OMOP Image Occurrence

Summary:

In the OMOP Common Data Model (CDM), the “Image_occurrence” table provides a structured way to represent medical imaging events, bridging the gap between imaging and observational research by integrating image-based measurements into the CDM.

The OMOP image_occurrence table captures records of medical imaging events at the series level, representing events where one or more related images are acquired as part of a clinical procedure, and it also provides a path to de-identified DICOM images.

Each record includes metadata such as the modality (e.g., CT, MRI) and the anatomic site of the image. The table maintains foreign key links to the procedure_occurrence and visit_occurrence tables, enabling contextual association with the related clinical procedure and patient encounter.

At Stanford, this table has been extended to include a link to the note table, allowing integration of imaging data with relevant clinical documentation. We expect to have larger number of images in the coming months.

As of November 2025, this table included images for 195,836 patients with 38,082,448 series. Series are defined as individual imaging acquisitions within a study.

📊 Data Volume

  • Patient Count: 195,836
  • Series Count: 38,082,448
  • Study UID: 3,926,143

🧬 Data Components

  • 163 Modality types
  • Anatomic sites

Image Occurrence Metrics

This visualization summarizes distribution of Modality categories and Anatomic sites by series and patient counts for all cancer types and for thoracic patients. Note: We expect the number of patients with imaging data to increase per release.

Modality Types

Top modality types for series above 5k occurrences are listed as follow.

Note: The following modality source values without image pixel were excluded from the analysis:

SR: Structured Report REG: Registration KO: Key Object Selection PR: Presentation State

Patient Population

Note: This visualization shows modality types with more than 1,000 patients.

Modality Descriptions and Frequency by Series and Patients

Please note that the modality descriptions are listed here: Modality Descriptions (DICOM Library).

Anatomic Sites

Top anatomic sites for series above 100k occurrences are listed as follow.

Patient Population

Note: This visualization shows anatomic sites with more than 100k series.

Anatomic Site Descriptions and Frequency by Series and Patients

Thoracic cancer patients are identified based on their primary site descriptions in the Neural Frame diagnoses data, which include diagnoses of lung, bronchus, or thymus cancers.

As of November 2025, from 14,208 patients with thoracic cancer, 13,607 had imaging data.

Modality Types

Top modality types for series above 5k occurrences are listed as follow.

Anatomic Sites

Top anatomic sites for series above 5k occurrences are listed as follow.

Source Code:

The source codes for this page can be found here, and the sql queries that support the metrics can be found here. The queries regarding thoracic cancer can be found here.