Concept Normalization Using External Evidence for Clinical Diagnosis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current intelligent clinical decision support tools face challenges in inferring accurate diagnoses due to the lack of large annotated training corpora and limitations in processing unstructured clinical data, particularly when patients describe symptoms in ambiguous or unconventional ways, leading to failures in capturing the complexities of clinical scenarios.
Innovation Solution
A method and system that normalize input data by extracting concepts from free text, querying external resources, and revising these concepts using a reinforcement learning agent to generate refined concepts, which are then processed to infer diagnoses, leveraging deep learning neural networks to optimize reward functions for accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If supervised classification models using neural networks are used for diagnostic inferencing, then the model can process structured clinical data, but it fails to capture ambiguities and uncertainties in unstructured clinical scenarios and requires large labeled data sets
Solution Approach 1:
The patent introduces an intermediary concept normalization module that acts as a bridge between unstructured patient descriptions and structured medical concepts. This module queries external knowledge bases (UMLS, SNOMED CT) to find standardized concepts that match patient descriptions, enabling the system to handle ambiguous language while maintaining diagnostic accuracy through structured processing.
Solution Approach 2:
The system performs preliminary concept normalization and disambiguation before the main diagnostic classification process. By pre-processing unstructured text to extract and normalize medical concepts using external knowledge bases, the system prepares the data in advance, reducing the burden on the classification model and improving its ability to handle ambiguous descriptions.
2Reliability
If medical personnel manually analyze clinical cases to infer diagnoses, then accurate diagnoses can be obtained, but significant time and cognitive burden are required
Solution Approach 1:
The system enables self-service by automatically performing concept extraction, normalization, and diagnostic inference without requiring manual medical expertise. The automated pipeline processes clinical cases through multiple stages including NLP, knowledge base querying, and machine learning-based classification, delivering accurate diagnoses rapidly and reducing dependency on manual analysis.
Solution Approach 2:
The system incorporates feedback mechanisms where diagnostic results and performance metrics are continuously used to improve the model. By monitoring diagnostic accuracy and adjusting the system based on outcomes, the patent ensures maintaining high reliability while automating the process to reduce time loss.
3Productivity
If concept extraction from free text is performed without external resource validation, then processing speed is maintained, but accuracy of concept identification decreases
Solution Approach 1:
The system performs preliminary querying of external knowledge bases to validate and refine extracted concepts before final diagnostic processing. By pre-checking concepts against UMLS and SNOMED CT, the system ensures accurate concept identification while maintaining efficiency through optimized query strategies and caching mechanisms.
Solution Approach 2:
External knowledge bases serve as intermediaries between raw text extraction and final concept identification. The system uses these knowledge bases to verify and disambiguate extracted concepts, improving measurement precision while maintaining productivity through efficient intermediary processing.
Data Source
AI summary
Various embodiments described herein relate to a method, system, and non-transitory machine-readable medium including one or more of the following: extracting a first concept from input data presented for processing by a downstream function; identifying external data from an external resource based on the first concept; extracting a second concept from the external data; revising the first concept based on the second concept to produce a revised concept, wherein revising includes: applying a machine learning agent to determine whether to keep the first concept or adopt the second concept, and adopting the second concept in place of the first concept for use as the revised concept based on a decision by the machine learning agent to adopt the second concept; and further processing the revised concept according to the downstream function to generate an output.