Medical Diagnostic Apparatus Bayesian Inference Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing medical diagnostic supporting apparatuses face reduced accuracy in inference when multiple pieces of information are not entered, as they often present less likely information as noteworthy, due to selecting based solely on high influence values without considering the likelihood of existence in medical images.
Innovation Solution
A medical diagnostic supporting apparatus that includes a medical image input unit, medical information acquisition unit, image feature amount acquisition unit, presented not-entered information candidate selection unit, inference unit, presented not-entered information selection unit, and presentation unit, which uses Bayesian networks to select and present not-entered information that is likely to exist and has a high influence on the inference result.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If not-entered information is selected solely based on high influence values, then the inference accuracy is improved, but irrelevant information that is less likely to exist in medical images is presented
Solution Approach 1:
The patent changes the selection parameters from solely using influence values to using a combined evaluation of influence values and likelihood of existence. This dual-parameter approach filters out irrelevant information while maintaining high inference accuracy by selecting not-entered information that is both influential and likely to exist in medical images.
Solution Approach 2:
The patent introduces an intermediary evaluation mechanism that assesses both the influence value and the likelihood of existence of not-entered information. This intermediary layer acts as a filter between the inference engine and the presentation layer, ensuring that only relevant and useful information is presented to doctors.
2Loss of information
If all not-entered information is presented to doctors, then complete diagnostic information is provided, but the information overload reduces efficiency in identifying critical information
Solution Approach 1:
The patent applies local quality by differentiating the presentation of not-entered information based on their evaluated relevance. Instead of uniform presentation, the system prioritizes displaying information with high influence values and high likelihood of existence, while de-prioritizing or filtering out irrelevant information. This selective presentation maintains information completeness while improving diagnostic efficiency.
Solution Approach 2:
The patent segments not-entered information into different categories based on their influence values and likelihood of existence. By dividing the information into prioritized groups, the system enables doctors to focus on critical information first while still having access to complete diagnostic data if needed, thus balancing information completeness with diagnostic efficiency.
Data Source
AI summary
A medical diagnostic supporting apparatus inputs a medical image to be a target of a medical diagnosis, acquires one or more pieces of medical information relating to the medical image as entered information, and acquires an image feature amount from the medical image. The medical diagnostic supporting apparatus selects a plurality of pieces of not-entered information associated with the acquired image feature amount from not-entered information that is medical information other than the entered information as presented not-entered information candidates that are candidates for presentation, and selects presented not-entered information from the presented not-entered information candidates based on a plurality of inference results acquired using the entered information and each of the presented not-entered information candidates. The medical diagnostic supporting apparatus presents the selected presented not-entered information to a doctor.