SciganticGuidesTopicsExplore

How to load CMMD mammography DICOM files and clinical data in a Jupyter notebook

Read the collection index with pandas (pd.read_parquet on series.parquet) and the images with pydicom (pydicom.dcmread). In a Scigantic notebook the collection tables and a sample mammography series are mounted at /mnt/archive, and each row of the index gives the public S3 location of its full series.

The code on this page runs in a Scigantic hosted notebook, where the dataset is mounted read-only at /mnt/archive. The same path is in the SCIGANTIC_MOUNT_PATH environment variable and linked from ~/archive. The public source is s3://scigantic-idc-open/cmmd/ (AWS us-east-1, public); full series in s3://idc-open-data.

Open IDC CMMD mammography on SciganticOpen the sample notebook: Mammograms joined to their clinical records

Code

Summarize the collection and read one mammogram

# CMMD: what is in the collection, and one real MG series from the sample folder
import os, glob, pandas as pd, numpy as np
M = os.environ.get('SCIGANTIC_MOUNT_PATH', '/mnt/archive')
series = pd.read_parquet(f'{M}/series.parquet')
print(series.groupby('Modality').agg(series=('SeriesInstanceUID', 'count'), patients=('PatientID', 'nunique'), GB=('series_size_MB', lambda s: round(s.sum() / 1024, 1))))
import pydicom
files = sorted(glob.glob(f'{M}/sample/MG_1.3.6.1.4.1.14519.5.2.1.1239.1759.817188338697704141156927137894/*.dcm'))
ds = pydicom.dcmread(files[0])
print(len(files), 'files;', ds.Modality, ds.get('Rows'), 'x', ds.get('Columns'), ds.get('ImageType'))

The starter cell from this dataset's schema card. It ran without error on 2026-10-08 in the Scigantic notebook image against the mounted data.

Read the patients table

import os
import pyarrow.parquet as pq
pf = pq.ParquetFile(os.path.join(os.environ['SCIGANTIC_MOUNT_PATH'], 'patients.parquet'))
print(pf.metadata)
df = pf.read_row_group(0).to_pandas()
df.head()

Run on 2026-10-08 in the Scigantic notebook image against the mounted data. It returned: Parquet 1,775 rows, 7 columns, 1 row groups; columns ['PatientID:large_string', 'PatientSex:large_string', 'PatientAge:large_string', 'n_studies:int64', 'n_series:int64', 'modalities:large_string', 'size_MB:double']

What the files are

CMMD is a breast cancer mammography collection in the NCI Imaging Data Commons (IDC): 1,775 patients, each with one study and one mammography series, 22.33 GB of images in total.

This archive holds Parquet tables that index the collection at patient, study and series level, the clinical table with its data dictionary, and one real mammography series of four DICOM files to test code against.

How the files are laid out

Path under the mountWhat it holds
series.parquet1,775 rows by 31 columns. One row per DICOM series, with modality, instance count, size, license and the s3://idc-open-data/... location of the files.
patients.parquet, studies.parquet1,775 rows each: patient-level and study-level summaries.
clinical/cmmd_clinical.parquet1,872 rows by 9 columns of clinical data; clinical/dictionary.parquet describes the columns.
sample/MG_<SeriesInstanceUID>/Four .dcm files, 17.6 MB: one complete mammography series.

Size, format and license

FormatParquet tables and DICOM
Size6 Parquet tables plus 4 sample DICOM files (18 MB). The full images stay in the IDC bucket.
LicenseCC BY 4.0 for all 1,775 series. Cite the source DOI 10.7937/tcia.eqde-4b16 and the IDC.
ProviderNCI Imaging Data Commons and the CMMD contributors; tables and sample by Scigantic
Public sources3://scigantic-idc-open/cmmd/ (AWS us-east-1, public); full series in s3://idc-open-data
Notebook imagePathology and medical imaging (pydicom and pandas preinstalled)

Things to know before you read it

Run it

Start a notebook from the IDC CMMD mammography archive page. The data is mounted when the notebook opens, so the code above runs as written and nothing is downloaded first. For a longer worked example, open the sample notebook Mammograms joined to their clinical records.

Related guides

Last checked 2026-11-21. All dataset guides.