Liver tumour CT with three radiologists segmentations

pl.load("liver_tumor_patient01") loads this dataset. One abdominal CT and three independent manual segmentations of the tumour, from the SimpleITK segmentation-evaluation notebook. The inter-observer agreement exercise (Dice, Hausdorff).

import pardesline as pl

ds = pl.load("liver_tumor_patient01")
print(ds.describe())  # roles, frames, units, labels, licence, citation
s = ds[0]
obj = s.ct  # fetched now, cached under ~/.cache/pardesline/datasets/liver_tumor_patient01

Provenance and licence

Source	direct HTTPS downloads with checksums: liverTumorSegmentations/Patient01Homo.mha, liverTumorSegmentations/Patient01Homo_Rad01.mha, liverTumorSegmentations/Patient01Homo_Rad02.mha, liverTumorSegmentations/Patient01Homo_Rad03.mha
Scheme	https
Size of the whole source	2 MB
Licence	not a standard licence: see the note
Redistribution	unknown
Licence note	Distributed with the SimpleITK notebooks (Apache-2.0 code; data terms not stated). Terms of use TO CONFIRM with the data owner before any redistribution.
Terms	https://github.com/InsightSoftwareConsortium/SimpleITK-Notebooks
Cite	SimpleITK Notebooks, Insight Software Consortium.
Samples grouped by	patient
Samples listed in the manifest	1

Roles (what each sample contains)

Role	Type	Frame	Units	Labels / layout	Pattern or file
ct	volume (CT)	from the file header	from the file header		liverTumorSegmentations/Patient01Homo.mha Abdominal CT (8-bit, windowed)
rad01	labels	from the file header	from the file header	1=tumor	liverTumorSegmentations/Patient01Homo_Rad01.mha Tumour segmentation by radiologist 1
rad02	labels	from the file header	from the file header	1=tumor	liverTumorSegmentations/Patient01Homo_Rad02.mha Tumour segmentation by radiologist 2
rad03	labels	from the file header	from the file header	1=tumor	liverTumorSegmentations/Patient01Homo_Rad03.mha Tumour segmentation by radiologist 3

Standardized layout

Files are fetched on first access into the dataset cache and stay in their original format; ds.export(folder) writes every sample as standard files (PLY for points and meshes, MetaImage for volumes and label maps) with a .pardesline.json sidecar carrying frame, units, labels, licence and provenance, plus a manifest for the folder. Label maps become exact surfaces with to_mesh(label=k), DICOM series become one volume, raw arrays are decoded through the layout above.

ds = pl.load("liver_tumor_patient01").take(2)
ds.export("./standardized")        # PLY / MetaImage + sidecars + manifest.json

Algorithms and models that run on it

• labels roles: mask.to_mesh(label=k), mask.to_numpy(), metrics: dice, hd95 (evaluation-metrics-3d) ; catalog modules: pyvista_volume_processing, simpleitk_registration, simpleitk_segmentation
• volume roles: volume.align(moving), volume.segment(lower, upper), volume.to_mesh(isovalue=...), volume.to_sitk() ; catalog modules: pyvista_volume_processing, simpleitk_registration, simpleitk_segmentation
• Supervised segmentation: prepare_nnunet_dataset.py writes this dataset in the nnU-Net v2 raw layout; PointNet trains on the label surfaces.
• Self-supervised pretraining: 3D JEPA reads the volumes directly.

Related

• Datasets catalog · How to download a medical imaging dataset in Python · Add a dataset
• Tags: medical, liver, ct, segmentation, inter-observer

_{Generated from src/pardesline/data/manifests/liver_tumor_patient01.json on 2026-10-10.}