Join

ProstateX

Notice

The PROSTATEx challenge will close for new submissions soon. The PROSTATEx training and testing cohorts will be part of the training dataset released for a new clinically significant prostate cancer detection challenge, PI-CAI 2022, deprecating the PROSTATEx benchmark. PI-CAI is planned to air in May 2022, and therefore submissions to PROSTATEx will close April 30th, 2022.

Decription

This challenge is the live continuation of the offline PROSTATEx Challenge ("SPIE-AAPM-NCI Prostate MR Classification Challenge”) that was held in conjunction with the 2017 SPIE Medical Imaging Symposium. In this challenge, the task is to predict the clinical significance of prostate lesions found in MRI data. 

Data

This collection is a retrospective set of prostate MR studies. Studies include T2-weighted (T2W), proton density-weighted (PD-W), dynamic contrast-enhanced (DCE), and diffusion-weighted (DW) imaging. The images were acquired on two different types of Siemens 3T MR scanners, the MAGNETOM Trio and Skyra. T2-weighted images were acquired using a turbo spin echo sequence and had a resolution of around 0.5 mm in-plane and a slice thickness of 3.6 mm. The DCE time series were acquired using a 3-D turbo flash gradient echo sequence with a resolution of around 1.5 mm in-plane, a slice thickness of 4 mm and a temporal resolution of 3.5 s. The proton density weighted image was acquired prior to the DCE time series using the same sequence with different echo and repetition times and a different flip angle. Finally, the DWI series were acquired with a single-shot echo planar imaging sequence with a resolution of 2 mm in-plane and 3.6 mm slice thickness and with diffusion-encoding gradients in three directions. Three b-values were acquired (50, 400, and 800), and subsequently, the ADC map was calculated by the scanner software. All images were acquired without an endorectal coil.

The images come in two encodings. The acquired MR is provided in DICOM encoding. Additionally, Ktrans images are provided. They come in MHD format. Ktrans is a key pharmacokinetic parameter computed from the available Dynamic contrast-enhanced T1-weighted series. Each patient has one study with several DICOM images and one Ktrans image. The Ktrans image is encoded in two files ProstateX-[ProxID]-Ktrans.[mhd/zraw], where ProxID is the ProstateX patient identifier. The DICOM images comprise several Series each comprising several Instances. The DICOM files are documented in the ProstateX-Images.csv file. The columns in that file encode the following:

  • ProxID – ProstateX patient identifier.

  • Name – Series Description

  • Studydate – Study Date

  • fid – Finding ID

  • Pos – Scanner Coordinate position of the finding

  • WorldMatrix – Matrix describing image orientation and scaling

  • ijk – image col,row,slice coordinate of finding

  • ImageUID – Image Identifier

  • TopLevel

    • 0 - Series forms one image

    • 1 – A set of Series forms a 4D image (e.g. Dynamic MR)

    • NA – Series form one image, but is part of a Level 1 4D image

  • SpacingBetweenSlices – Scalar Spacing between slices

  • VoxelSpacing – Vector with x,y,z spacing scalars

  • Dim – Vector with 4D dimensions of the image

  • DCMSerDescr – The original DICOM Series Description

  • DCMSerUID – The DICOM Series UID

  • DCMSerNum – The DICOM Series Number

  • InstanceUIDList – DICOM Instances that make up this series

  • ImageUIDList – TopLevel-NA Images the make up this Toplevel 1 image

For example, to get the ADC image of Patient ProstateX-0123 do the following. After you imported the DICOM files into your environment, go to patient ProstateX-0123 and find the series with ADC in it. In this case it is ‘ep2d_diff_tra_DYNDIST_ADC’. It has SeriesNumber 8. The DICOM images in that series form the ADC image for this challenge. Image slice j at coordinate i,j contains a finding fid. See findings for more details.

The data can be found and downloaded from The Cancer Imaging Archive PROSTATEx webpage under the tab 'Detailed Description'.

Findings

The findings are documented in the ProstateX-Findings.csv table. Documentation for the columns in that table is as follows:

  • ProxID – ProstateX patient identifier
  • fid - Finding ID

  • pos - Scanner Coordinate position of the finding

  • ClinSig – Identifier available in training set that identifies whether this is a clinically significant finding. Either the biopsy GleasonScore was 7 or higher. Findings with a PIRADS score 2 were not biopsied and are not considered clinically significant. In our center the occurrence of clinically significant cancer in PIRADS 2 lesions is less than 5%.

Evaluation

The metric used for evaluation is the Area under the ROC curve (AUC). Predictions for all the lesions in the test data set are required.

The submission file requires the following three columns:

  • ProxID
  • fid
  • ClinSig

ProxID and fid are the ProstateX patient identifier and Finding ID that are given in ProstateX-Findings-Test.csv.

ClinSig is the predicted clinical significance that is generated by your algorithm (must be between 0 and 1).

For an example submission file see this page.

Acknowledgments

The prostate MR imaging was performed at the Radboud University Medical Centre (Radboudumc) in the Prostate MR Reference Center under the supervision of prof.dr. Jelle Barentsz. The Radboudumc is located in Nijmegen, The Netherlands. The dataset was collected and curated for research in computer-aided diagnosis of prostate MR under the supervision of Henkjan Huisman, Radboudumc. 

The SPIE and AAPM are acknowledged for their support in setting up this challenge. The TCIA is acknowledged for anonymizing and hosting the data. The organizers of the initial offline challenges are acknowledged: Samuel G. Armato, Lubomir Hadjiyski, and Karen Drukker and all other members from SPIE, AAPM, NIH, TCIA involved.