Medical Conversation Concept Linking for Structured Data Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automated extraction of structured data from conversational text in medical records faces challenges such as variety in vocabularies and sentence structures, medically irrelevant information, noise, and linking distant pieces of information due to unreliable sentence delimitation by automatic speech recognition.
Innovation Solution
A system and method that utilizes a machine learning model to identify and combine transcript concepts based on medical concepts and labels, determining valid concept combinations to generate structured data, leveraging contextual information to link relevant information across the conversation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional automated extraction methods are used, then processing speed is maintained, but accuracy of extracting structured data deteriorates due to variety in vocabularies and sentence structures
Solution Approach 1:
The system transforms the extraction task from rule-based pattern matching to a machine learning classification task, changing the fundamental parameter of how extraction is performed. The ML model learns to handle vocabulary and sentence structure variations through training on diverse data, achieving both accuracy and adaptability simultaneously
Solution Approach 2:
The patent introduces an intermediary processing layer that combines multiple transcript concepts before final extraction. This intermediary step allows the system to handle complex relationships between concepts and resolve ambiguities in vocabulary and sentence structures through contextual analysis
2Loss of information
If comprehensive concept combination is performed, then completeness of structured data improves, but computational resources required increase
Solution Approach 1:
The system segments the extraction process into distinct stages: concept identification, concept combination, and final extraction. By dividing the task, the system can apply targeted processing at each stage, ensuring completeness while managing computational resources efficiently through staged processing
Solution Approach 2:
The system performs preliminary concept identification and labeling before combining concepts. This preliminary action prepares the data in advance, allowing the combination stage to focus only on integrating pre-processed concepts, thereby reducing overall computational requirements while maintaining completeness
3Measurement precision
If noisy transcript data is processed directly, then processing efficiency is maintained, but extraction accuracy deteriorates due to fillers, garbles, and restating
Solution Approach 1:
The system extracts and isolates relevant medical concepts from noisy transcript data, separating signal from noise. By focusing only on identified concepts rather than processing entire sentences, the system achieves high accuracy while maintaining efficiency through selective processing
Solution Approach 2:
The system applies partial processing to transcript data - not every word or phrase is analyzed in depth, but only those identified as potential medical concepts through initial scanning. This partial action approach maintains efficiency while achieving accuracy through focused analysis of relevant portions
4Loss of information
If distant pieces of information are linked, then completeness of structured data improves, but complexity of processing increases due to unreliable sentence delimitation
Solution Approach 1:
The machine learning model serves multiple functions simultaneously: it identifies concepts, determines their relationships, and handles sentence boundary issues all within a single unified framework. This multi-functionality reduces overall system complexity while achieving complete information linking
Solution Approach 2:
The system introduces an intermediary concept combination stage that acts as a mediator between raw transcript concepts and final structured data. This intermediary layer handles the complexity of linking distant information and resolving delimitation issues, isolating complexity from the rest of the system
Data Source
AI summary
A system for determining structured data includes a processor configured to receive a plurality of medical concepts and a corresponding plurality of labels. The processor is further configured to receive a transcript of a conversation between a physician and a patient. The processor is further configured to determine a plurality of transcript concepts in the transcript based on the plurality of medical concepts. The processor is further configured to assign each transcript concept with the label associated with the corresponding medical concept. The processor is further configured to determine a plurality of concept combinations by combining the plurality of transcript concepts. The processor is further configured to determine, via a machine learning model, a combination label for each concept combination. The combination label is a valid label or invalid label. The processor is further configured to generate an output structured data based on the concept combinations having the valid label.