Pathology Image Diagnosis Support Using Basis Image Evidence
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional diagnosis support systems using learning models for pathology images do not provide a basis for deriving estimation results, making it difficult for users to determine the reliability of the estimation results.
Innovation Solution
An information processing apparatus that derives an estimation result using a trained model and identifies a basis image from training data, presenting both the estimation result and the basis image to users for improved reliability assessment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If only an estimation result is output from the learning model, then the diagnosis support system is simple and easy to operate, but the user cannot determine the reliability of the estimation result
Solution Approach 1:
The patent introduces a basis image as an intermediary element that mediates between the learning model's estimation result and the user's reliability assessment. The basis image serves as evidence that connects the abstract estimation result to concrete observable features in the pathology image, allowing users to verify the reliability without requiring complex technical knowledge of the learning model's internal workings.
2Reliability
If a basis image is provided to show the basis for derivation, then the reliability of estimation result can be determined, but the device complexity increases
Solution Approach 1:
The patent extracts only the essential basis image from the complex learning model processing, presenting it to the user without requiring them to understand the underlying complex algorithms. This extraction approach provides the necessary reliability information while avoiding the complexity of explaining the entire learning model architecture and processing steps.
3Ease of operation
If detailed basis information is provided, then the user can effectively utilize the estimation result, but the information processing becomes more complex
Solution Approach 1:
The patent applies local quality by providing basis information specifically at the locations in the pathology image where the learning model made its estimation. Instead of providing generic or global information, the basis image highlights the specific local regions and features that contributed to the estimation result, making the information both useful and efficiently processed.
Data Source
AI summary
An estimation result can be more effectively utilized. An information processing apparatus includes a deriving unit (100) that derives an estimation result of diagnosis for a second pathology image using a trained model on which learning has performed using training data including a plurality of first pathology images, and an identifying unit (100) that identifies a basis image that serves as a basis for derivation of the estimation result by the trained model from the plurality of first pathology images.