CT-ICH (PhysioNet)
Computed Tomography Images for Intracranial Hemorrhage Detection and Segmentation
Head CT of patients with traumatic brain injury with radiologist masks of intracranial hemorrhage and per-slice labels for five hemorrhage types and skull fracture, for hemorrhage detection and segmentation. 82 patients from a hospital in Iraq; 75 scans are on PhysioNet.
Overview
CT-ICH is a small head CT collection for detecting and outlining intracranial hemorrhage after traumatic brain injury. It was gathered by researchers at Florida Atlantic University and the University of Technology in Baghdad together with radiologists in Babil, Iraq, and published on PhysioNet with a data descriptor that also reports a U-Net baseline. At release it was presented as the first public head CT set with hemorrhage masks.
Composition
The study covers 82 patients (46 male, 36 female) admitted to the emergency unit with a head injury, aged from one day to 72 years with a mean of 27.8; 27 were under 18. Each has one non-contrast CT scan of about 30 to 34 slices. 36 patients have a hemorrhage: intraventricular in 5, intraparenchymal in 16, subarachnoid in 7, epidural in 21 and subdural in 4, with some slices showing more than one type. 22 have a skull fracture. Slices without hemorrhage dominate: 2,173 slices are hemorrhage-free against a few dozen to a few hundred per hemorrhage type. The PhysioNet release contains CT volumes and masks for 75 patients; the volumes of patients 59 to 65 are missing, while the demographics and label tables list all 82.
Acquisition
Scans were collected retrospectively between February and August 2018 at Al Hilla Teaching Hospital on a Siemens SOMATOM Definition AS at 100 kV with 5 mm slices. The DICOM files were converted to NIfTI. Faces were blurred and overlaid with random noise for de-identification in version 1.3.1, which ships NIfTI files only.
Annotations
Two radiologists read each scan together without access to clinical history. After agreeing on the diagnosis, they recorded the hemorrhage types and fractures per slice and drew the hemorrhage regions on windowed images in a custom Matlab tool. The masks are stored as NIfTI. One scan with a chronic hemorrhage was excluded from the descriptor's study.
Known limitations
- Single hospital, single scanner and a young, trauma-only population.
- Some hemorrhage types appear in only a handful of patients, and only one consensus annotation exists per slice.
- Seven of the 82 CT volumes are missing from the release.
- The descriptor names CC BY 4.0, but PhysioNet distributes the data under its Restricted Health Data License.
Cohort
Aggregate numbers from the sources below. Bars are relative to the 82 subjects.
Sex
- Female 36 44%
- Male 46 56%
Contrast / sequence
Groups can overlap
- Non-contrast CT 82 100%
Section 3.1 in Hssayeni et al. 2020, Intracranial Hemorrhage Segmentation Using A Deep Convolutional Model (Data 5, 14)
Scanner vendor
- Siemens Healthineers 82 100%
Section 3.1 in Hssayeni et al. 2020, Intracranial Hemorrhage Segmentation Using A Deep Convolutional Model (Data 5, 14)
Country
- Iraq 82 100%
Section 3.1 in Hssayeni et al. 2020, Intracranial Hemorrhage Segmentation Using A Deep Convolutional Model (Data 5, 14)
Age
mean 27.8 ± 19.5, up to 72
Condition
Groups can overlap
- Traumatic brain injury 82 100%
- Intracranial hemorrhage 36 44%
- Bone fracture 22 27%
- Intracerebral hemorrhage 16 20%
Section 3.1, Table 4 in Hssayeni et al. 2020, Intracranial Hemorrhage Segmentation Using A Deep Convolutional Model (Data 5, 14)
Contrast combinations
How many subjects have exactly each set of contrasts.
| ct_noncontrast | Subjects with exactly this set |
|---|---|
82 |
Section 3.1 in Hssayeni et al. 2020, Intracranial Hemorrhage Segmentation Using A Deep Convolutional Model (Data 5, 14)
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
Sign a data use agreement, often reviewed by the provider
PhysioNet account and the PhysioNet Restricted Health Data Use Agreement 1.5.0, signed online.
PhysioNet Restricted Health Data License 1.5.0
Research-only license for registered PhysioNet users who sign the matching data use agreement online. Unlike the credentialed license it asks for no identity check or human-subjects training. You must not share the data or try to identify people, and you must release the code behind your publications.
PhysioNet lists the PhysioNet Restricted Health Data License 1.5.0 and the matching data use agreement. The data descriptor in Data (2020) instead names the Creative Commons Attribution 4.0 license for version 1.3.0; the PhysioNet project page, which grants access, is taken as authoritative.
What you can do
- Not stated
- Not stated
- Not stated
- Create derived data Not stated
- Conditional
What you can share
- No
- Not stated
- Share trained models Not stated
What you must do
- Not stated
- Share alike No
- Yes
- Ethics approval No
- Manuscript review No
- Yes
- No
- No
Limits
- Yes
- Location limits No
Citation
Hssayeni MD, Croock MS, Salman AD, Al-khafaji HF, Yahya ZA, Ghoraani B. Intracranial Hemorrhage Segmentation Using A Deep Convolutional Model. Data 5(1), 14 (2020). doi:10.3390/data5010014. Hssayeni M. Computed Tomography Images for Intracranial Hemorrhage Detection and Segmentation (version 1.3.1). PhysioNet (2020). doi:10.13026/4nae-zg36
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.