Entity Detection for Electronic Medical Records
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current electronic medical record systems require clinicians to use structured data formats, which can be cumbersome and time-consuming, especially for free-form notes, and lack efficient methods for extracting discrete medical facts from verbal dictations.
Innovation Solution
A method and apparatus for enhancing electronic medical records by automatically extracting discrete medical facts from free-form narratives using natural language understanding techniques, involving token matching with ontologies, hierarchical concept identification, and statistical modeling to determine entity types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If clinicians use structured data formats for electronic medical records, then data management and accessibility are improved, but the time and effort required to document patient encounters increases
Solution Approach 1:
The system enables self-service by automatically extracting discrete medical facts from free-form clinician notes and populating structured electronic medical record fields without requiring manual data entry. The entity detection system processes the narrative text and autonomously identifies and categorizes medical entities, allowing the documentation system to serve itself rather than requiring clinician intervention for data structuring.
Solution Approach 2:
The patent replaces the mechanical process of manual data entry and structured form filling with an automated natural language processing system. The entity detection apparatus substitutes the manual mechanical action of clinicians typing structured data with an automated computational process that analyzes free-form text and extracts relevant medical information, thereby eliminating the time-consuming manual documentation process.
2Ease of operation
If free-form narratives are used for clinical notes, then ease of documentation is improved, but automatic extraction of discrete medical facts becomes more difficult
Solution Approach 1:
The patent introduces an intermediary entity detection system that bridges free-form narratives and structured medical records. This intermediary apparatus processes the unstructured text through tokenization, entity recognition, and classification stages, acting as a mediator that translates clinician-friendly free-form notes into machine-structured data without requiring clinicians to change their documentation style.
Solution Approach 2:
The entity detection system applies segmentation by breaking down free-form clinical notes into discrete tokens and identifying individual medical entities within the narrative. The system segments the continuous text stream into extractable factual elements such as patient conditions, medications, procedures, and outcomes, making the extraction process systematic rather than attempting to parse the entire narrative as a single unit.
3Measurement precision
If hierarchical ontology concepts are included as features, then entity detection accuracy is improved, but processing complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-processing the input text through tokenization and matching tokens against a hierarchical ontology before performing entity detection. The system prepares the data structure in advance by organizing tokens and their corresponding ontology concepts, which simplifies the subsequent entity recognition process and improves accuracy by having the hierarchical relationships established beforehand.
Data Source
AI summary
Techniques for entity detection include matching a token from at least a portion of a text string with a matching concept in an ontology. A first concept may be identified as being hierarchically related to the matching concept within the ontology, and a second concept may be identified as being hierarchically related to the first concept within the ontology. The first and second concepts may be included in a set of features of the token. Based at least in part on the set of features of the token, a measure related to a likelihood that the at least a portion of the text string corresponds to a particular entity type may be determined.