DICOM Enrichment with EMR Data via Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning approaches for medical imaging struggle to incorporate additional medical information about subjects, such as age, sex, and medical history, due to difficulties in obtaining and correlating this information with medical imaging data, while adhering to privacy laws like HIPAA.
Innovation Solution
A software and/or hardware facility that enriches DICOM objects with additional medical information by using a DICOM enrichment template to specify fields to be added, retrieving values from Electronic Medical Record (EMR) databases, and optionally anonymizing identifying information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If additional medical information is incorporated into DICOM objects to improve machine learning training quality, then prediction accuracy is improved, but data processing complexity and privacy protection requirements increase
Solution Approach 1:
The patent segments the data enrichment process by introducing a separate enrichment template file that specifies which EMR fields to retrieve. This template acts as an independent configuration layer, separating the complex data integration logic from the core DICOM processing pipeline, thereby managing complexity while enabling comprehensive data incorporation.
Solution Approach 2:
The patent introduces an intermediary enrichment process that acts as a mediator between EMR databases and DICOM objects. This intermediary layer retrieves and correlates additional medical information (age, sex, medical history) before integrating it into DICOM objects, simplifying the overall data integration complexity while improving prediction accuracy.
2Measurement precision
If identifying information is retained in DICOM objects for accurate data correlation, then data correlation accuracy is improved, but subject privacy protection deteriorates
Solution Approach 1:
The patent extracts identifying information from DICOM objects into a separate subject identifier field, removing sensitive data from the main DICOM structure. This extraction allows accurate correlation through the identifier while protecting subject privacy by isolating sensitive information in a controlled manner.
Solution Approach 2:
The patent creates a copy of the subject identifier for correlation purposes while maintaining the original DICOM object structure. This copying approach enables accurate data matching without exposing sensitive identifying information in the enriched DICOM objects, thereby protecting subject privacy.
3Loss of information
If live EMR connections are maintained for real-time data retrieval, then data freshness is improved, but system operational efficiency deteriorates
Solution Approach 1:
The patent performs preliminary data retrieval by incorporating additional medical information into DICOM objects at the time of imaging or during pre-processing. This preliminary action ensures data freshness is captured when relevant, while avoiding the need for continuous live EMR connections during subsequent machine learning operations, thereby improving operational efficiency.
Solution Approach 2:
The patent creates enriched copies of DICOM objects with embedded medical information, eliminating the need for repeated live EMR connections. These copied and enriched objects can be used offline for training and inference, maintaining data freshness from the time of enrichment while dramatically improving system operational efficiency.
Data Source
AI summary
A facility for performing an enrichment process is disclosed. The facility accesses a starting file emitted by a medical imaging device, which contains (1) medical imaging information captured from a subject, and information identifying the subject. The facility uses the information identifying the subject in the starting file to extract medical information from an electronic medical record corresponding to the subject. The facility then creates an enriched file by making a copy of the starting file and adding the extracted medical information to the copy.