Skip to content
Open Imaging Index
MRI Brain

BONBID-HIE

BOston Neonatal Brain Injury Data for Hypoxic Ischemic Encephalopathy (BONBID-HIE): I. MRI and Manual Lesion Annotation

Neonatal brain diffusion MRI from 133 term-born infants with hypoxic ischemic encephalopathy at Massachusetts General Hospital: skull-stripped ADC maps, ZADC deviation maps and expert lesion masks, used in the MICCAI 2023 BONBID-HIE lesion segmentation challenge.

Go to dataset Edit Report

Overview

BONBID-HIE Part I is a public set of neonatal brain diffusion MRI with manual lesion masks for hypoxic ischemic encephalopathy (HIE), released by researchers at Boston Children's Hospital, Harvard Medical School and Massachusetts General Hospital. It served as the data of the first BONBID-HIE lesion segmentation challenge at MICCAI 2023. HIE lesions in this cohort tend to be small and spread across several regions, which makes the set a test case for segmenting small diffuse lesions.

Composition

The cohort has 133 term-born infants (74 male, 59 female) with a clinical diagnosis of HIE, treated at Massachusetts General Hospital between 2001 and 2018 and scanned on postnatal day 0 to 14 (mean 3.9 days). Each patient has a skull-stripped apparent diffusion coefficient (ADC) map, a ZADC map that expresses each voxel as standard deviations from a normative neonatal ADC atlas, and a binary lesion mask. A clinical spreadsheet with maternal and neonatal variables and a normal and a lesion atlas are included. The challenge split is 85 training, 4 validation and 44 test cases. The paper reports that the median lesion covers 0.63% of the brain and 74 patients have lesions below 1% of brain volume.

Acquisition

Diffusion MRI came from the clinical archive: 52 patients on a GE 1.5 T Signa scanner (2001 to 2012) and 81 on a Siemens 3 T TrioTim or PrismaFit (2012 to 2018), with b = 1000 s/mm². ADC maps were computed on the scanners. Preprocessing covered bias correction, skull stripping and DRAMMS registration to the atlas.

Annotations

A physician with more than three years of experience drew the lesion masks on the 3D ADC maps in native space, guided by the clinical neuroradiology reports. Uncertain cases (27 patients) were settled by consensus of three pediatric neuroradiologists.

Known limitations

Single-center, retrospective data with only diffusion-derived maps; no T1- or T2-weighted images. Race is unknown for 57 of 133 mothers. Some mild cases have no visible lesion (minimum lesion volume 0 mm³). The Zenodo files are in MetaImage format, while the paper describes NIfTI. License statements conflict: the Zenodo license field gives CC BY-NC-ND 2.5, its description text says CC BY 4.0, and the paper and challenge page say CC BY-NC-ND; this entry records the NoDerivs license.

Cohort

Aggregate numbers from the sources below. Bars are relative to the 133 subjects.

Sex

  • Male 74 56%
  • Female 59 44%

Table 1A in Bao et al. 2023, bioRxiv preprint 2023.06.30.546841 (v1)

Age

0
50
100
133
0-0.5

Retrospective Data Collection in Bao et al. 2023, bioRxiv preprint 2023.06.30.546841 (v1)

Contrast combinations

How many subjects have exactly each set of contrasts.

ADCSubjects with exactly this set
133

Data Records in Bao et al. 2023, bioRxiv preprint 2023.06.30.546841 (v1)

Contrast / sequence

subjects, values can overlap

  • ADC map 133 100%

Data Records in Bao et al. 2023, bioRxiv preprint 2023.06.30.546841 (v1)

Condition

subjects, values can overlap

  • Hypoxic ischemic encephalopathy 133 100%

Table 1 in Bao et al. 2023, bioRxiv preprint 2023.06.30.546841 (v1)

Scanner vendor

subjects

  • Siemens Healthineers 81 61%
  • GE HealthCare 52 39%

Table 1D in Bao et al. 2023, bioRxiv preprint 2023.06.30.546841 (v1)

Field strength

subjects

  • 3 T 81 61%
  • 1.5 T 52 39%

Table 1D in Bao et al. 2023, bioRxiv preprint 2023.06.30.546841 (v1)

Country

subjects

  • United States 133 100%

Retrospective Data Collection in Bao et al. 2023, bioRxiv preprint 2023.06.30.546841 (v1)

Split

subjects

  • Training 85 64%
  • Test 44 33%
  • Validation 4 3%

Data in BONBID-HIE 2023 challenge data page (grand-challenge.org)

License and access

Our reading of the license, not legal advice. Before you use the data, read the original license and confirm that your use is allowed. We take no responsibility for how you use a dataset. Full disclaimer

Access
Open download

Download without an account

Access page

Creative Commons Attribution-NonCommercial-NoDerivs 2.5 Generic

Older version of CC BY-NC-ND. Share the unchanged data with credit for non-commercial purposes. The legal code grants no right to make derivative works, and Creative Commons advises against training models on NoDerivatives material.

Original license text Version read: 2.5 Checked 2026-10-09

What you can do

  • No
  • Not stated
  • No
  • No
  • Yes

What you can share

  • Conditional
  • No
  • No

What you must do

  • Yes
  • Share alike No
  • No
  • No
  • Manuscript review No
  • Release code No
  • Return results No
  • Delete after use No

Limits

  • No
  • Location limits No

Citation

Bao R, Song Y, Bates SV, Weiss RJ, Foster AN, Jaimes Cobos C, Sotardi S, Zhang Y, Gollub RL, Grant PE, Ou Y. BOston Neonatal Brain Injury Dataset for Hypoxic Ischemic Encephalopathy (BONBID-HIE): Part I. MRI and Manual Lesion Annotation. bioRxiv 2023.06.30.546841 (2023).

@article{bao2023boston,
  title={BOston Neonatal Brain Injury Dataset for Hypoxic Ischemic Encephalopathy (BONBID-HIE): Part I. MRI and Manual Lesion Annotation},
  author={Bao, Rina and Song, Ya'nan and Bates, Sara V and Weiss, Rebecca J and Foster, Anna N and Cobos, Camilo Jaimes and Sotardi, Susan and Zhang, Yue and Gollub, Randy L and Grant, P Ellen and others},
  journal={bioRxiv},
  year={2023},
  publisher={Cold Spring Harbor Laboratory Preprints}
}

Sources

Every number on this page comes from one of these documents. Each chart names the table or page it is taken from. The raw numbers are in stats.csv.