13 CDM map study, series, image

13.1 Which DICOM tags?

Note

Q: Please recommend the appropriate DICOM tag that match each of these:

Slice-level or Image-level
Series-level
Study-level

Here are the key DICOM tags for each hierarchical level:

13.1.1 Study-level

StudyInstanceUID (0020,000D) - PRIMARY IDENTIFIER
├── StudyDate (0008,0020)
├── StudyTime (0008,0030)
├── StudyDescription (0008,1030)
├── AccessionNumber (0008,0050)
└── PatientID (0010,0020)

13.1.2 Series-level

SeriesInstanceUID (0020,000E) - PRIMARY IDENTIFIER
├── SeriesNumber (0020,0011)
├── SeriesDescription (0008,103E)
├── Modality (0008,0060)
├── BodyPartExamined (0018,0015)
└── SeriesDate (0008,0021)

13.1.3 Slice/Image-level

SOPInstanceUID (0008,0018) - PRIMARY IDENTIFIER
├── InstanceNumber (0020,0013)
├── ImagePositionPatient (0020,0032)
├── SliceLocation (0020,1041)
└── SliceThickness (0018,0050)

13.1.4 DICOM Hierarchy Visualization

Patient
└── Study (StudyInstanceUID)
    └── Series (SeriesInstanceUID)
        ├── Image/Slice 1 (SOPInstanceUID)
        ├── Image/Slice 2 (SOPInstanceUID)
        └── Image/Slice N (SOPInstanceUID)

13.1.5 For Your OMOP CDM Mapping:

Study-level mapping:

StudyInstanceUID → Primary key for imaging studies
AccessionNumber → Links to RIS/hospital systems

Series-level mapping:

SeriesInstanceUID → Groups related images
Modality → Maps to OMOP procedure concepts

Image-level mapping:

SOPInstanceUID → Individual measurements/observations
InstanceNumber → Ordering/sequencing

13.2 Design

Note

Q: Please help me design how these identifiers should flow into OMOP CDM tables:

StudyInstanceUID
SeriesInstanceUID
SOPInstanceUID

Here’s a comprehensive design for mapping DICOM identifiers into OMOP CDM tables:

13.2.1 Core OMOP Tables Mapping

13.2.1.1 1. PROCEDURE_OCCURRENCE (Study-level)

-- One row per imaging study
procedure_occurrence_id    -- OMOP auto-generated
person_id                 -- Patient identifier  
procedure_concept_id      -- Maps to imaging procedure (e.g., "CT Head")
procedure_date           -- From StudyDate
procedure_source_value   -- StudyInstanceUID
modifier_source_value    -- AccessionNumber

13.2.1.2 2. MEASUREMENT (Series & Image-level)

-- Multiple rows per study (one per series or per image)
measurement_id           -- OMOP auto-generated
person_id               -- Patient identifier
measurement_concept_id  -- Maps to measurement type
measurement_date        -- From SeriesDate or ContentDate
value_as_number         -- Quantitative measurements (if any)
measurement_source_value -- SeriesInstanceUID or SOPInstanceUID

13.2.2 Custom Extension Strategy

13.2.2.1 3. DICOM_HIERARCHY (Custom table)

CREATE TABLE dicom_hierarchy (
    hierarchy_id             BIGINT PRIMARY KEY,
    study_instance_uid       VARCHAR(200) NOT NULL,
    series_instance_uid      VARCHAR(200),
    sop_instance_uid         VARCHAR(200),
    procedure_occurrence_id  BIGINT,  -- Links to OMOP
    measurement_id           BIGINT,  -- Links to OMOP
    hierarchy_level          VARCHAR(10), -- 'STUDY'/'SERIES'/'IMAGE'
    file_path               VARCHAR(500),
    created_date            DATE
);

13.2.3 Mapping Flow Diagram

DICOM Study (CT Brain)
├── StudyInstanceUID: 1.2.3.4.5
│   └── PROCEDURE_OCCURRENCE
│       ├── procedure_source_value: "1.2.3.4.5"
│       └── procedure_concept_id: [CT Head concept]
│
├── SeriesInstanceUID: 1.2.3.4.5.100
│   └── MEASUREMENT (Series-level)
│       ├── measurement_source_value: "1.2.3.4.5.100"
│       └── measurement_concept_id: [Imaging series concept]
│
└── SOPInstanceUID: 1.2.3.4.5.100.001
    └── MEASUREMENT (Image-level)
        ├── measurement_source_value: "1.2.3.4.5.100.001"
        └── measurement_concept_id: [CT slice concept]

13.2.4 Python Implementation Example

def map_dicom_to_omop(dicom_metadata):
    """Map DICOM hierarchy to OMOP CDM tables"""
    
    # 1. Create study-level procedure
    procedure_data = {
        'person_id': dicom_metadata['patient_id'],
        'procedure_concept_id': get_procedure_concept('CT_HEAD'),
        'procedure_date': dicom_metadata['study_date'],
        'procedure_source_value': dicom_metadata['study_instance_uid']
    }
    procedure_id = insert_procedure_occurrence(procedure_data)
    
    # 2. Create series-level measurements
    for series in dicom_metadata['series']:
        series_measurement = {
            'person_id': dicom_metadata['patient_id'],
            'measurement_concept_id': get_measurement_concept('IMAGING_SERIES'),
            'measurement_date': series['series_date'],
            'measurement_source_value': series['series_instance_uid']
        }
        series_measurement_id = insert_measurement(series_measurement)
        
        # 3. Create image-level measurements (if needed)
        for image in series['images']:
            image_measurement = {
                'person_id': dicom_metadata['patient_id'],
                'measurement_concept_id': get_measurement_concept('CT_SLICE'),
                'measurement_date': image['content_date'],
                'measurement_source_value': image['sop_instance_uid'],
                'value_as_number': image.get('slice_thickness')
            }
            image_measurement_id = insert_measurement(image_measurement)
            
            # 4. Maintain hierarchy in custom table
            insert_dicom_hierarchy({
                'study_instance_uid': dicom_metadata['study_instance_uid'],
                'series_instance_uid': series['series_instance_uid'],
                'sop_instance_uid': image['sop_instance_uid'],
                'procedure_occurrence_id': procedure_id,
                'measurement_id': image_measurement_id,
                'hierarchy_level': 'IMAGE'
            })

Key Design Decisions

Study → PROCEDURE_OCCURRENCE (represents the imaging exam)
Series → MEASUREMENT (represents acquisition parameters)
Images → MEASUREMENT (individual slices/observations)
Custom table maintains DICOM hierarchy and file references