QIN-PROSTATE-Repeatability

Annotated test-retest multi-parametric MRI of the prostate. Fifteen treatment-naive subjects each scanned twice within two weeks on the same 3T scanner with an endorectal coil, yielding 30 imaging studies. Each study contains T2-weighted axial, diffusion-weighted (b=0, b=1400), apparent diffusion coefficient (ADC), and dynamic contrast-enhanced (DCE) subtraction series, with manual whole-gland, peripheral-zone, suspected-lesion, and normal-PZ segmentations stored as DICOM SEG.

Dataset Details

No predefined train/val/test split — this is a repeatability cohort. The natural split is two timepoints per subject (Time Point 1 vs Time Point 2). All 15 subjects are treatment-naive; 11 have at least one suspected lesion (yielding 22 lesion-positive study-timepoints).

Segmentation Classes

A single radiologist (10+ years abdominal-MRI experience) annotated each study in 3D Slicer, blinded to subject ID and to baseline/repeat status. SegmentAlgorithmType in every SEG is MANUAL. Per study:

1. Whole prostate gland — annotated on T2W/ADC/DCE
2. Whole peripheral zone — annotated on T2W/ADC/DCE
3. Suspected lesion (tumor) — annotated when present (22 of 30 study-timepoints)
4. Normal PZ tissue — annotated on ADC only

Bundled DICOM SEG objects are the canonical and only ground truth — no alternative mask source exists for this collection.

Structure

images/<PatientID>/<StudyInstanceUID>/<SeriesInstanceUID>/*.dcm           # MR
segmentations/<PatientID>/<StudyInstanceUID>/<SeriesInstanceUID>/*.dcm    # SEG (multi-frame)
structured_reports/<PatientID>/<StudyInstanceUID>/<SeriesInstanceUID>/*.dcm  # SR (TID 1500)
series_to_patient.json                                                    # series-level metadata

Each SEG SOP references its source-image series via DICOM metadata (DerivationImageSequence -> SourceImageSequence -> ReferencedSOPInstanceUID). SR objects (TID 1500/1411) carry per-ROI volume (mm³) and mean ADC (µm²/s) with Time Point (1 or 2) and Tracking UID linkage to the segmented ROIs.

Important Notes for Loaders

• mpMRI is NOT co-registered. T2W, DWI, ADC, and DCE have different geometries. Use T2W axial as the geometric reference and resample/align other modalities to it rather than stacking as channels.
• DICOM SEG ⇄ ITK conversion is needed to get a labelmap volume; use dcmqi's segimage2itkimage or pydicom-seg.
• Test/retest pairs: each subject has two studies (TP1 vs TP2). Decide whether to treat as 30 independent samples or 15 paired samples for evaluation.
• Lesion mask absent in 4 subjects (8 study-timepoints) — handle missing-class gracefully.

Source

• TCIA collection: https://www.cancerimagingarchive.net/collection/qin-prostate-repeatability/
• DOI: 10.7937/K9/TCIA.2018.MR1CKGND
• Companion paper: https://pmc.ncbi.nlm.nih.gov/articles/PMC6278692/

Citation

@article{fedorov2018qinprostate,
  author  = {Fedorov, Andrey and Schwier, Michael and Clunie, David and
             Herz, Christian and Pieper, Steve and Kikinis, Ron and
             Tempany, Clare and Fennessy, Fiona},
  title   = {An annotated test-retest collection of prostate multiparametric MRI},
  journal = {Scientific Data},
  volume  = {5},
  pages   = {180281},
  year    = {2018},
  doi     = {10.1038/sdata.2018.281}
}

@misc{qinprostaterepeatability2018tcia,
  author    = {Fedorov, A. and Schwier, M. and Clunie, D. and Herz, C. and
               Pieper, S. and Kikinis, R. and Tempany, C. and Fennessy, F.},
  title     = {Data From QIN-PROSTATE-Repeatability [Dataset]},
  year      = {2018},
  publisher = {The Cancer Imaging Archive},
  doi       = {10.7937/K9/TCIA.2018.MR1CKGND}
}

@article{fedorov2017repeatability,
  author  = {Fedorov, Andrey and Vangel, Mark G. and Tempany, Clare M. and
             Fennessy, Fiona M.},
  title   = {Multiparametric Magnetic Resonance Imaging of the Prostate:
             Repeatability of Volume and Apparent Diffusion Coefficient
             Quantification},
  journal = {Investigative Radiology},
  volume  = {52},
  number  = {9},
  pages   = {538--546},
  year    = {2017},
  doi     = {10.1097/RLI.0000000000000382}
}