Home

Welcome to the OHDSI Medical Imaging Workgroup (ImageWG) wiki!

How to use this wiki

This wiki is intended for individuals integrating imaging with clinical EHR data. That is, individuals who are implementing the OHDSI OMOP Medical Imaging Extensions to the Common Data Model (MI CDM).

Navigation of pages is along the right hand side.

Have a question? Post questions or issues to forums.ohdsi.org with the tag #imaging and we will get right back to you!

This wiki consists of:

1. Goals of incorporating imaging into Observational Research (below)
2. Book of OHDSI Medical Imaging Chapter - the high level introduction (link coming soon)
3. Charter of Imaging Evidence Network members - how we work together to produce evidence
4. The DICOM Standard - Background and Complexities
5. Getting started: Decision tree, ETL methods, and ETL conventions - technical implementation - not sure where to start implementation after Book of OHDSI? START HERE. The diagram explains the ETL process. The paragraphs below are numbered to match the steps in the diagram and contain more details.
6. There are several wiki pages with detailed information. All begin with "Convention: ...". These links are contained within the Getting Started page above.
7. Examples of use (use cases) - cohort discovery with Atlas - coming soon.
8. Reference papers and materials

---

1. Goals of incorporating imaging into Observational Research

Why is imaging research important to combine with EHR research

Medical Imaging is an essential part of medicine for detecting disease and for assessing the response to treatment of patients. In US on there are over 1 billion medical imaging studies performed on a yearly bases. Imaging data is generated by leveraging how energy interacts with the human body to look for anatomical and/or physiological information about the body's function. This can leverage visible light data for imaging in dermatology and viewing anatomical pathology slides under a microscope. This also can include x-rays used in Computed Tomography, ultrasound waves in Ultrasound, radiofrequency waves in Magnetic Resonance, to radioisotypes in Nuclear Medicine. Imaging data can provide highly granular information in studying disease and treatments.

Medical Imaging Research has accelerated greatly since 2012 and the invention of the deep convolutional neural networks for their ability be able to learn how to segment and classify medical images. In 2012, DCNN were introduced in winning the ImageNet challenge used to categorize visible light images. In 2024 over 10,000 papers were published in medical imaging (cite pubmed).

To date, the traditional way to bring radiological concepts into the OMOP data model has been through text extraction techniques from radiological notes. Radiology notes acquired clinically often lack the structured morphological features that are of benefit to research. The work described here is an Imaging extension to the OMOP model to enable source imaging data to be indexable into the OMOP model along with features derived from segmentation algorithms.

Have a question? Post questions or issues to forums.ohdsi.org with the tag #imaging and we will get right back to you!

Vision for federated learning with imaging data we can accomplish - needs an author

Scalability - needs an author - does this really go here?? if not, then where?