PI-CAI — 2D Slice-Based T2-Weighted Prostate MRI Dataset with csPCa Labels

Axial T2-weighted prostate MRI slices from the PI-CAI (Prostate Imaging: Cancer AI) Public Training and Development Dataset, converted to a 2D, slice-level dataset by Chehab Lab. Each row is one slice with the patient's case-level label for clinically significant prostate cancer (csPCa) and their clinical metadata.

📦 Dataset Structure

• 1,476 patients, 1,476 studies, 33,085 slices. One study per patient (the earliest MRI).
• Label: cspca, with 1,057 negative and 419 positive patients.

column	type	description
volume_id	int32	PI-CAI patient ID
study_id	int32	PI-CAI study ID
slice_id	int32	axial slice index in the original volume
image	image	T2W slice, 8-bit grayscale PNG
cspca	int32	1 = csPCa, 0 = not
case_isup	int32	case-level ISUP grade
patient_age	float32	age in years
psa	float32	prostate-specific antigen (ng/mL)
psad	float32	PSA density (ng/mL²)
prostate_volume	float32	prostate volume (mL)
center	string	acquisition centre

⚙️ Preprocessing

• Only the axial T2W sequence (*_t2w.mha) is included. ADC, HBV, sagittal and coronal sequences are not.
• When a patient has several studies, only the earliest MRI is kept.
• Per volume, intensities are clipped to the 1st–99th percentiles, scaled to 0–255 and saved as 8-bit PNG.
• The original in-plane resolution is kept with no resizing (mostly 384×384, 512×512 or 1024×1024).
• Constant (empty) slices are removed. slice_id keeps the index of the slice in the original volume.

🚀 Usage

from datasets import load_dataset
import matplotlib.pyplot as plt

ds = load_dataset("chehablab/PI-CAI", split="train")
sample = ds[1500]

plt.imshow(sample["image"], cmap="gray")
plt.title(
    f"patient {sample['volume_id']} | slice {sample['slice_id']} | "
    f"csPCa={sample['cspca']} | ISUP {sample['case_isup']}"
)
plt.axis("off")
plt.show()

🏷️ Label Descriptions

csPCa (cspca)

Clinically significant prostate cancer, taken from case_csPCa in the PI-CAI marksheet. A case is positive when it has a lesion with ISUP grade ≥ 2.

ISUP Grade (case_isup)

grade	meaning
0	benign or no cancer
1	Gleason 3+3 (not clinically significant)
2	Gleason 3+4
3	Gleason 4+3
4	Gleason 8
5	Gleason 9–10

🧪 Use Cases

• Patient-level csPCa classification from 3D volumes (multiple instance learning, slice aggregation)
• Slice-level representation learning and benchmarking of pretrained encoders
• Studying the effect of clinical metadata (PSA, PSA density, prostate volume, age)
• Multi-centre domain shift (center)

⚠️ Important Notes

• Split by volume_id, not by row, so slices of the same patient are never in both train and test.
• cspca and case_isup are case-level labels: every slice of a patient has the same label, even slices without a lesion.
• case_isup, psa, psad, prostate_volume and patient_age can be missing for some patients.
• Labels come from the public PI-CAI marksheet (a mix of histopathology and radiology-derived labels), not from the hidden challenge test set.
• Images are not resized. Resize them before training if your model needs a fixed input size.

📚 Citation

If you use this dataset, please cite Chehab Lab and the original PI-CAI work:

@article{saha2024picai,
  title   = {Artificial intelligence and radiologists in prostate cancer detection on MRI (PI-CAI): an international, paired, non-inferiority, confirmatory study},
  author  = {Saha, Anindo and Bosma, Joeran S. and Twilt, Jasper J. and others},
  journal = {The Lancet Oncology},
  volume  = {25},
  number  = {7},
  pages   = {879--887},
  year    = {2024},
  doi     = {10.1016/S1470-2045(24)00220-1}
}

Original sources: PI-CAI challenge · images (Zenodo) · labels (GitHub)

📜 License

CC BY-NC 4.0, the same as the original PI-CAI public dataset: attribution is required and commercial use is not permitted.

Chehab Lab @ 2026