Cancer Imaging Phenomics Toolkit (CaPTk)  1.8.0.Beta
Brain Cancer: Glioblastoma Infiltration Index (Recurrence)

This application provides a probability map of deeply infiltrating tumor in the peritumoral edema/invasion region that largely agrees with subsequent recurrence in de novo glioblastoma patients, via multi-parametric MRI analysis, as shown in [1-3].

REQUIREMENTS:

  1. Co-registered Multimodal MRI: T1, T1-Gd, T2, T2-FLAIR, DSC-MRI, DTI-AX, DTI-FA, DTI-RAD, DTI-TR. Ensure that these are the identified modalities in the drop-down menus next to each loaded image.
  2. Segmentation labels of the tumor sub-regions in a single NIfTI (.nii.gz) file: Non-enhancing tumor core (Label=1), Enhancing tumor core (Label=4), Edema (Label=2)
  3. (For training only) Segmentation labels of the near-region and far-region within the Edema. ROI (Label=1), other (Label=0)
  4. The data for each patient should be organized in the following directory structure. When running in the command-line, filenames must i nclude words in BOLD to be identified as respective required files.
    • Subject_ID
      1. features.csv
      2. CONVENTIONAL
        • my_t1_file.nii.gz
        • my_t2_file.nii.gz
        • my_t1ce_file.nii.gz
        • my_flair_file.nii.gz
      3. DTI
        • my_axial_file.nii.gz
        • my_fractional_file.nii.gz
        • my_radial_file.nii.gz
        • my_trace_file.nii.gz
      4. PERFUSION
        • my_perfusion_file.nii.gz
      5. SEGMENTATION
        • my_tumor_segmentation_label_file.nii.gz
        • my_near_region_file.nii.gz (only for training a new model)
        • my_far_region_file.nii.gz (only for training a new model)
  5. The data of single or multiple patients should be organized in the above mentioned structure and reside under the same folder, e.g.:
    • Input_Directory
      1. Subject_ID1
      2. Subject_ID2
      3. ...
      4. Subject_IDn

USAGE:

  • Infiltration prediction on loaded subject.
    1. Load the required images in CaPTk and correctly assign the modality label in the drop-down menu.
    2. Load the segmentation labels of the tumor sub-regions from a single NIfTI (.nii.gz) file. The labels included in the file should represent the Non-enhancing tumor core (Label=1), Enhancing tumor core (Label=4), and Edema (Label=2).
    3. Select the "Model Directory". Note that a model trained on a cohort of HUP can be found in ftp://www.nitrc.org/home/groups/captk/downloads/models/recurrence.zip
    4. Select the "Output Directory" and click on "Confirm".
    5. The result is saved in the output folder and also loaded in the list of modalities (within ~2 minutes).
  • Infiltration prediction on a batch of subjects.
    • Train a new model:
      1. Select the "Training Directory", e.g., Data_of_multiple_patients (confirm the data follows the structure instructed above).
      2. Select the "Output Directory" where the trained model should be saved.
      3. Click on 'Confirm'.
      4. A pop-up window will confirm the completion of model training (~1.5*NoOfSubjects minutes).
      • This application is also available as with a stand-alone CLI for data analysts to build pipelines around.
        ${CaPTk_InstallDir}/bin/RecurrenceEstimator.exe -t 0 -i C:/RecurrenceSubjects -o C:/RecurrenceModel
        
    • Use existing model:
      1. Select the "Model Directory". Note that a model trained on a cohort of HUP can be found in ftp://www.nitrc.org/home/groups/captk/downloads/models/recurrence.zip
      2. Select the "Test Directory", e.g., Data_of_multiple_patients (confirm the data follows the structure instructed above).
      3. Select the "Output Directory", where the user wants to save the infiltration maps.
      4. Click on 'Confirm'.
      5. A pop-up window will confirm the completion of infiltration map calculations (~1.5*NoOfSubjects minutes).
      • This application is also available as with a stand-alone CLI for data analysts to build pipelines around:
        ${CaPTk_InstallDir}/bin/RecurrenceEstimator.exe -t 1 -i C:/RecurrenceSubjects -o C:/RecurrenceOutput -m C:/RecurrenceModel
        



References:

  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
  2. H.Akbari, L.Macyszyn, J.Pisapia, X.Da, M.Attiah, Y.Bi, S.Pal, R.Davuluri, L.Roccograndi, N.Dahmane, R.Wolf, M.Bilello, D.M.O'Rourke, C.Davatzikos, "Survival Prediction in Glioblastoma Patients Using Multi-parametric MRI Biomarkers and Machine Learning Methods", American Society of Neuroradiology, O-525:2042-2044, 2015. (http://www.asnr.org/sites/default/files/proceedings/2015_Proceedings.pdf)
  3. H.Akbari, L.Macyszyn, X.Da, M.Bilello, R.L.Wolf, M.Martinez-Lage, G.Biros, M.Alonso-Basanta, D.M.O'Rourke, C.Davatzikos. "Imaging Surrogates of Infiltration Obtained Via Multiparametric Imaging Pattern Analysis Predict Subsequent Location of Recurrence of Glioblastoma", Neurosurgery. 78(4):572-80, 2016, DOI:10.1227/NEU.0000000000001202