Imaging Module

Parthenon's Imaging module integrates DICOM medical imaging data into the OMOP CDM research platform. It provides DICOM file import, study/series/instance hierarchy management, a Cornerstone3D-powered web viewer with clinical tools, radiology NLP feature extraction, population-level imaging analytics, and imaging cohort criteria for the cohort builder.

Module Architecture

DICOM Source  -->  Import & Index  -->  Study/Series/Instance DB  -->  WADO-RS Endpoint
     |                                          |                          |
Local files                              Feature Extraction         Cornerstone3D Viewer
DICOMweb server                                 |
                                         Imaging Features
                                                |
                                    Cohort Criteria  /  Population Analytics

DICOM Data Model

Parthenon follows the standard DICOM information hierarchy:

Studies

A study represents a single imaging examination (e.g., a chest CT scan ordered for a patient):

Field	Description
Study Instance UID	Globally unique DICOM identifier
Accession Number	Hospital-assigned order number
Modality	Imaging modality (CT, MR, US, XA, etc.)
Body Part Examined	Anatomical region (CHEST, HEAD, ABDOMEN, etc.)
Study Description	Free-text description from the DICOM header
Study Date	Date the examination was performed
Person ID	Linked OMOP person (if matched)
Series Count	Number of series in the study
Image Count	Total number of images across all series
Status	indexed / processed / error

Series

A series is a set of images acquired with the same parameters (e.g., axial CT slices at 5mm thickness):

Field	Description
Series Instance UID	Unique series identifier
Series Description	Protocol name or series label
Modality	Series-level modality (may differ from study)
Series Number	Ordering index within the study
Image Count	Number of instances in this series
Slice Thickness	Slice thickness in mm
Manufacturer	Scanner manufacturer
Model	Scanner model name

Instances

Individual DICOM image files with SOP Instance UIDs, instance numbers, and slice locations.

Supported Modalities

Code	Modality	Description
CT	Computed Tomography	Cross-sectional X-ray imaging
MR	Magnetic Resonance	MRI scans
PT	Positron Emission Tomography	PET scans
US	Ultrasound	Sonographic imaging
CR	Computed Radiography	Digital X-rays
DX	Digital Radiography	Direct digital X-rays
MG	Mammography	Breast imaging
XA	X-Ray Angiography	Vascular imaging
NM	Nuclear Medicine	Scintigraphy, SPECT
RF	Radiofluoroscopy	Real-time X-ray

Importing DICOM Data

Local File Import

Import DICOM files from a directory on the Parthenon server:

1. Navigate to Imaging from the main navigation.
2. Click Import Local in the toolbar.
3. Configure:
   • Data Source -- target OMOP source
   • Directory -- relative path to DICOM files on the server (default: dicom_samples)
4. Click Import.

The DicomFileService scans the directory recursively, parses DICOM headers from each .dcm file using a PHP-native DICOM reader, and upserts study/series/instance records in the database.

Import results show:

• Studies imported (new or updated)
• Series imported
• Instances imported

DICOMweb Indexing

For integration with PACS or Orthanc servers, Parthenon can pull study metadata via the DICOMweb (QIDO-RS) protocol:

1. Click Index from DICOMweb in the toolbar.
2. Configure:
   • Data Source -- target OMOP source
   • Limit -- maximum number of studies to index (default: 100, max: 1000)
   • Modality filter -- optionally restrict to a specific modality (e.g., CT)
3. Click Index.

The DicomwebService queries the configured DICOMweb endpoint, retrieves study-level metadata, and upserts records. Series can be indexed separately per study.

DICOMweb configuration

The DICOMweb endpoint URL is configured per data source in the source settings. Parthenon supports Orthanc's built-in DICOMweb plugin and any standard WADO-RS/QIDO-RS compliant server.

Series Indexing

After a study is indexed, retrieve detailed series information:

1. Open a study in the study browser.
2. Click Index Series.
3. Parthenon queries the DICOMweb endpoint for series-level metadata and populates the series table.

Study Browser

The study browser provides a searchable, filterable list of all indexed imaging studies:

Filters

Filter	Description
Data Source	Studies from a specific OMOP source
Modality	Filter by imaging modality (CT, MR, etc.)
Person ID	Studies for a specific patient
Date Range	Studies within a time period
Body Part	Studies of a specific anatomical region
Search	Free-text search across study description and accession number

Study Detail Page

Click any study to open the detail page showing:

• Full study metadata (all DICOM header fields)
• Series list with thumbnails and metadata
• Instance list (sortable by instance number and slice location)
• Imaging features extracted from NLP or AI (see below)
• Link to the Cornerstone3D viewer

Cornerstone3D DICOM Viewer

Parthenon includes an integrated web-based DICOM viewer built on Cornerstone3D, providing clinical-grade image viewing directly in the browser:

Viewer Features

Tool	Description	Shortcut
Window/Level	Adjust brightness and contrast (Hounsfield units for CT)	Left-click drag
Zoom	Magnify or reduce image size	Scroll wheel
Pan	Move the image within the viewport	Middle-click drag
Scroll	Navigate through slices in a series	Scroll wheel (over image)
Length Measurement	Measure distance between two points in mm	L key
Angle Measurement	Measure angle between three points	A key
Reset	Reset all viewport adjustments	R key

WADO-RS Endpoints

The viewer retrieves DICOM images via Parthenon's built-in WADO-URI endpoint:

GET /api/v1/imaging/wado/{sopInstanceUid}

This endpoint streams the raw DICOM file with Content-Type: application/dicom and appropriate caching headers. The Cornerstone3D dicomImageLoader uses this endpoint to load images on demand.

Viewer performance

For large studies (500+ images), the viewer loads images progressively as you scroll through slices. Initial viewport rendering begins immediately with the first available image. Enable browser caching to avoid re-downloading images on repeat views.

Imaging Features

The imaging module extracts structured features from imaging studies for use in research queries:

Feature Types

Type	Source	Description
NLP Finding	Radiology NLP service	Findings extracted from radiology report text
AI Classification	AI model inference	Classification results from imaging AI models
Radiomic	Quantitative analysis	Radiomic features (texture, shape, intensity)
Manual	User annotation	Manually entered imaging observations

NLP Feature Extraction

For studies linked to OMOP patients, Parthenon can extract imaging findings from associated radiology notes:

1. Open a study detail page.
2. Click Extract NLP Features.
3. The RadiologyNlpService queries the OMOP note table for radiology reports matching the patient.
4. NLP extracts structured findings (e.g., "pulmonary nodule", "pleural effusion") with confidence scores.
5. Extracted features are stored in the imaging_features table.

Feature Table

Each feature record contains:

Field	Description
Feature Name	Descriptive name (e.g., "Pulmonary Nodule")
Feature Type	nlp_finding / ai_classification / radiomic / manual
Algorithm	Name of the extraction algorithm or model
Value (numeric)	Quantitative value (e.g., nodule size in mm)
Value (string)	Qualitative value (e.g., "present", "absent")
Unit	Unit of measurement (mm, HU, etc.)
Body Site	Anatomical location of the finding
Confidence	Algorithm confidence score (0-1)
OMOP Concept ID	Mapped OMOP concept for standardized querying

Imaging Cohort Criteria

The Imaging module extends the cohort builder with imaging-specific criteria:

Criteria Type	Description	Example
modality	Patients with studies of a specific modality	Patients with at least one CT scan
anatomy	Patients with studies of a specific body part	Patients with chest imaging
quantitative	Patients with imaging features meeting numeric thresholds	Patients with nodule size > 8mm
ai_classification	Patients with specific AI classification results	Patients classified as "high risk" by lung screening AI
dose	Patients based on radiation dose parameters	Patients with cumulative CT dose > threshold

Creating Imaging Criteria

1. Navigate to Imaging > Criteria (or use the Cohort Builder's imaging criteria panel).
2. Click New Criterion.
3. Select a criteria type and configure parameters.
4. Optionally mark as Shared for organization-wide use.

Imaging criteria combine with standard clinical criteria in the cohort builder for multi-modal research.

Population Imaging Analytics

The Population Analytics view provides aggregate imaging statistics across a data source:

By Modality

Bar chart showing study counts and unique patient counts per modality:

Modality	Studies	Unique Patients
CT	--	--
MR	--	--
US	--	--
...	--	--

By Body Part

Top 15 body parts by study count, revealing the most common examination types in your database.

Top Features

Most frequently extracted imaging features across all studies, grouped by feature type. This identifies the most common radiological findings in your population.

OMOP CDM Integration

Imaging studies can be linked to the OMOP CDM through:

• Person matching -- linking person_id from patient demographics
• Image occurrence -- mapping to the OMOP image_occurrence extension table (if available)
• Visit linking -- associating studies with OMOP visits via accession number or date matching
• Measurement mapping -- imaging features written to the measurement table with appropriate concepts

API Reference

Key imaging API endpoints:

Endpoint	Method	Description
/api/v1/imaging/stats	GET	Module statistics
/api/v1/imaging/studies	GET	List/filter studies
/api/v1/imaging/studies/{id}	GET	Study detail with series
/api/v1/imaging/studies/{id}/instances	GET	Instance list for viewer
/api/v1/imaging/studies/{id}/index-series	POST	Index series from DICOMweb
/api/v1/imaging/studies/{id}/extract-nlp	POST	Extract NLP features
/api/v1/imaging/studies/index-from-dicomweb	POST	Index studies from DICOMweb
/api/v1/imaging/features	GET	List imaging features
/api/v1/imaging/criteria	GET, POST	List/create cohort criteria
/api/v1/imaging/analytics/population	GET	Population-level analytics
/api/v1/imaging/wado/{sopUid}	GET	WADO-URI DICOM file stream
/api/v1/imaging/import-local/trigger	POST	Trigger local DICOM import

DICOM data privacy

DICOM files may contain patient-identifying information embedded in the DICOM headers (patient name, date of birth, medical record number). Parthenon does not automatically de-identify DICOM data. Ensure your DICOM files are appropriately de-identified before importing into research environments, or configure your import pipeline to strip PHI-containing DICOM tags.