Cancer Imaging Phenomics Toolkit (CaPTk)  1.9.0
Miscellaneous: Perfusion PCA Parameter Extractor

This application extracts the principal components from DSC-MRI scans as mentioned in [1].

REQUIREMENTS: Directory containing the following images in a sub-folder:

  1. A single DSC-MRI image (filename must contain the text 'perf' or 'PERF' or 'DSC' to be correctly detected e.g. AAAA_perf.nii.gz )
  2. Its corresponding mask for which the measurements need to be extracted (filename must contain the text 'label' or 'segmentation') to be correctly detected e.g. AAA_label.nii.gz)
  3. An example directory structure for the input data is as follows:
      - RootFolder (this is the one you select for input)
        - Subject_AAAA
          - AAAA_perf.nii.gz
          - AAAA_label.nii.gz


  1. Launch the application from "Applications" -> "Perfusion PCA".
  2. Specify the input directory containing a DSC-MRI and the mask (Label:1) defining the voxels to extract the measurements.
  3. Specify the output directory to save results.
  4. Select Extract new parameters to extract parameters or Apply extracted parameters to apply the extracted parameters.
  5. Specify the number of PCA images to produce [default value = 0] (optional) or
  6. Specify the variance threshold (optional)
  7. Specify the 'Extracted parameters directory' (if in Apply extracted parameters mode)
  8. Press the "Confirm" button.
  9. The principal components will be extracted/applied and the results will be saved at the specified location.


  1. In extract new parameters mode
    • If none of the optional parameters are specified, default value of 0 is used for the number of PCA images to produce.
  2. In apply extracted parameters mode
    • In this mode one of the optional parameters is required.

This application is also available as with a stand-alone CLI for data analysts to build pipelines around, using the following example commands:

  • The following command will train a new model based on the samples in the inputDir using 5 principal components:
    ${CaPTk_InstallDir}/bin/PerfusionPCA.exe -t 0 -i C:/properly/formatted/inputDir -o C:/outputDir -n 5
  • The following command will train a new model based on the samples in the inputDir for a variance threshold of 99:
    ${CaPTk_InstallDir}/bin/PerfusionPCA.exe -t 0 -i C:/properly/formatted/inputDir -o C:/outputDir -vt 99
  • The following command will perform inference using an existing model for inputs in C:/input based on C:/modelDir using 5 principal components:
    ${CaPTk_InstallDir}/bin/PerfusionPCA.exe -t 1 -i C:/properly/formatted/inputDir -m C:/inputModelDir -o C:/outputDir -n 5


  1. H.Akbari, L.Macyszyn, X.Da, R.L.Wolf, M.Bilello, R.Verma, D.M.O'Rourke, C.Davatzikos, "Pattern analysis of dynamic susceptibility contrast-enhanced MR imaging demonstrates peritumoral tissue heterogeneity", Radiology. 273(2):502-10, 2014, DOI:10.1148/radiol.14132458