manual interaction.
Innovation Solution
A computer-implemented method and system that segments and labels anatomical structures using evaluation algorithms, applies inference rules to generate comprehensible evaluation information, and allows user interaction for modification, integrating deep learning and traditional rule-based systems to provide traceable and understandable assessments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If artificial intelligence evaluation algorithms are used to automate medical image assessment, then productivity increases, but the comprehensibility and explainability of results deteriorates
Solution Approach 1:
The patent introduces an intermediary layer between the deep learning algorithm and the final evaluation result. This intermediary consists of rule-based systems that translate the black-box AI outputs into comprehensible medical concepts and criteria, allowing users to understand how conclusions were reached while maintaining automated evaluation efficiency
Solution Approach 2:
The evaluation system is segmented into multiple independent components: deep learning algorithms for feature extraction, rule-based systems for logical reasoning, and separate modules for different evaluation criteria. This segmentation allows each component to specialize while maintaining overall system transparency and explainability
2Loss of information
If manual evaluation by physicians is performed, then comprehensibility of results is maintained, but productivity and efficiency deteriorate
Solution Approach 1:
The patent merges the strengths of manual evaluation (comprehensibility, medical expertise) with automated systems (efficiency, consistency). Physician knowledge is encoded into rule-based systems that work alongside AI algorithms, creating a hybrid system that maintains human-level comprehensibility while achieving automated productivity
Solution Approach 2:
The system performs preliminary automated evaluation using deep learning and rule-based systems before final physician review. This preliminary action handles routine assessments efficiently, allowing physicians to focus only on complex or ambiguous cases, thereby improving overall productivity while maintaining comprehensibility
3Productivity
If deep learning algorithms are used to mimic complex disease assessment schemes, then productivity increases, but device complexity deteriorates
Solution Approach 1:
Rule-based systems serve as intermediaries that simplify the interface between complex deep learning algorithms and users. These rules translate complex AI outputs into straightforward medical criteria and decision pathways, reducing perceived system complexity while maintaining automated evaluation capabilities
Solution Approach 2:
The system incorporates feedback mechanisms where evaluation results and intermediate findings are continuously fed back through rule-based systems that can adjust and refine outputs. This feedback loop simplifies complex decision-making processes by making them iterative and transparent, reducing overall system complexity
Data Source
AI summary
At least one example embodiment provides a computer-implemented method for evaluating at least one image data set of an imaging region of a patient, wherein at least one evaluation information describing at least one medical condition in an anatomical structure of the imaging region is determined.