Diagnosis Object Extraction Using Dynamic Entity Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for extracting diagnosis objects from medical documents group entities based on medical concepts without considering the diagnoses made, leading to inconsistent results for the same body part entities across different medical documents.
Innovation Solution
A method that extracts body part entities and non-body-part entities along with their relations, computes relevance scores using historical medical documents, and clusters non-body-part entities to form diagnosis-dependent groups, allowing for accurate diagnosis object extraction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If body part entities are grouped according to medical concepts, then the extraction process is simple and consistent, but the diagnosis objects become inconsistent across different medical documents with the same body parts
Solution Approach 1:
The patent changes the grouping parameters from static medical concepts to dynamic diagnosis-dependent parameters. By computing relevance scores based on diagnoses and using these scores as new grouping parameters, the system achieves both operational simplicity through automated scoring and reliability through diagnosis-specific grouping that adapts to different medical documents.
Solution Approach 2:
The patent introduces dynamic grouping where diagnosis objects are not fixed by medical concepts alone but vary dynamically based on the specific diagnoses present in each medical document. The relevance scores and clustering results change dynamically according to the diagnostic context, enabling the same body parts to be grouped differently for different diagnoses.
2Ease of manufacture
If diagnosis objects are extracted independently of diagnoses, then the extraction method is straightforward, but it fails to capture diagnosis-specific entity groupings
Solution Approach 1:
The patent implements feedback by computing relevance scores that incorporate diagnostic information from the medical document. The extraction process uses diagnosis-related terms and their relationships with body parts to generate relevance scores, which then feed back into the grouping process. This feedback mechanism ensures that diagnosis-specific information is preserved while maintaining an automated extraction workflow.
Solution Approach 2:
The patent performs preliminary extraction of diagnosis-related terms and their relationships with body parts before final grouping. By pre-computing relevance scores based on diagnoses and storing these scores, the system prepares diagnosis-specific information in advance, which is then used during the clustering phase to form accurate diagnosis objects without losing diagnostic context.
3Stability of the object's composition
If the same body part entities are grouped together based on medical concepts, then grouping is consistent, but different diagnoses involving the same body parts are not distinguished
Solution Approach 1:
The patent applies local quality by making grouping characteristics diagnosis-specific rather than uniformly applying medical concept grouping across all documents. Each diagnosis creates a local grouping context where body parts are clustered based on their relevance to that specific diagnosis, allowing the same body parts to have different groupings in different diagnostic contexts while maintaining consistency within each diagnostic framework.
Data Source
AI summary
This invention provides a method for extracting a diagnosis object from a medical document comprises: extracting, from an input medical document, body part entities and at least one type of non-body-part entities and the relations between the body part entities and the non-body-part entities; obtaining, for each pair of all possible pairs of the non-body-part entities, a relevance score between two non-body-part entities within one pair, wherein the relevance score is obtained by using the relations between the two non-body-part entities within one pair and one or more body part entities in a plurality of historical medical documents; clustering the non-body-part entities in the input medical document into one or more clusters based on the relevance scores of the all possible pairs; and outputting one or more body part entities related to one or more non-body-part entities clustered in each of the clusters as one diagnosis object.