Biomarker Analysis Apparatus Using Multi-Model Inference for Tumor Site Determination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current liquid biopsy technologies face challenges in accurately determining tumor incidence sites, often leading to false determinations and increased patient load due to repeated detailed examinations, as existing models may incorrectly identify tumor sites with high probabilities.
Innovation Solution
A medical information processing apparatus that utilizes a series of trained models (zeroth, first, second, and third determination models) to infer disease types based on biomarker data, with the second and third models adjusting for potential false determinations by excluding specific disease-type elements or modifying biomarker data to improve accuracy and reduce unnecessary examinations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single trained model is used to determine disease type based on biomarker data, then the determination process is simple and quick, but the accuracy is low leading to false determinations
Solution Approach 1:
The determination system is segmented into multiple specialized models: a zeroth model for initial determination and first/third models for re-determination after defying results. Each model handles specific scenarios, dividing the complex determination task into manageable segments that collectively improve accuracy while maintaining operational simplicity.
Solution Approach 2:
The system changes the parameter of model selection based on the determination flow state. Different models are activated depending on whether it's the initial determination or a re-determination after a defying result, optimizing accuracy for each specific situation rather than using a single static model.
2Reliability
If multiple detailed examinations are performed to verify tumor incidence sites, then the accuracy of diagnosis is improved, but the physical and financial load on patients increases
Solution Approach 1:
The system performs preliminary filtering through the zeroth model to identify high-probability tumor incidence sites before detailed examinations are conducted. By pre-screening and prioritizing likely candidates, the system reduces the number of detailed examinations needed while maintaining high diagnostic reliability.
Solution Approach 2:
The system uses feedback from examination results to trigger re-determination processes. When examination results defy the initial determination, the system automatically initiates a re-determination flow with specialized models, ensuring reliability only when necessary rather than through routine repeated examinations for all patients.
3Adaptability or versatility
If the examination system performs comprehensive analysis of all possible disease types, then the coverage of potential diagnoses is complete, but the time and resources required increase significantly
Solution Approach 1:
The comprehensive analysis is segmented into hierarchical stages: the zeroth model provides broad initial coverage of multiple disease types, then the first and third models focus re-determination on specific defied disease types. This segmentation maintains comprehensive coverage capability while improving efficiency by concentrating detailed analysis only where needed.
Solution Approach 2:
The system performs partial analysis comprehensively at the initial stage through the zeroth model, then applies excessive action (re-determination) only partially to cases with defying results. This selective application of comprehensive analysis maintains overall efficiency while ensuring thoroughness for uncertain cases.
Data Source
AI summary
A medical information processing apparatus according to an embodiment includes a processing circuitry. The processing circuitry acquires an examination value of a biomarker collected from an examination target subject; determines a first disease type based on an inference result of a first trained model, the inference result being obtained by inputting the acquired examination value to the first trained model, the first trained model being configured to infer the first disease type of an affected disease upon inputting of the examination value; receives input of an examination result related to the first disease type of the subject and obtained through examination; and determines, when the examination result of the subject defies the first disease type, a second disease type of the disease with which the subject is potentially affected by using a second trained model in accordance with the defied first disease type.