Overview

Schema Version

https://atlas.rsna.org/schemas/2025-11/dataset.json

Name

TICH-2 baseline noncontrast CT scans with ICH, PHE, and IVH annotations

Link

https://dx.doi.org/10.1148/ryai.220096

Indexing

Keywords: Head/Neck, Brain, Hemorrhage, Segmentation, Quantification, Convolutional Neural Network, Deep Learning, Machine Learning

Content: NR, CT

RadLex: RID49570, RID7540, RID16789, RID16633, RID10782, RID39452, RID4710, RID6476

Author(s)

Yong En Kok

Stefan Pszczolkowski

Zhe Kang Law

Azlinawati Ali

Kailash Krishnan

Philip M. Bath

Nikola Sprigg

Robert A. Dineen

Andrew P. French

Organization(s)

University of Nottingham

NIHR Nottingham Biomedical Research Centre

National University of Malaysia

Universiti Sultan Zainal Abidin

Nottingham University Hospitals NHS Trust

License

Text: Data available upon written request to the TICH-2 Chief Investigator; proposals assessed by the CI (and Steering Committee if required); Data Transfer Agreement required before sharing.

Contact

Data sharing contact: Nikola Sprigg (email presented in article: ku.ca.mahgnitton@ggirpS.alokiN); Corresponding author: Yong En Kok (ku.ca.mahgnitton@kok.gnoy)

Funding

UK National Institute for Health Research Health Technology Assessment Programme (project code 11_129_109).

Ethical review

Ethical approval from the UK Health Research Authority and relevant national or local institutional review boards; written informed consent obtained from participants or relatives before enrollment.

Comments

Baseline noncontrast head CT scans from the TICH-2 international multicenter clinical trial with expert annotations for intracerebral hemorrhage (ICH), perihematomal edema (PHE), and intraventricular hemorrhage (IVH); used to develop and evaluate deep learning segmentation models.

Date

Published: 2022-09-28

Created: 2022-09-28

References

[1] Kok YE, Pszczolkowski S, Law ZK, Ali A, Krishnan K, Bath PM, Sprigg N, Dineen RA, French AP. "Semantic Segmentation of Spontaneous Intracerebral Hemorrhage, Intraventricular Hemorrhage, and Associated Edema on CT Images Using Deep Learning". Radiology: Artificial Intelligence. 2022-09-28. doi:10.1148/ryai.220096. PMID: 36523645. PMCID: PMC9745441.

Dataset

Motivation

Enable automated, accurate segmentation and quantification of ICH, PHE, and IVH on CT for clinical trials and large cohort studies.

Sampling

1732 eligible TICH-2 participants with valid baseline scans from 124 participating centers; minimum requirement was axial orientation; any scanner manufacturer, settings, or section thickness included.

Partitioning scheme

Random split into training (90%) and test (10%) cohorts at the patient level.

Missing information

Image file formats, voxel resolutions, and scanner settings not detailed in the article.

Relationships between instances

Single baseline scan per included participant; lesions (ICH, PHE, IVH) annotated per scan.

Noise

Multicenter variability in scanners, acquisition settings, and section thickness; PHE boundaries noted as low-contrast and heterogeneous.

External data

No external datasets reported; all data from TICH-2 trial baseline noncontrast CT scans.

Confidentiality

Clinical trial imaging data with anonymized ground truth segmentations; shared only under Data Transfer Agreement upon approval by CI.

Sensitive data

Medical imaging data of patients with acute spontaneous ICH.