extractSegmentationInfo

R2026b

Extract segmentation information of DICOM-SEG file

Since R2026b

Syntax

segInfo = extractSegmentationInfo(medVol)

Description

segInfo = extractSegmentationInfo(medVol) extracts metadata related to segmentation labels from a DICOM-SEG file, imported as a medical volume, medVol.

Examples

Read and Visualize DICOM-SEG Data

Download and unzip an MRI prostrate volume [1].

zipFile = matlab.internal.examples.downloadSupportFile("medical","ProstatexMRIData.zip");
filepath = fileparts(zipFile);
unzip(zipFile,filepath)

Create a medical volume object for the MRI prostate volume.

filepathMRI = fullfile(filepath,"ProstatexMRIData","5.000000-t2tsetra-75680");
mriVol = medicalVolume(filepathMRI)

mriVol = 
  medicalVolume with properties:

                  Voxels: [384×384×19 int16]
          VolumeGeometry: [1×1 medicalref3d]
            SpatialUnits: "mm"
             Orientation: "transverse"
            VoxelSpacing: [0.5000 0.5000 4.5000]
            NormalVector: [1.4081e-17 0.2419 0.9703]
        NumCoronalSlices: 384
       NumSagittalSlices: 384
     NumTransverseSlices: 19
            PlaneMapping: ["sagittal"    "coronal"    "transverse"]
    DataDimensionMeaning: ["left"    "posterior"    "superior"]
                Modality: "MR"
           WindowCenters: [19×1 double]
            WindowWidths: [19×1 double]

Unzip the prostrate segmentation volume attached to this example as a supporting file.

unzip("ProstateXSegData.zip",pwd)

Create a medical volume object for the prostate segmentation volume.

filepathSEG = fullfile(pwd,"ProstateX-0004","10-18-2011-NA-MR prostaat kanker detectie WDSmc MCAPRODETW-45493", ...
    "300.000000-Segmentation-3.549","1-1.dcm");
segVol = medicalVolume(filepathSEG)

segVol = 
  medicalVolume with properties:

                  Voxels: [384×384×171 logical]
          VolumeGeometry: [1×1 medicalref3d]
            SpatialUnits: "mm"
             Orientation: "transverse"
            VoxelSpacing: [0.5000 0.5000 0.5000]
            NormalVector: [2.6606e-17 0.2419 0.9703]
        NumCoronalSlices: 384
       NumSagittalSlices: 384
     NumTransverseSlices: 171
            PlaneMapping: ["sagittal"    "coronal"    "transverse"]
    DataDimensionMeaning: ["left"    "posterior"    "superior"]
                Modality: "SEG"
           WindowCenters: []
            WindowWidths: []

Extract segmentation information from the segmentation volume.

segInfo = extractSegmentationInfo(segVol)

segInfo = struct with fields:
         SegmentNumber: 1
          SegmentLabel: "Prostate"
    SegmentDescription: "Prostate"
           ColorCIELab: [43803 26565 37722]
         AlgorithmType: "MANUAL"
     SegmentationColor: [128 174 128]
      SegmentationType: 'BINARY'

Observe that the voxel spacing for the MRI and segmentation volumes is different. Resample the segmentation volume to align with the MRI volume.

segVol = resample(segVol,mriVol.VolumeGeometry)

segVol = 
  medicalVolume with properties:

                  Voxels: [384×384×19 logical]
          VolumeGeometry: [1×1 medicalref3d]
            SpatialUnits: "mm"
             Orientation: "transverse"
            VoxelSpacing: [0.5000 0.5000 4.5000]
            NormalVector: [1.4081e-17 0.2419 0.9703]
        NumCoronalSlices: 384
       NumSagittalSlices: 384
     NumTransverseSlices: 19
            PlaneMapping: ["sagittal"    "coronal"    "transverse"]
    DataDimensionMeaning: ["left"    "posterior"    "superior"]
                Modality: "SEG"
           WindowCenters: []
            WindowWidths: []

The segmentation data contains only one label. Create a label map for the segmentation.

labelMap = segVol.Voxels;

Visualize the label map overlaid on the MRI volume.

tf = oneSliceIntrinsicToWorldMapping(mriVol.VolumeGeometry,1);
vol = volshow(mriVol.Voxels, ...
    Transformation=tf, ...
    DisplayRangeMode="data-range", ...
    OverlayData=labelMap, ...
    OverlayDisplayRangeMode="data-range", ...
    OverlayColormap=[1 0 0], ...
    OverlayAlpha=0.5);

References

[1] Meyer, A., Schindele, D., von Reibnitz, D., Rak, M., Schostak, M., & Hansen, C. (2020). PROSTATEx Zone Segmentations [Data set]. The Cancer Imaging Archive. https://doi.org/10.7937/TCIA.NBB4-4655.

Input Arguments

medVol — Medical volume imported from DICOM-SEG file

medicalVolume object

Medical volume imported from a DICOM-SEG file, specified as a medicalVolume object.

Output Arguments

segInfo — Segmentation label information

structure

Segmentation label information, returned as a structure with these fields.

Field	Description
SegmentNumber	Segment numbers, returned as an S-by-1 vector containing the numeric identifier of each of the S segments.
SegmentLabel	Segment labels, returned as an S-by-1 vector of strings containing the label of each of the S segments.
SegmentDescription (if available)	Segment descriptions, returned as an S-by-1 vector of strings containing a detailed description of each of the S segments.
AlgorithmType (if available)	Type of algorithm, returned as an S-by-1 vector of strings containing the algorithm type of each of the S segments as "MANUAL", "AUTOMATIC", or "SEMIAUTOMATIC".
AlgorithmName (if available)	Algorithm name, returned as an S-by-1 vector of strings containing the algorithm name of each of the S segments.
SegmentationType	Type of segmentation, returned as "BINARY" or "FRACTIONAL".
SegmentationFractionalType	Subtype of fractional segmentation, returned as "PROBABILITY" or "OCCUPANCY". This field exists only if the segmentation type is "FRACTIONAL".
MaximumFractionalValue	Maximum permissible fractional value in raw pixel data, returned as a numeric scalar. This field exists only if the segmentation type is "FRACTIONAL".
ColorCIELab (if available)	LAB color triplets for visualization, returned as an S-by-3 array with each of the S rows representing the LAB color triplet of the corresponding segment.
SegmentationColor (if available)	RGB color triplets for visualization, returned as an S-by-3 matrix in which each row represents the RGB color triplet of the corresponding segment.
SegmentsOverlap (if available)	Indicates whether any pixel in the image is or might be in more than one segment, returned as "YES", "NO", or "UNDEFINED".

Version History

Introduced in R2026b