Clinical Note Generation Using Historical Note Overlap Synthesis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Clinical notes generated solely based on doctor-patient conversation (DoPaCo) often lack important clinical information present in historical notes stored in electronic health records (EHRs).
Innovation Solution
A system and method that trains a machine learning model to generate clinical notes by identifying and synthesizing overlapping data from historical notes, incorporating missing information to create augmented clinical notes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If clinical notes are generated solely based on doctor-patient conversation (DoPaCo), then the generation process is simple and fast, but the clinical notes lack important clinical information present in historical notes
Solution Approach 1:
The patent merges DoPaCo-based note generation with historical note analysis by integrating a machine learning model that processes both current conversation data and historical clinical notes to generate comprehensive clinical notes that include both new and relevant historical information
Solution Approach 2:
The system performs preliminary analysis of historical notes before generating the final clinical note. The machine learning model pre-processes historical clinical notes to identify and extract relevant information that should be incorporated into the new clinical note, ensuring no important information is lost
2Loss of information
If historical notes are integrated into clinical note generation, then information completeness improves, but processing time and computational resources increase
Solution Approach 1:
The machine learning model extracts only the relevant information from historical notes that is necessary for the current clinical context, rather than processing or copying all historical data. This selective extraction reduces processing time while maintaining information completeness
Solution Approach 2:
The system applies partial action by focusing processing efforts only on the portions of historical notes that are relevant to the current clinical situation, avoiding unnecessary processing of unrelated historical information and thus reducing time loss
3Loss of substance
If overlapping data is removed from synthesized input notes, then data redundancy decreases, but the complexity of data processing increases
Solution Approach 1:
The machine learning model creates a synthesized version of the input clinical note that copies only the non-redundant information, effectively removing overlapping data while maintaining the essential clinical information. This copying approach simplifies the output while reducing redundancy
Data Source
Figure 1
Figure 2
Figure 3A~3C
AI summary
A system includes a non-transitory storage having stored thereon instructions that when executed by a processor cause the processor to train a machine learning model to generate clinical notes by repeatedly: receiving a plurality of clinical notes associated with a patient; selecting a final clinical note; selecting a predecessor clinical note chronologically preceding the final clinical note; determining overlapping data between the final clinical note and the predecessor clinical note; performing an action on the identified overlapping data to generate a synthesized input clinical note; providing the synthesized input clinical note and the predecessor clinical note to the machine learning model; receiving a predicted final clinical note from the machine learning model; and updating the machine learning model based on the predicted final clinical note and the final clinical note.