Segmentations

Segmentation labelmap file format (.seg.nrrd)

Segmentation is stored in a standard NRRD image file with custom fields to store additional metadata.

The slicerio Python package can read .seg.nrrd files and import segments into numpy arrays and dictionary objects. The package can be used in any Python environment (installed using pip install slicerio) - not just in 3D Slicer.

Image data

If segments do not overlap then the file has 3 spatial dimensions. Even if a single-slice image is segmented, the spatial dimension must be still 3 because that is required for specification origin, spacing, and axis directions in 3D space. If segments overlap then the file has one list dimension and 3 spatial dimensions. Each 3D volume in the list is referred to as a layer.

If the image contains a single voxel then it has a special meaning: it means that no image data is available and the image geometry (origin, spacing, axis directions, extents) will be ignored. This special case is necessary because it is not possible to create a nrrd image file without any voxel data. Therefore, a segmentation without segments or with only empty segments is saved to file (e.g., to be used as a template for new segmentations) then it will be saved with an image data containing a single voxel.

Metadata

Additional metadata is stored in custom data fields (starting with Segmentation_ or SegmentN_ prefixes), which provide hints on how the segments should be displayed or what they contain.

Common custom fields

  • Segmentation_ContainedRepresentationNames: names of segmentation representations (separated by | character) that should be displayed. Common representation names include Binary labelmap, Closed surface, Planar contours, Fractional labelmap.

  • Segmentation_ConversionParameters: parameters for algorithms that compute requested representations. Each parameter definition is separated by the & character. A parameter definition consists of parameter name, value, and description text, separated by | character.

  • Segmentation_SourceRepresentation: defines what representation is stored in the file. It is most commonly Binary labelmap, but there are other representations, too, which are stored in an image volume - such as Fractional labelmap.

  • Segmentation_ReferenceImageExtentOffset: This field allows storing an image with arbitrary extent in a NRRD file. NRRD file only stores size of the volume. This field stores voxel coordinates (separated by spaces), therefore the extent can be computed as [ReferenceImageExtentOffsetI, ReferenceImageExtentOffsetI+sizeI-1, ReferenceImageExtentOffsetJ, ReferenceImageExtentOffsetJ+sizeJ-1, ReferenceImageExtentOffsetK, ReferenceImageExtentOffsetk+sizeK-1].

Segment custom fields

These fields specify properties for each segment. N refers to the segment index, which is an integer between 0 and NumberOfSegments - 1.

  • SegmentN_ID: identifier of the segment, it is unique within the segmentation. Can be used for unanmiguously refer to a segment even if the segment’s display name is changed.

  • SegmentN_Name: display name of the segment (the name that is shown to users).

  • SegmentN_NameAutoGenerated: if value is 0 then it means that SegmentN_Name was chosen manually by the user, if value is 1 then it means that the segment’s name is generated automatically (for example, determined from terminology).

  • SegmentN_Color: recommended segment display color, defined with red, green, blue values (between 0.0 - 1.0) separated by space character.

  • SegmentN_ColorAutoGenerated: if value is 0 then it means that SegmentN_Color was chosen manually by the user, if value is 1 then it means that the segment’s color is generated automatically (for example, color defined based on terminology).

  • SegmentN_Extent: 6 space-separated values [minI, maxI, minJ, maxJ, minZ, maxZ] defining extent of non-empty region within the segment.

  • SegmentN_Tags: List of key-value pairs for storing additional information about the segment. Key and value is separated by the : character, pairs are separated from each other by the | character.

  • SegmentN_Layer: Index of the 3D volume that contains the segment. Its value is between 0 and DimensionAlongTheListImageAxis - 1. For segmentations in which the segments do not overlap, the segmentation can be represented as a single 3D volume with 1 layer. Upon saving, segments are typically collapsed to as few layers as possible.

  • SegmentN_LabelValue: The scalar value used to represent the segment within its own layer. Segments in separate layers can have the same label value.

TerminologyEntry tag

A frequently used key is TerminologyEntry, which defines what the segment contains using DICOM compliant terminology. Value stores 7 parts: terminology context name, category, type, type modifier, anatomic context name, anatomic region, and anatomic region modifier. Parts are separated from each other by ~ character. Five of these parts - category, type, type modifier, anatomic region, and anatomic region modifier are defined by a triplet of (coding scheme designator, code value, and code meaning). Components of the triplet are separated by ^ character.

Example:

A mass in the right side of the adrenal gland would be encoded with the following the tag:

TerminologyEntry:
Segmentation category and type - 3D Slicer General Anatomy list
~SCT^49755003^Morphologically Altered Structure
~SCT^4147007^Mass
~^^
~Anatomic codes - DICOM master list
~SCT^23451007^Adrenal gland
~SCT^24028007^Right"

Interpretation:

  • terminology context name: Segmentation category and type - 3D Slicer General Anatomy list

  • category: codingScheme=SCT (Snomed Clinical Terms), codeValue=49755003, codeMeaning=Morphologically Altered Structure

  • type: codingScheme=SCT, codeValue=4147007, codeMeaning=Mass

  • type modifier: not specified

  • anatomic context name: Anatomic codes - DICOM master list

  • anatomic region: codingScheme=SCT, codeValue=23451007, codeMeaning=Adrenal gland

  • anatomic region modifier: codingScheme=SCT, codeValue=24028007, codeMeaning=Right

Segmentation surface file format (.seg.vtm)

Segmentation is stored in a standard VTK multiblock data set file format. Multiblock data set consists of an index (.vtm) file that stores the file path of each block. Each block file is a VTK polydata (.vtp) file, typically stored in a subdirectory.

Metadata

Metadata fields are stored in the polydata files (.vtp) as arrays in FieldData:

  • Segmentation_SourceRepresentation: type: String, definition: see in “Common custom fields” section above

  • Segmentation_ConversionParameters: type: String, definition: see in “Common custom fields” section above

  • Segmentation_ContainedRepresentationNames: type: String, definition: see in “Common custom fields” section above

  • Segment_ID: type = String, definition: see in “Segment custom fields” section above

  • Segment_Name: type = String, definition: see in “Segment custom fields” section above

  • Segment_NameAutoGenerated: type = String, definition: see in “Segment custom fields” section above

  • Segment_Color: type = Float64, number of components = 3, definition: see in “Segment custom fields” section above

  • Segment_ColorAutoGenerated: type = String, definition: see in “Segment custom fields” section above

  • Segment_Tags: type = String, definition: see in “Segment custom fields” section above