Pathology Image Diagnosis with Basis Image Evidence
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional diagnosis support systems using learning models for pathology images lack transparency, making it difficult for users to determine the reliability of estimation results, as the basis for these results is not presented.
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 the user for enhanced reliability assessment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a learning model is used to derive diagnosis estimation results from pathology images, then diagnostic productivity is improved, but the reliability of the estimation result cannot be determined because the basis for derivation is not presented
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 trust assessment. The basis image serves as tangible evidence that connects the abstract AI decision to the user's understanding, allowing verification without requiring direct inspection of model parameters or training data.
Solution Approach 2:
The system implements feedback by presenting the basis image to the user, who can then assess whether the estimation result is appropriate based on visual evidence. This creates a closed loop where the user's understanding of the model's reasoning influences their acceptance and potential correction of the diagnosis, improving overall system reliability.
2Device complexity
If only the estimation result is output without presenting the basis for derivation, then device complexity is reduced, but the user cannot sufficiently utilize the estimation result
Solution Approach 1:
The patent extracts only the essential basis information needed for user verification - specifically the basis image that directly supports the estimation result - without requiring extraction or presentation of the entire model architecture, training dataset, or parameter configurations. This selective extraction maintains simplicity while providing sufficient utility.
Data Source
AI summary
An estimation result can be more effectively utilized. An information processing apparatus includes a deriving unit 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 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.