Hi everyone,
I’m currently working on transforming radiology data into the OMOP Common Data Model. I have structured data such as patient birth date, sex, procedure occurrence, procedure name, interpreter, and city. Additionally, I have the full radiology report written by the radiologist (interpreter), which is unstructured text.
We are using NLP to extract clinical elements from these reports — such as measurements, conditions, observations, and sometimes devices when they are mentioned. The extracted terms are mapped to standard vocabularies like SNOMED CT and LOINC, and we are considering incorporating RadLex for more radiology-specific coverage.
The extracted concepts are being recorded in the NOTE_NLP table, and then inserted into the appropriate domain tables.
However, I have a specific question regarding devices: when a device (e.g., a pacemaker or stent) is mentioned by the radiologist in the report, but no explicit implantation date is provided, we are currently storing that information as an OBSERVATION record, since we lack structured data or DICOM metadata to place it into the DEVICE_EXPOSURE table.
Does this approach align with OMOP recommendations?
Is it appropriate to store the mention of a device (without implantation date) as an OBSERVATION, or is there a more suitable alternative in the model?
This approach is inspired by the article:
Representing and Utilizing Clinical Textual Data for Real World Studies: An OHDSI Approach
We also reviewed:
Development of Medical Imaging Data Standardization for Imaging-Based Observational Research: OMOP Common Data Model Extension
While it proposes CDM extensions for imaging metadata, our current project does not use DICOM files, so this does not yet apply to us.
One of our main goals is to use OMOP to associate structured clinical labels with each radiology report, based on the radiologist’s interpretation, in a standardized and queryable format.
Any feedback or guidance on this approach — especially regarding handling device mentions — would be greatly appreciated.
Thanks so much!