Medical Data Analysis Using Semantic Subspace and Knowledge Graph
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current disease diagnosis methods using medical data analysis require setting up separate models for each disease and rely on detection indexes, which is inefficient and increases medical costs, failing to utilize comprehensive patient information like symptoms effectively.
Innovation Solution
A method and device that analyze medical data by locating a semantic subspace based on physical parameters and utilizing a medical knowledge graph to determine probabilities, integrating implicit and explicit knowledge for disease analysis, thereby reducing the need for multiple models and unnecessary tests.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If separate models are set up for each disease using detection indexes, then disease diagnosis can be performed, but the system complexity increases and medical costs increase
Solution Approach 1:
The patent merges multiple disease-specific models into a unified semantic subspace model. Instead of maintaining separate diagnostic models for each disease, the system represents all diseases within a common semantic framework where diseases are subspaces in a high-dimensional space defined by medical features. This consolidation reduces model complexity while preserving diagnostic capability through unified mathematical operations.
Solution Approach 2:
The patent creates a universal diagnostic model that can handle multiple diseases simultaneously. The semantic subspace framework provides a multi-functional system where a single model structure can diagnose various diseases by projecting patient data into appropriate disease subspaces. This universal approach eliminates the need for disease-specific model implementations.
2Reliability
If separate models are set up for each disease, then specific disease analysis is possible, but the quantity of models required increases
Solution Approach 1:
Multiple disease models are merged into a single semantic subspace model. The system represents diseases as subspaces within a unified framework, allowing one model to perform the function of many separate models. The mathematical formulation h=DX enables a single model to capture patterns for multiple diseases through the matrix D that encodes disease-specific knowledge.
3Measurement precision
If detection indexes are used for diagnosis, then objective measurement is achieved, but comprehensive patient information like symptoms is not utilized effectively
Solution Approach 1:
The patent transforms the diagnostic approach by changing the parameter representation from traditional detection indexes alone to a comprehensive feature space that includes symptoms, lab results, and other patient information. The semantic subspace model processes these diverse parameters uniformly, converting qualitative symptom data into the same mathematical framework as quantitative lab values, thereby utilizing all patient information effectively while maintaining measurement objectivity.
Data Source
AI summary
The present disclosure provides a medical data analysis method and device. The method comprises locating a semantic subspace of the subject in a medical data set by taking the physical parameter as a feature; and analyzing the probability P1 of the subject being in the semantic subspace that the subject belongs to by judging the semantic consistency of the semantic subspace where the physical parameter of the subject exists. In addition, it is also possible to analyze the probability P2 of the subject being in the node that the subject belongs to based on the evidence transference score of the physical parameter of the subject on the medical knowledge graph. P=α×P1+(1−α)×P2 The probability P of the subject being in the semantic subspace or node that the subject belongs to can be determined by P=α×P1+(1−α)×P2, wherein α is a reconciling parameter, 0<α<1. Through these solutions, the analysis accuracy and efficiency can be improved and the cost can be decreased.