Medical ontology alignment method based on Point Set Registration
An ontology and medical technology, applied in the field of medical knowledge graphs, can solve the problems of low fusion accuracy of medical knowledge graphs, heterogeneity of medical ontology data, poor data matching accuracy, etc., and achieve simple and easy algorithms, rigorous interpretability, The effect of improving accuracy
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
specific Embodiment approach 1
[0058] Specific implementation mode 1: In this implementation mode, the specific process of a medical ontology alignment method based on Point Set Registration is as follows:
[0059] Point SetRegistration is point cloud registration;
[0060] Step 1. Embedding each concept in the two sets of medical ontology datasets to obtain a vector representation of the concept;
[0061] Step 2. For the ontology alignment problem, a mixed Gaussian model is established based on Step 1;
[0062] Step 3, use the EM algorithm to solve the mixed Gaussian model obtained in step 2, and obtain the transformation relationship T between two sets of medical ontology datasets (medical ontology dataset FMA and medical ontology dataset NCI) θ (y m );
[0063] Step 4, the vectors of the two groups of medical ontology obtained in step 1 (using the TransE method obtained in step 1 to embed each concept in the ontology data set FMA and NCI, the vector representation of the concept obtained;) represents ...
specific Embodiment approach 2
[0068] Embodiment 2: The difference between this embodiment and Embodiment 1 is that in the first step, each concept in the two sets of medical ontology data sets is embedded to obtain a vector representation of the concept; the specific process is:
[0069] Using the TransE method, using the triplet relationship contained in the medical ontology dataset FMA as input, embedding each concept (such as: Monoblast) in the medical ontology dataset FMA, and obtaining the vector representation of the concept X N×D ;
[0070] Using the TransE method, using the triplet relationship contained in the medical ontology dataset NCI as input, embedding each concept (such as: Chondroblast) in the medical ontology dataset NCI, and obtaining the vector representation of the concept Y M×D ;
[0071] x N×D and Y M×D The expression is:
[0072] x N×D =(x 1 ,...,x n ,...,x N ) T ;
[0073] Y M×D =(y 1 ,...,y m ,...,y M ) T ;
[0074] where x 1 is the concept vector obtained by embe...
specific Embodiment approach 3
[0076] Embodiment 3: The difference between this embodiment and Embodiment 1 or 2 is that in Step 2, a mixed Gaussian model is established based on Step 1 for ontology alignment; the specific process is:
[0077] Will Y M×D y in 1 ,...,y M The vector is regarded as the centroid of the mixture Gaussian model, X N×D x in 1 ,...,x N Vectors are viewed as points generated by a mixture of Gaussian models;
[0078] Thus, the original ontology alignment problem is transformed into a problem of solving the parameters of the mixed Gaussian model;
[0079] Establish the probability density function of the mixed Gaussian model, the expression is as follows:
[0080]
[0081] In the formula, p(m) is the prior probability of the mth model, p(x n |m) is x given the mth Gaussian model n The conditional probability distribution of x n is the concept vector obtained by embedding the nth concept in the medical ontology dataset FMA, and M is the size of the ontology dataset NCI (y 1...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More