Skip to content

Medical Segmentation Decathlon, Task02 Heart: left-atrium segmentation on 30 cardiac MRIs

pl.load("msd_task02_heart") loads this dataset. Twenty training MRIs with the left atrium labelled (King's College London) and ten unlabelled test MRIs, read member by member out of the official 456 MB tar on S3 (HTTP Range; the 61 headers cost 32 KB, each volume its own size). The reference cardiac-segmentation exercise: mask.to_mesh(label=1) gives the atrium surface in LPS millimetres. ds.filter(split='Tr') keeps the labelled cases.

import pardesline as pl

ds = pl.load("msd_task02_heart")
print(ds.describe())  # roles, frames, units, labels, licence, citation
s = ds[0]
obj = s.mri  # fetched now, cached under ~/.cache/pardesline/datasets/msd_task02_heart

Provenance and licence
Source remote archive read by HTTP Range: https://msd-for-monai.s3-us-west-2.amazonaws.com/Task02_Heart.tar
Scheme archive
Size of the whole source 456 MB
Licence CC-BY-SA-4.0
Commercial use allowed by the licence
Redistribution allowed
Licence note CC-BY-SA 4.0 as stated in the archive's dataset.json (checked 2026-10-13); share-alike applies to derived data.
Terms http://medicaldecathlon.com/
Homepage http://medicaldecathlon.com/
Cite Antonelli M., Reinke A., Bakas S., et al. The Medical Segmentation Decathlon. Nat Commun 13:4128, 2022. Simpson A.L. et al. A large annotated medical image dataset for the development and evaluation of segmentation algorithms. arXiv:1902.09063, 2019.
Samples grouped by id

Roles (what each sample contains)

Role Type Frame Units Labels / layout Pattern or file
mri volume (MR) from the file header from the file header Task02_Heart/images{split}/{id}.nii.gz
Cardiac MRI (NIfTI, LPS mm from the header); split Tr = labelled, Ts = test
mask labels from the file header from the file header 1=left_atrium Task02_Heart/labelsTr/{id}.nii.gz
Left-atrium label map (training cases only)

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("msd_task02_heart").take(2)
ds.export("./standardized")        # PLY / MetaImage + sidecars + manifest.json

Algorithms and models that run on it

Generated from src/pardesline/data/manifests/msd_task02_heart.json on 2026-10-14.