anonymous-neurips-ED/CTSpinoPelvic1K-Sample

CTSpinoPelvic1K

A fused spine + pelvis 3D CT segmentation dataset built by patient-level crosswalk between three public sources:

1. TCIA CT COLONOGRAPHY — DICOM CT volumes (prone + supine per patient)
2. CTSpine1K (COLONOG subset) — VerSe-convention vertebral label masks
3. CTPelvic1K dataset2 — sacrum + bilateral hip label masks

Annotations are placed onto the TCIA CT volume with the highest bone coverage (HU > 200), separately per anatomy. For ~650 patients both annotations land on the same series (fused cases); for the rest, spine and pelvic labels target different prone/supine acquisitions (separate cases).

Labels (10-class)

ID	Name	Source
0	background	—
1	L1	CTSpine1K (VerSe label 20 → 1)
2	L2	CTSpine1K (VerSe label 21 → 2)
3	L3	CTSpine1K (VerSe label 22 → 3)
4	L4	CTSpine1K (VerSe label 23 → 4)
5	L5	CTSpine1K (VerSe label 24 → 5)
6	L6 / LSTV	CTSpine1K (VerSe label 25 → 6) — lumbarized S1
7	sacrum	CTPelvic1K (dataset2 label 1 → 7)
8	left hip	CTPelvic1K (dataset2 label 2 → 8)
9	right hip	CTPelvic1K (dataset2 label 3 → 9)

CTPelvic1K's sacrum takes priority over CTSpine1K's sacrum (label 26) to avoid the two labelling conventions colliding in cases of lumbosacral transitional vertebrae.

Orientation

All volumes are canonicalised to PIR (Posterior-Inferior-Right). The CT and its label map share exactly the same 4×4 affine; no resampling is needed before training.

LSTV annotation

Each case carries two complementary LSTV (lumbosacral transitional vertebra) annotations:

• lstv_vertebral — derived from CTSpine1K by counting lumbar labels in the segmentation (4 → sacralization, 5 → normal, 6 → lumbarization).
• lstv_pelvic — derived from CTPelvic1K filename qualifiers (any substring containing "sacralization" → sacralization).
• lstv_agreement — True when both sources agree, False when they disagree, None when either side is uninformative.
• lstv_class — integer 0–3 summarising the dominant call (0=normal, 1=lumbarization, 2=semi-sacralization, 3=sacralization). Pelvic label takes priority.

Splits

70 / 15 / 15 train / val / test, stratified by (lstv_class × match_type) so each split contains the rare sacralization and lumbarization classes.

File format

Each case is a single .npz file under data/<split>/token_<N>.npz:

import numpy as np, json

d = np.load("token_17.npz", allow_pickle=False)
ct     = d["ct"]       # int16  (Z, Y, X)   HU
label  = d["label"]    # uint8  (Z, Y, X)   0..9
affine = d["affine"]   # float32 (4, 4)    RAS affine
meta   = json.loads(str(d["meta"]))

print(meta["match_type"], meta["lstv_class"], meta["spine_bone_pct"])

Quickstart — PyTorch

from dataset_interface import CTSpinoPelvicDataset
from torch.utils.data import DataLoader

ds  = CTSpinoPelvicDataset(
    root      = "anonymous-mlhc/CTSpinoPelvic1K",
    split     = "train",
    cache_dir = "~/.cache/ctspinopelvic1k",
)
dl  = DataLoader(ds, batch_size=1, shuffle=True)

for batch in dl:
    ct, label = batch["ct"], batch["label"]   # (B,1,Z,Y,X) / (B,Z,Y,X)
    ...

Quickstart — MONAI

from monai.transforms import (
    Compose, RandCropByPosNegLabeld, RandFlipd, NormalizeIntensityd,
)
from dataset_interface import CTSpinoPelvicDataset

transforms = Compose([
    NormalizeIntensityd(keys="ct", subtrahend=0, divisor=1000),
    RandCropByPosNegLabeld(keys=("ct","label"), label_key="label",
                           spatial_size=(96,96,96), pos=2, neg=1, num_samples=2),
    RandFlipd(keys=("ct","label"), prob=0.5, spatial_axis=(0,1,2)),
])

ds = CTSpinoPelvicDataset(root="anonymous-mlhc/CTSpinoPelvic1K",
                          split="train", transform=transforms)

Citation

Please cite the source datasets (CTSpine1K, CTPelvic1K, TCIA CT COLONOGRAPHY) alongside this derivative release. BibTeX entries are provided in CITATION.cff.

License

• Source datasets — CT COLONOGRAPHY (TCIA), CTSpine1K, CTPelvic1K — retain their respective licenses.
• Derivative fused labels, splits, and code: CC BY-NC 4.0 (non-commercial).