ImagingDataCommons/idc-index-data

NCI Imaging Data Commons (IDC) index

This dataset is a catalog. It contains metadata and cloud locations for every DICOM series in the NCI Imaging Data Commons; it does not contain pixel data.

IDC is an NCI Cancer Research Data Commons repository of publicly available cancer imaging data, co-located with analysis tools in the cloud. To explore it interactively instead, use the IDC portal. Without downloading anything, any image in IDC can be viewed in the browser. To query IDC in plain language, point an AI assistant at its agent interfaces -- a hosted MCP server, an agent skill, and a REST API over the same metadata.

This catalog describes IDC v24 (released 2026-04-16): 1,032,911 series across 166,740 studies, 85,362 patients and 176 collections, totalling 99.3 TB of imaging data.

One row is one DICOM series, with its collection, patient, study and series attributes, its license and source DOI, and the S3 URL to fetch it from. Use it to find the data you want here, then download only that -- the alternative is sifting through 99 TB.

These are the same Parquet files published with each idc-index-data release, republished here for the dataset viewer, the SQL Console, automatic Croissant metadata, and a citable, versioned record you can pin.

Every config has a single split named train, the Hub default, because many downstream tools assume it exists. It carries no train/test meaning.

Quickstart

pip install datasets idc-index

datasets reads this catalog; idc-index is the client that downloads the DICOM files it points at. Filter here, download there:

from datasets import load_dataset

idx = load_dataset("ImagingDataCommons/idc-index-data", "idc_index", split="train")
sel = idx.filter(
    lambda r: r["collection_id"] == "nsclc_radiomics" and r["Modality"] == "SEG"
)

from idc_index import IDCClient

client = IDCClient()
client.download_from_selection(
    seriesInstanceUID=sel["SeriesInstanceUID"], downloadDir="./idc_data"
)

Downloads come directly from IDC's public AWS and GCS buckets at no cost to you. What lands on disk is DICOM; Loading images as tensors below turns it into arrays.

Every series in this catalog can also be looked at without downloading anything. IDC streams the pixels to a zero-footprint browser viewer, and get_viewer_URL builds a link to any series you have selected:

print(client.get_viewer_URL(seriesInstanceUID=sel["SeriesInstanceUID"][0]))

It picks the viewer that fits the data -- OHIF for radiology, Slim for slide microscopy -- and opens the enclosing study with your series selected. Passing a segmentation, as above, brings it up overlaid on the images it segments.

Query the catalog without downloading anything, using DuckDB:

SELECT collection_id, COUNT(*) AS series, SUM(series_size_MB) / 1e6 AS size_TB
FROM 'hf://datasets/ImagingDataCommons/idc-index-data/idc_index.parquet'
GROUP BY 1 ORDER BY size_TB DESC LIMIT 10;

The same queries run in the SQL Console tab on this page, with no local setup.

Loading images as tensors

A DICOM series is not an array yet. For CT, MR and PET it is usually one file per slice. The slices must be ordered by their position in space, not by file name, and their stored values rescaled to physical units, such as Hounsfield units for CT. Not every series is a volume at all: localizers, uneven slice spacing and gantry tilt are all common.

highdicom handles this, and returns a Volume that keeps voxel spacing and the patient-space affine next to the array. The examples below were tested with highdicom 0.28.2.

pip install "highdicom>=0.28.2" duckdb idc-index torch

Start in the catalog. volume_geometry_index flags every CT, MR and PET series whose slices form a regularly spaced 3D grid, so series that won't load as a volume are never downloaded:

import duckdb

hf = "hf://datasets/ImagingDataCommons/idc-index-data"
query = f"""
    SELECT SeriesInstanceUID
    FROM '{hf}/idc_index.parquet'
    JOIN '{hf}/volume_geometry_index.parquet' USING (SeriesInstanceUID)
    WHERE collection_id = 'nsclc_radiomics' AND Modality = 'CT'
      AND regularly_spaced_3d_volume
    LIMIT 3
"""
uids = [row[0] for row in duckdb.sql(query).fetchall()]

Download each series into its own directory, then load it:

from pathlib import Path

import highdicom as hd
import numpy as np
import pydicom
import torch
from idc_index import IDCClient

client = IDCClient()
client.download_from_selection(
    seriesInstanceUID=uids, downloadDir="idc_data", dirTemplate="%SeriesInstanceUID"
)


def load_volume(uid):
    files = Path("idc_data", uid).glob("*.dcm")
    return hd.get_volume_from_series(
        [pydicom.dcmread(f) for f in files], dtype=np.float32
    )


def volume_to_channel_first_tensor(vol):
    # Convert to a tensor with a leading channel dimension, as is typically
    # required in pytorch

    # The result of a match_geometry operation may be permuted/flipped
    # so the resulting numpy array is non-contiguous
    arr = np.ascontiguousarray(vol.array)

    if vol.number_of_channel_dimensions == 0:
        # Volume has no channel -> add one
        # Result is (channels, slices, rows, columns)
        t = torch.from_numpy(arr).unsqueeze(0)
    elif vol.number_of_channel_dimensions == 1:
        # Volume has a trailing channel -> permute to the front
        t = torch.from_numpy(arr).permute([3, 0, 1, 2])
    else:
        raise ValueError("Expected at most one channel dimension")

    return t


vol = load_volume(uids[0])
image = volume_to_channel_first_tensor(vol)

print(image.shape, vol.spacing)  # spacing in mm, of the three spatial axes

get_volume_from_series raises ValueError for a series that is not a regular grid. Those are the series the geometry filter above leaves out. Pass dtype explicitly; the default is float64.

get_volume_from_series has other parameters, for example to choose which pixel transforms are applied. See its documentation.

Segmentations

seg_index has a row for each segmentation (DICOM SEG) series, and segmented_SeriesInstanceUID names the image series it segments. This example takes one from NLSTSeg, expert segmentations of lung lesions in NLST CT, downloads it with its CT, and puts the mask on the CT's grid:

query = f"""
    SELECT s.SeriesInstanceUID, s.segmented_SeriesInstanceUID
    FROM '{hf}/seg_index.parquet' s
    JOIN '{hf}/idc_index.parquet' i USING (SeriesInstanceUID)
    JOIN '{hf}/volume_geometry_index.parquet' g
      ON g.SeriesInstanceUID = s.segmented_SeriesInstanceUID
    WHERE i.analysis_result_id = 'nlstseg' AND s.total_segments > 1
      AND g.regularly_spaced_3d_volume
    LIMIT 1
"""
seg_uid, image_uid = duckdb.sql(query).fetchone()
client.download_from_selection(
    seriesInstanceUID=[seg_uid, image_uid],
    downloadDir="idc_data",
    dirTemplate="%SeriesInstanceUID",
)

ct = load_volume(image_uid)
seg = hd.seg.segread(next(Path("idc_data", seg_uid).glob("*.dcm")))
labels = seg.get_volume(combine_segments=True)
labels = labels.match_geometry(ct)

image = volume_to_channel_first_tensor(ct)
mask = volume_to_channel_first_tensor(labels)  # (1, slices, rows, columns)
print({s.SegmentNumber: s.SegmentLabel for s in seg.SegmentSequence})

A segmentation often covers fewer slices than its image, is rotated relative to the source image, or stores the slices in the opposite order; match_geometry pads, flips and rotates it (as required) onto the image grid. It raises RuntimeError when the two grids are offset by a fraction of a voxel, and rounding in stored positions alone can cause that: some nsclc_radiomics segmentations sit 2e-5 voxels off their CT, just beyond the default tolerance, and match_geometry(ct, tol=1e-4) accepts them. Loosen tol no further than you need: a large tolerance can hide a segmentation that is genuinely misaligned with its image.

combine_segments=True returns a label map, in which each voxel holds the number of the segment it belongs to, or 0. NLSTSeg segments each lesion separately, so mask numbers the lesions.

A label map cannot hold a voxel that belongs to two segments, and segments may overlap. The expert segmentations in nsclc_radiomics, for example, outline the primary tumor inside the lung that contains it, and combine_segments=True raises RuntimeError on them. For those, seg.get_volume() returns one binary mask per segment down the last axis, which volume_to_channel_first_tensor moves to the front: (segments, slices, rows, columns). Losses differ in which form they take. PyTorch's CrossEntropyLoss takes class indices, as in a label map, while others, such as the default behavior of MONAI's DiceLoss, take one channel per segment. However, be aware that many common loss functions (such as Dice) expect the segments to be non-overlapping even when using the one-channel-per-segment representation, the so-called "one-hot" format.

get_volume can also select a subset of the segments. See its documentation for this and its other parameters.

Training and other image types

To feed a DataLoader, call load_volume from a torch.utils.data.Dataset. Volumes differ in shape, so bring them to a common one before batching, for example with vol.pad_or_crop_to_spatial_shape((64, 256, 256)) or by resampling.

• Radiographs and mammograms (CR, DX, MG) are one 2D image per file: hd.imread(path).get_frame(1).
• Slide microscopy (SM) is a multi-resolution pyramid with one file per level, and the full-resolution level is usually too large for one array. Read a region of one level with hd.imread(path).get_total_pixel_matrix(), whose row_start, row_end, column_start and column_end bounds are 1-based, end excluded. sm_instance_index gives each file's TotalPixelMatrixRows, TotalPixelMatrixColumns and PixelSpacing_0, so you can pick the level before downloading.

Indices

Each index is a separate config (subset). Load one with the name argument of load_dataset, or select it from the dropdown in the dataset viewer.

Config	Rows	Size	Description
idc_index (default)	1,032,911	73.5 MB	This is the main metadata table provided by idc-index. Each row corresponds to a DICOM series, and contains attributes at the collection, patient, study, and series levels. The table also contains download-related attributes, such as the AWS S3 bucket and URL to download the series.
analysis_results_index	24	0.0 MB	This table contains metadata about the analysis results collections available in IDC. Each row corresponds to an analysis results collection, and contains attributes such as the collection name, types of cancer represented, number of subjects, and pointers to the resources to learn more about the content of the collection
ann_group_index	15,924	0.5 MB	This table contains detailed metadata about individual annotation groups within Microscopy Bulk Simple Annotations (ANN) series in IDC. Each row corresponds to a single annotation group, providing granular information about the graphic type, number of annotations, property codes, and algorithm details. This table can be joined with ann_index using SeriesInstanceUID for series-level context. Note: ANN series are assumed to contain a single instance.
ann_index	7,102	0.2 MB	This table contains metadata about the Microscopy Bulk Simple Annotations (ANN) series available in IDC. Each row corresponds to a DICOM series containing annotations, and includes attributes such as the annotation coordinate type and references to the annotated image series. For detailed group-level information (counts, graphic types, property codes), join with ann_group_index using SeriesInstanceUID. This table can be joined with the main idc_index table using the SeriesInstanceUID column. Note: ANN series are assumed to contain a single instance.
clinical_index	5,756	0.1 MB	This table contains metadata about the tabular data, including clinical data, accompanying images that is available in IDC. Think about this table as a dictionary containing information about the columns for all of the tabular data accompanying individual collections in IDC. Each row corresponds to a unique combination of collection, clinical data table that is available for that collection, and a column from that table. Individual tables referenced from this table can be retrieved using idc-index get_clinical_table() function.
collections_index	176	0.1 MB	This table contains metadata about the collections available in IDC. Each row corresponds to a collection, and contains attributes such as the collection name, types of cancer represented, number of subjects, and pointers to the resources to learn more about the content of the collection.
contrast_index	63,576	1.1 MB	This table contains one row per DICOM series that has contrast agent information. It captures contrast bolus metadata from CT, MR, PT, XA, and RF imaging modalities, including the agent name, ingredient, and administration route. Only series with at least one non-null contrast attribute are included. This table can be joined with the main idc_index table using the SeriesInstanceUID column.
ct_index	266,232	6.3 MB	This table contains one row per CT Image Storage (SOPClassUID 1.2.840.10008.5.1.4.1.1.2) DICOM series in IDC, capturing acquisition and reconstruction parameters that are not included in the main idc_index table. The index can be joined to idc_index on SeriesInstanceUID to combine universal series metadata with CT-specific acquisition parameters. For XRayTubeCurrent, Exposure, and ExposureTime — which vary across instances within a series due to dose modulation — both min and max values are reported. All other attributes are aggregated with ANY_VALUE (one representative instance).
mr_index	123,481	2.9 MB	This table contains one row per MR Image Storage (SOPClassUID 1.2.840.10008.5.1.4.1.1.4) DICOM series in IDC, capturing MR acquisition and sequence parameters that are not included in the main idc_index table. The index can be joined to idc_index on SeriesInstanceUID to combine universal series metadata with MR-specific acquisition parameters. EchoTime and DiffusionBValue are reported as arrays of all distinct per-instance values because they legitimately differ across instances within multi-echo and diffusion-weighted series respectively. All other attributes are aggregated with ANY_VALUE (one representative instance).
pt_index	4,865	0.3 MB	This table contains one row per Positron Emission Tomography Image Storage (SOPClassUID 1.2.840.10008.5.1.4.1.1.128) DICOM series in IDC, capturing PET acquisition, reconstruction and radiopharmaceutical parameters that are not included in the main idc_index table. The index can be joined to idc_index on SeriesInstanceUID to combine universal series metadata with PET-specific acquisition parameters. ActualFrameDuration is reported as an array of all distinct per-instance values because it legitimately varies across frames in dynamic (multi-frame) PET acquisitions. All other attributes are constant within a series and are aggregated with ANY_VALUE.
rtstruct_index	19,358	0.4 MB	This table contains one row per DICOM RT Structure Set (RTSTRUCT) SeriesInstanceUID available from IDC, and captures key metadata about the structure set including the number of ROIs, ROI names, generation algorithms, interpreted types, and the referenced image series. Note: multi-valued columns (ROINames, ROIGenerationAlgorithms, RTROIInterpretedTypes) are aggregated with DISTINCT independently, so positional correspondence between columns is not preserved.
seg_index	190,146	6.5 MB	This table contains one row per DICOM Segmentation SeriesInstanceUID available from IDC, and captures key metadata about the segmentation series including the number of segments, segmentation type, algorithm type and name, and the segmented image series. Note: multi-valued columns (AlgorithmType, AlgorithmName, SegmentedPropertyCategory_CodeMeanings, etc.) are aggregated with DISTINCT independently, so positional correspondence between columns is not preserved. For example, the first value in AlgorithmType does not necessarily pair with the first value in AlgorithmName.
sm_index	76,299	2.2 MB	This table contains metadata about the slide microscopy (SM) series available in IDC. Each row corresponds to a DICOM series, and contains attributes specific to SM series, such as the pixel spacing at the maximum resolution layer, the power of the objective lens used to digitize the slide, and the anatomic location from where the imaged specimen was collected. This table can be joined with the main index table using the SeriesInstanceUID column.
sm_instance_index	391,932	17.9 MB	This table contains metadata about the slide microscopy (SM) series available in IDC. Each row corresponds to an instance from a DICOM Slide Microscopy series available from IDC, identified by SOPInstanceUID, and contains attributes specific to SM series, such as the pixel spacing at the maximum resolution layer, the power of the objective lens used to digitize the slide, and the anatomic location from where the imaged specimen was collected. This table can be joined with the main index table and/or with sm_index using the SeriesInstanceUID column.
version_metadata_index	24	0.0 MB	This table contains metadata about each IDC data release version. Each row corresponds to one IDC version and captures when that version was created. This index can be used to correlate data in other indexes (which include idc_version columns) with the corresponding release timestamps.
volume_geometry_index	319,472	5.1 MB	This table contains one row per DICOM series from IDC for single-frame CT, MR, and PT SOP classes, with boolean columns characterizing the geometric properties of each series. The checks determine whether the series forms a regularly-spaced rectilinear 3D volume (consistent orientation, spacing, dimensions, and slice positions). Series that do not pass all checks may still be usable with additional processing such as resampling or acquisition geometry correction (e.g., for variable-spacing or gantry-tilted acquisitions). Oblique-aware: uses projection-based slice position computation, which handles gantry-tilted CT, oblique MR, and axial PET uniformly.

clinical_index is a dictionary of the clinical tables and columns available per collection -- not the clinical data itself. The clinical tables are not among these artifacts; retrieve them with IDCClient.get_clinical_table().

Data fields

Columns of idc_index, the default config:

Column	Type	Description
collection_id	STRING	short string with the identifier of the collection the series belongs to
analysis_result_id	STRING	this string is not empty if the specific series is part of an analysis results collection; analysis results can be added to a given collection over time
PatientID	STRING	identifier of the patient within the collection (DICOM attribute)
SeriesInstanceUID	STRING
StudyInstanceUID	STRING	unique identifier of the DICOM study (DICOM attribute)
source_DOI	STRING	Digital Object Identifier of the dataset that contains the given series; follow this DOI to learn more about the activity that produced this series
PatientAge	STRING	age of the subject at the time of imaging (DICOM attribute)
PatientSex	STRING	subject sex (DICOM attribute)
StudyDate	STRING	date of the study (de-identified) (DICOM attribute)
StudyDescription	STRING	textual description of the study content (DICOM attribute)
BodyPartExamined	STRING	body part imaged (not applicable for SM series) (DICOM attribute). For derived series (SEG, RTSTRUCT) this reflects the source acquisition, not the segmented anatomy -- what was segmented is recorded in the seg_index table (SegmentedPropertyType_CodeMeanings) and in rtstruct_index.
Modality	STRING	acquisition modality (DICOM attribute)
SOPClassUID	STRING	SOP Class UID identifying the type of DICOM object (e.g., CT Image Storage, Segmentation Storage); more specific than Modality for distinguishing object types (DICOM attribute)
sop_class_name	STRING	human-readable name of the SOP Class (e.g., "CT Image Storage", "Segmentation Storage"); derived from SOPClassUID
TransferSyntaxUID	STRING	Transfer Syntax UID identifying the encoding of the stored instances (e.g., Explicit VR Little Endian, JPEG 2000, HTJ2K); comma-separated when a series contains instances with different encodings, which is common for SM (DICOM attribute)
transfer_syntax_name	STRING	human-readable name of the Transfer Syntax (e.g., "JPEG 2000", "Explicit VR Little Endian"); comma-separated when a series contains instances with different encodings; derived from TransferSyntaxUID
PhotometricInterpretation	STRING	intended interpretation of the pixel data, e.g., MONOCHROME2 (grayscale), RGB or YBR_FULL_422 (color); NULL for non-image objects (e.g., SEG, SR, RTSTRUCT); comma-separated when a series contains instances with different values (DICOM attribute)
PixelRepresentation	STRING	data representation of pixel sample values: "0" for unsigned integers, "1" for signed integers; NULL for non-image objects; comma-separated when a series contains instances with different values (DICOM attribute)
Manufacturer	STRING	manufacturer of the equipment that produced the series (DICOM attribute)
ManufacturerModelName	STRING	model name of the equipment that produced the series (DICOM attribute)
SeriesDate	STRING	date of the series (de-identified) (DICOM attribute)
SeriesDescription	STRING	textual description of the series content (DICOM attribute)
SeriesNumber	STRING	series number (DICOM attribute)
instanceCount	INTEGER	number of instances in the series
license_short_name	STRING	short name of the license that applies to this series
series_init_idc_version	INTEGER	IDC data release version number when this series first appeared in IDC (integer, e.g., 1 for v1)
series_revised_idc_version	INTEGER	IDC data release version number when this series was most recently revised in IDC (integer, e.g., 24 for v24)
aws_bucket	STRING	name of the AWS S3 bucket that contains the series
crdc_series_uuid	STRING	unique identifier of the series within the IDC
series_aws_url	STRING	public AWS S3 URL to download the series in bulk (each instance is a separate file)
series_size_MB	FLOAT	total size of the series in megabytes

Every other config is described by a <config>_schema.json sidecar in this repository, carrying the same table and column descriptions:

Config	Columns	Schema
analysis_results_index	13	analysis_results_index_schema.json
ann_group_index	12	ann_group_index_schema.json
ann_index	3	ann_index_schema.json
clinical_index	6	clinical_index_schema.json
collections_index	12	collections_index_schema.json
contrast_index	4	contrast_index_schema.json
ct_index	21	ct_index_schema.json
mr_index	22	mr_index_schema.json
pt_index	21	pt_index_schema.json
rtstruct_index	6	rtstruct_index_schema.json
seg_index	9	seg_index_schema.json
sm_index	20	sm_index_schema.json
sm_instance_index	15	sm_instance_index_schema.json
version_metadata_index	2	version_metadata_index_schema.json
volume_geometry_index	11	volume_geometry_index_schema.json

They are plain JSON, so you can read one without downloading the data:

import json, urllib.request

url = "https://huggingface.co/datasets/ImagingDataCommons/idc-index-data/resolve/main/seg_index_schema.json"
schema = json.load(urllib.request.urlopen(url))
print(schema["table_description"])
for column in schema["columns"]:
    print(column["name"], "--", column.get("description", ""))

Licensing

The images are not covered by a single license. Every row carries a license_short_name giving the license of that series; the YAML above lists all of them so the dataset appears under each one's Hub filter. Check it per series before redistributing or using data commercially.

License	Series	Commercial use
CC BY 4.0	865,935	allowed
CC BY 3.0	132,303	allowed
CC BY-NC 4.0	28,783	not allowed
CC BY-NC 3.0	5,851	not allowed
National Library of Medicine Terms and Conditions; May 21, 2019	39	see terms

Series under National Library of Medicine Terms and Conditions are governed by https://www.nlm.nih.gov/databases/download/terms_and_conditions.html.

Every license IDC uses -- CC BY and CC BY-NC alike -- requires attribution. See IDC licensing and attribution.

The index files in this repository are a factual catalog of that content and are distributed under the license of the idc-index-data repository. That license covers the tables only, never the referenced images.

Attribution and citation

Attribution is required by every license in this catalog, and it is owed to the source dataset, not to IDC. Each row's source_DOI identifies the dataset the series came from; resolve it to a formatted citation with IDC's citations API or IDCClient.citations_from_selection().

Many IDC collections originate from The Cancer Imaging Archive (TCIA); IDC is a TCIA Data Analysis Center. Those collections additionally carry TCIA's data usage policies and restrictions, including obligations on downstream attribution.

Please also acknowledge IDC itself:

@article{fedorov2023idc,
  title   = {National Cancer Institute Imaging Data Commons: Toward Transparency,
             Reproducibility, and Scalability in Imaging Artificial Intelligence},
  author  = {Fedorov, Andrey and Longabaugh, William J. R. and Pot, David and
             Clunie, David A. and Pieper, Steven D. and Gibbs, David L. and
             Bridge, Christopher and Herrmann, Markus D. and Homeyer, Andr\'e and
             Lewis, Rob and Aerts, Hugo J. W. L. and Krishnaswamy, Deepa and
             Thiriveedhi, Vamsi K. and Ciausu, Cosmin and Schacherer, David P. and
             Bontempi, Dennis and Pihl, Todd and Wagner, Ulrike and
             Farahani, Keyvan and Kim, Erika and Kikinis, Ron},
  journal = {RadioGraphics},
  volume  = {43},
  number  = {12},
  year    = {2023},
  doi     = {10.1148/rg.230180}
}

Versioning

Tags on this repo match the idc-index-data releases one for one, and main always holds the most recent published release. Pin a version to keep results reproducible:

load_dataset("ImagingDataCommons/idc-index-data", "idc_index", revision="24.2.2")

This release, 24.2.2, indexes IDC v24 (released 2026-04-16). Not every idc-index-data release is published here; tags on this repo are a subset of the GitHub releases.

This card is generated, not maintained here. Every publish regenerates README.md -- YAML front matter and all -- from the release artifacts and commits it over whatever the Hub currently holds. Edits made through the Hub UI and community pull requests merged into this card are reverted by the next publish, with no warning and no notification to whoever made them. The old text survives only in this repo's commit history.

So please don't send card fixes as pull requests here; they will not last. Open them against the template the wording comes from, scripts/hf/card_template.md, and they will appear at the next release. Nothing else on the Hub is affected: discussions persist, and only the Parquet files, their *_schema.json sidecars and this card are ever written or removed by the publishing job.

Links

• IDC portal -- browse the data and build cohorts interactively
• Visualizing IDC images -- how the browser viewers work; get a link to any series with IDCClient.get_viewer_URL()
• IDC agent interfaces -- search IDC, size a cohort and get a download command by asking: hosted MCP server, agent skill, or REST API
• IDC documentation
• idc-index Python package -- the download client (pip install idc-index)
• idc-index-data on GitHub -- how these tables are built (SQL included)
• highdicom Python package -- used to read DICOM and arrange as tensors
• GCS mirror of the release artifacts -- fetch a single file directly, e.g. https://storage.googleapis.com/idc-index-data-artifacts/current/release_artifacts/idc_index.parquet
• IDC user forum