Medical Image Processing Device for Explainable AI Diagnosis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computer-aided diagnosis (CAD) systems face challenges in achieving both high accuracy and explainability, particularly in medical image processing, where AI-based systems struggle to interpret features used for decision-making and human clinicians find it difficult to understand AI-driven diagnostics due to the 'black box' nature of AI algorithms.
Innovation Solution
A medical image processing device and endoscope system that acquires and processes multiple types of medical images under varying conditions, generating diagnosis and reference information by combining feature amounts using a layered model structure, allowing for the conversion of feature amounts into interpretable reference information, thereby enhancing diagnostic accuracy and explainability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If AI-based CAD systems are used to achieve high accuracy in medical image processing, then diagnostic accuracy is improved, but explainability of the determination deteriorates due to the black box nature of AI algorithms
Solution Approach 1:
The patent introduces an intermediary component that extracts and explains the features used by the AI model for determination. This mediator layer translates the black box AI decisions into interpretable feature information that clinicians can understand, thereby maintaining high diagnostic accuracy while improving explainability.
Solution Approach 2:
The patent segments the AI determination process into distinct feature components that can be individually analyzed and explained. By dividing the complex AI decision-making into separable feature contributions, the system maintains accuracy while enabling interpretability of each feature's role in the diagnosis.
2Measurement precision
If complex feature amounts are used to enhance accuracy in AI-based CAD, then estimation accuracy is improved, but the complexity of the system increases making it more difficult to explain
Solution Approach 1:
The patent employs an intermediary mechanism that simplifies the representation of complex feature amounts while preserving their diagnostic value. This intermediary layer transforms complex multi-dimensional features into more manageable and explainable forms without sacrificing estimation accuracy.
Solution Approach 2:
The patent applies parameter changes to transform complex feature representations into simplified forms that maintain accuracy. By adjusting the parameters and dimensions of feature representation, the system reduces complexity while preserving the essential information needed for high-accuracy estimation.
3Measurement precision
If multiple types of medical images under varying conditions are processed, then diagnostic accuracy is improved, but the complexity of image processing increases
Solution Approach 1:
The patent implements a universal image processing framework that handles multiple types of medical images under varying conditions through a single integrated system. This multi-functional approach processes diverse image inputs (different modalities, conditions, and formats) through unified processing pipelines, improving diagnostic accuracy while managing complexity through standardization.
Data Source
AI summary
The medical image processing device acquires the plurality of types of medical images obtained by imaging a subject under imaging conditions different from each other. In a case where a first medical image that is one type of the plurality of types of medical images is input, diagnosis information regarding a diagnosis of the subject shown in the first medical image is generated. Diagnostic reference information is generated using assigned reference information that is the reference information assigned to a second medical image which is included in the medical image and which has a type different from the first medical image.