Representation

The medical imaging extension proposal represents imaging characteristics through two new tables—Image_occurrence and Image_features.


Schema

Image_occurrence


Table Description

The Image_occurrence table describes imaging events and provides data lineage to the imaging study stored in DICOM format.


User Guide

Each row in Image_occurrence represents a collection of images acquired on an imaging modality using a contiguous imaging technique. This is referred to as a DICOM series. Each DICOM series can be an independent modality and acquisition technique grouped within a DICOM study (e.g., PET/CT scan).


cdmTableName cdmFieldName isRequired cdmDatatype userGuidance
image_occurrence image_occurrence_id Yes integer The unique key is given to an imaging study record (often referred to as the accession number or imaging order number)
image_occurrence person_id Yes integer The person_id of the Person for whom the procedure is recorded. This can be a system-generated code or adopted from original source.
image_occurrence procedure_occurrence_id Yes integer The unique key is given to a procedure record for a person. Link to the Procedure_occurrence table.
image_occurrence visit_occurrence_id No integer The unique key is given to the visit record for a person. Link to the Visit_occurrence table.
image_occurrence anatomic_site_concept_id No integer Anatomical location of the imaging procedure by the medical acquisition device (gross anatomy). It maps the ANATOMIC_SITE_SOURCE_VALUE to a Standard Concept in the Spec Anatomic Site domain. This should be coded at the lowest level of granularity.
image_occurrence wadors_uri No varchar (max) A Web Access to DICOM Objects via Restful Web Services Uniform Resource Identifier on study level.
image_occurrence local_path Yes varchar (max) “Universal Naming Convention (UNC) path to the folder containing the image object file access via a storage block access protocol. (e.g., \Server)”
image_occurrence image_occurrence_date Yes date The date the imaging procedure occurred.
image_occurrence image_study_UID Yes varchar (250) DICOM Study UID
image_occurrence image_series_UID Yes varchar (250) DICOM Series UID
image_occurrence modality Yes varchar (250) “DICOM-defined value (e.g., US, CT, MR, PT, DR, CR, NM)”


Image_feature


Table Description

Imaging features are comprised of segmentation algorithm results executed on the images, image acquisition parameters, and structured radiology reports. The Image_feature table describes the characteristics of the images and their provenance.


User Guide

Each row will contain a uniquely identified feature with links to the source imaging as well as the clinical domain table the feature is located in. The image_feature_type_concept_id will describe the method used to create that feature (e.g., machine learning algorithm, DICOM structured report). Image acquisition parameters will also be stored in the clinical domain tables and linked to the source images through the image feature table. This allows the cardinality to include multiple acquisition parameters as well as the benefit of being able to query them with existing tooling in the OHDSI platform.


cdmTableName cdmFieldName isRequired cdmDatatype userGuidance
image_feature image_feature_id Yes integer The unique key is given to an imaging feature.
image_feature person_id Yes integer The person_id of the Person table for whom the the procedure is recorded. This can be a system-generated code or adopted from original source.
image_feature image_occurrence_id Yes integer The unique key of the Image_occurrence table.
image_feature table_concept_id Yes integer “The concept_id of the domain table that feature is stored in Measurement, Observation, etc. This concept should be used with the table_row_id.”
image_feature table_row_id Yes integer The row_id of the domain table that feature is stored.
image_feature image_feature_concept_id Yes integer Concept_id of standard vocabulary—often a LOINC or RadLex of image features
image_feature image_feature_type_concept_id Yes integer “This field can be used to determine the provenance of the imaging features (e.g., DICOM SR, algorithms used on images)”
image_feature image_finding_concept_id No integer “RadLex or other terms of the groupings of image feature (e.g., nodule)”
image_feature image_finding_id No integer Integer for linking related image features. It should not be interpreted as an order of clinical relevance.
image_feature anatomic_site_concept_id No integer This is the site on the body where the feature was found. It maps the ANATOMIC_SITE_SOURCE_VALUE to a Standard Concept in the Spec Anatomic Site domain.
image_feature alg_system No varchar(max) “URI of the algorithm that extracted features, including version information”
image_feature alg_datetime No datetime The date and time of the algorithm processing.