Overview

Schema Version

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

Name

Northwestern Memorial Hospital noncontrast head CT intracranial hemorrhage (ICH) dataset (2008–2012)

Link

https://pubs.rsna.org/doi/10.1148/ryai.230296

Indexing

Keywords: intracranial hemorrhage, noncontrast head CT, weakly supervised learning, attention, bidirectional LSTM, Grad-CAM, study-level labels, CQ500, RSNA ICH

Content: CT, NR, ER

RadLex: RID28768

SNOMED: 1386000

Author(s)

Yunan Wu

Michael Iorga

Suvarna Badhe

James Zhang

Donald R. Cantrell

Elaine J. Tanhehco

Nicholas Szrama

Andrew M. Naidech

Michael Drakopoulos

Shamis T. Hasan

Kunal M. Patel

Tarek A. Hijaz

Eric J. Russell

Shamal Lalvani

Amit Adate

Todd B. Parrish

Aggelos K. Katsaggelos

Virginia B. Hill

Organization(s)

Northwestern University Feinberg School of Medicine

Northwestern Memorial Hospital

License

Text: Data generated by the authors or analyzed during the study are available upon request.

Contact

Virginia B. Hill, MD, Department of Radiology, Northwestern University Feinberg School of Medicine, 676 N St. Clair St, Ste 1400, Chicago, IL 60611, USA

Funding

Supported by the fiscal year 2020 Dixon Translational Research Grants Initiative and the Northwestern Memorial Foundation “Using Artificial Intelligence to Detect and Triage Emergencies on Noncontrast Head CTs.”

Ethical review

Approved by the Northwestern University institutional review board with waivers of Health Insurance Portability and Accountability Act authorization and informed consent.

Comments

Retrospective institutional dataset used for model fine-tuning and evaluation in a weakly supervised ICH detection study; all 2008–2012 noncontrast head CT scans at Northwestern Memorial Hospital were included regardless of patient history, image quality/artifacts, or scanner/technique.

Date

Published: 2024-08-28

Created: 2008-01-01

References

[1] Wu Y, Iorga M, Badhe S, et al.. "Precise Image-level Localization of Intracranial Hemorrhage on Head CT Scans with Deep Learning Models Trained on Study-level Labels". Radiology: Artificial Intelligence. 2024-11-01. doi:10.1148/ryai.230296. PMID: 39194400. PMCID: PMC11605431.

[2] . "RSNA Intracranial Hemorrhage Detection (pretraining dataset)". Kaggle. . Available from: https://kaggle.com/competitions/rsna-intracranial-hemorrhage-detection

[3] Chilamkurthy S, Ghosh R, Tanamala S, et al.. "Deep learning algorithms for detection of critical findings in head CT scans: a retrospective study". Lancet. 2018. PMID: 30318264.

Dataset

Motivation

Enable generalizable, weakly supervised detection and image-level localization of ICH to prioritize reading of positive noncontrast head CT studies while minimizing false-positive interruptions.

Sampling

All noncontrast head CT scans performed at Northwestern Memorial Hospital from 2008–2012 were included regardless of clinical indications, image quality, artifacts, scanner, or technique.

Partitioning scheme

Fivefold cross-validation on the training data (80% train, 20% validation within training folds); a fixed held-out local test set; an external independent test set (CQ500).

Missing information

Image file formats, pixel spacing/resolution, scanner manufacturers and acquisition parameters are not specified in the article body; additional details deferred to Appendix S1 (not included here).

Relationships between instances

Each study comprises sequential axial CT images; interimage (slice-to-slice) relationships leveraged via bidirectional LSTM and attention.

Noise

Images include variable quality and artifacts; scans from multiple scanners and techniques were included without exclusion.

External data

Model pretraining used the public RSNA ICH dataset; external generalizability evaluated on the public CQ500 dataset.

Confidentiality

Retrospective clinical imaging dataset from a single institution; IRB-approved with waivers of HIPAA authorization and informed consent.

Sensitive data

Clinical imaging data potentially containing PHI; handled under IRB approval with HIPAA waiver.