MITEA
MITEA (MR-Informed Three-dimensional Echocardiography Analysis)
3D transthoracic echocardiography of 134 adults from Auckland (82 healthy, 52 with cardiac disease), scan and rescan at end-diastole and end-systole, with left ventricular myocardium and cavity labels registered from each subject's own cardiac MRI. 536 labelled volumes.
Overview
MITEA is a set of labelled 3D echocardiograms of the left ventricle, made at the Auckland Bioengineering Institute of the University of Auckland and shared through the Cardiac Atlas Project. Each participant also had a cardiac MRI on the same day, and the ventricle shapes traced on MRI were aligned to the echo volumes. The labels therefore follow the MRI anatomy rather than an observer's reading of the ultrasound. The dataset is meant for training and testing automatic segmentation of the myocardium and cavity and for measuring volumes, ejection fraction and mass from 3D echo.
Composition
144 adults were recruited and 134 kept after 10 were dropped for poor echo quality. Of these, 82 are healthy controls and 52 have acquired cardiac disease: left ventricular hypertrophy, cardiac amyloidosis, aortic regurgitation, hypertrophic or dilated cardiomyopathy, and two heart transplant recipients. Ages range from 18 to 84 (mean 47) and 81 participants are male. Every participant has two echo clips, scan and rescan in random order, and each clip is given at end-diastole and end-systole, for 536 labelled volumes. The paper's own experiment grouped both clips of a person together and used 107 participants for training and 27 for testing.
Acquisition
Transthoracic single-beat 3D echo was recorded from the apical window during breath-holds on a Siemens ACUSON SC2000 with a 4Z1c matrix probe. An experienced sonographer tuned depth, gain, sector width and harmonic settings per person. Volumes were resampled to Cartesian grids with 1 mm isotropic voxels, with zeros outside the imaging pyramid. The paired cine MRI used Siemens 1.5 T and 3 T scanners and was done within two hours of the echo; the MRI images are not part of the release.
Annotations
One analyst built left ventricular models from the MRI with guide-point modelling, counting papillary muscles and trabeculations as cavity. After an automatic rough alignment, the same person adjusted each model by hand inside the echo volume at both frames. Masks hold myocardium (value 1) and cavity (value 2).
Known limitations
- One centre, one ultrasound vendor and one observer for labels and alignment.
- Labels assume the ventricle has the same shape in both scans, although posture, breath-hold and heart rate differ.
- Parts of the labelled ventricle can lie outside the acquired pyramid.
- Only end-diastole and end-systole are labelled.
Cohort
Aggregate numbers from the sources below. Bars are relative to the 134 subjects.
Sex
- Male 81 100%
Covers 81 of 134 subjects.
Age
mean 47± 20 · range 18 to 84No age bins reported.
Modality
subjects
- Ultrasound 134 100%
Condition
subjects, values can overlap
- Healthy control 82 61%
- Left ventricular hypertrophy 14 10%
- Cardiac amyloidosis 12 9%
- Aortic regurgitation 10 7%
- Hypertrophic cardiomyopathy 8 6%
- Dilated cardiomyopathy 6 4%
Scanner vendor
subjects
- Siemens Healthineers 134 100%
Country
subjects
- New Zealand 134 100%
Split
subjects
- Training 107 80%
- Test 27 20%
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
A research proposal is reviewed and approved
Different parts of the data carry different licenses. The summary on the right shows the most restrictive answer per rule.
Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International
Non-commercial use only, with credit. Anything you share that is built on the data must use the same license.
The dataset page states that MITEA can be provided under CC BY-NC-SA 4.0 for research purposes and that commercial or other use needs contact with the data contributors. The same page's intellectual property notice says no part of the dataset may be copied, distributed or disclosed to others without consent.
What you can do
- No
- Conditional
- Yes
- Yes
What you can share
- Conditional
- Conditional
- Conditional
What you must do
- Yes
- Yes
- No
- No
- Manuscript review No
- Release code No
- Return results No
- Delete after use No
Limits
- No
- Location limits No
Cardiac Atlas Project Policies and Procedures for Data Distribution to Users
Users submit a research project to the CAP Steering Committee and each contributing study, then sign a Data Distribution Agreement per study. Data may be used only for that project, not passed on without CAP approval, not used commercially without the contributors' permission, and are deleted when the project ends.
Access is requested through an online form that describes the intended research and data handling. The requester must agree to the CAP Policy Statements, not to share the data outside their research group and to use it only for the purpose stated in the request. Institutional email addresses are required.
What you can do
- Conditional
- Not stated
- Conditional
- Conditional
What you can share
- Conditional
- Conditional
- Share trained models Not stated
What you must do
- Yes
- Share alike No
- Yes
- No
- Conditional
- Release code No
- Yes
- Yes
Limits
- Yes
- No
Citation
Zhao D, Ferdian E, Maso Talou GD, Quill GM, Gilbert K, Wang VY, Babarenda Gamage TP, Pedrosa J, D'hooge J, Sutton TM, Lowe BS, Legget ME, Ruygrok PN, Doughty RN, Camara O, Young AA, Nash MP. MITEA: A dataset for machine learning segmentation of the left ventricle in 3D echocardiography using subject-specific labels from cardiac magnetic resonance imaging. Front Cardiovasc Med 9, 1016703 (2023). https://doi.org/10.3389/fcvm.2022.1016703
All numbers
Every number on this page, as stored in stats.csv, with its source.
| Measure | Breakdown | Value | Source |
|---|---|---|---|
| Subjects | total 144 recruited; 10 excluded for poor 3D echo image quality | 134 | zhao2023 Section 2 |
| Subjects | modality=US Transthoracic real-time 3D echocardiography | 134 | zhao2023 Section 2.1 |
| Subjects | vendor=siemens Siemens ACUSON SC2000 with 4Z1c transducer | 134 | zhao2023 Section 2.1 |
| Subjects | country=NZ Auckland, New Zealand | 134 | zhao2023 Section 2 |
| Subjects | condition=healthy | 82 | zhao2023 Table 1 |
| Subjects | condition=left_ventricular_hypertrophy | 14 | zhao2023 Section 3.1 |
| Subjects | condition=cardiac_amyloidosis | 12 | zhao2023 Section 3.1 |
| Subjects | condition=aortic_regurgitation | 10 | zhao2023 Section 3.1 |
| Subjects | condition=hypertrophic_cardiomyopathy | 8 | zhao2023 Section 3.1 |
| Subjects | condition=dilated_cardiomyopathy | 6 | zhao2023 Section 3.1 |
| Subjects | sex=male | 81 | zhao2023 Table 1 |
| Subjects | sex=male;condition=healthy | 42 | zhao2023 Table 1 |
| Subjects | split=train Split used in the paper's nnU-Net experiment | 107 | zhao2023 Section 2.4 |
| Subjects | split=test Split used in the paper's nnU-Net experiment | 27 | zhao2023 Section 2.4 |
| Scans | total 3D image volumes: 268 clips (scan and rescan per subject), each at end-diastole and end-systole | 536 | zhao2023 Section 2.3 |
| Mean age | total | 47 | zhao2023 Table 1 |
| Age SD | total | 20 | zhao2023 Table 1 |
| Minimum age | total | 18 | zhao2023 Table 1 |
| Maximum age | total | 84 | zhao2023 Table 1 |
Sources
The keys used in the table above.
- zhao2023 Zhao et al. 2023, Frontiers in Cardiovascular Medicine (MITEA data descriptor) paper
- cap-mitea Cardiac Atlas Project, MITEA page website