Deep learning interpretation of echocardiograms - PubMed Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2020 Jan 24:3:10.
doi: 10.1038/s41746-019-0216-8. eCollection 2020.

Deep learning interpretation of echocardiograms

Amirata Ghorbani #  1 , David Ouyang #  2 , Abubakar Abid  1 , Bryan He  3 , Jonathan H Chen  2 , Robert A Harrington  2 , David H Liang  2 , Euan A Ashley  2 , James Y Zou  1   3   4   5
Affiliations

Deep learning interpretation of echocardiograms

Amirata Ghorbani et al. NPJ Digit Med. .

Abstract

Echocardiography uses ultrasound technology to capture high temporal and spatial resolution images of the heart and surrounding structures, and is the most common imaging modality in cardiovascular medicine. Using convolutional neural networks on a large new dataset, we show that deep learning applied to echocardiography can identify local cardiac structures, estimate cardiac function, and predict systemic phenotypes that modify cardiovascular risk but not readily identifiable to human interpretation. Our deep learning model, EchoNet, accurately identified the presence of pacemaker leads (AUC = 0.89), enlarged left atrium (AUC = 0.86), left ventricular hypertrophy (AUC = 0.75), left ventricular end systolic and diastolic volumes ( R 2 = 0.74 and R 2 = 0.70), and ejection fraction ( R 2 = 0.50), as well as predicted systemic phenotypes of age ( R 2 = 0.46), sex (AUC = 0.88), weight ( R 2 = 0.56), and height ( R 2 = 0.33). Interpretation analysis validates that EchoNet shows appropriate attention to key cardiac structures when performing human-explainable tasks and highlights hypothesis-generating regions of interest when predicting systemic phenotypes difficult for human interpretation. Machine learning on echocardiography images can streamline repetitive tasks in the clinical workflow, provide preliminary interpretation in areas with insufficient qualified cardiologists, and predict phenotypes challenging for human evaluation.

Keywords: Cardiovascular diseases; Image processing; Machine learning.

PubMed Disclaimer

Conflict of interest statement

Competing InterestsThe authors declare no competing interests.

Figures

Fig. 1
Fig. 1. EchoNet machine learning pipeline for outcome prediction.
a EchoNet workflow for image selection, cleaning, and model training. b Comparison of model performance with different cardiac views as input. c Examples of data augmentation. The original frame is rotated (left to right) and its intensity is increase (top to bottom) as augmentations.
Fig. 2
Fig. 2. EchoNet performance and interpretation for three clinical interpretations of local structures and features.
For each task, representative positive examples are shown side-by-side with regions of interest from the respective model. Shaded areas indicate 95% confidence intervals.
Fig. 3
Fig. 3. EchoNet performance and interpretation for ventricular size and function.
EchoNet performance for a predicted left ventricular end systolic volume, b predicted end diastolic volume, c calculated ejection fraction from predicted ESV and EDV, and d predicted ejection fraction. e Input image, interpretation, and overlap for ejection fraction model.
Fig. 4
Fig. 4. EchoNet performance and interpretation for systemic phenotypes.
a EchoNet performance for prediction of four systemic phenotypes (sex, weight, height and age) using apical-4-chamber view images. Shaded areas indicate 95% confidence intervals. b Interpretation of systemic phenotype models with representative positive examples shown side-by-side with regions of interest.
Fig. 5
Fig. 5. Bland-Altman plotsBland-Altman plots of EchoNet performance for regression predictiontasks.
The solid black line indicates the median. Orange, red, and blue dashed lines delineate the central 50%, 75%, and 95% of cases based on differences between automated and measured values.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Heidenreich P, et al. Forecasting the future of cardiovascular disease in the united states: a policy statement from the american heart association. Circulation. 2011;123:933–944. doi: 10.1161/CIR.0b013e31820a55f5. - DOI - PubMed
    1. Cohen M, et al. Racial and ethnic differences in the treatment of acute myocardial infarction: findings from the get with the guidelines-coronary artery disease program. Circulation. 2010;121:2294–2301. doi: 10.1161/CIRCULATIONAHA.109.922286. - DOI - PubMed
    1. Havranek E, et al. Social determinants of risk and outcomes of cardiovascular disease a scientific statement from the american heart association. Circulation. 2015;132:873–898. doi: 10.1161/CIR.0000000000000228. - DOI - PubMed
    1. Madani A, Ong JR, Tiberwal A, Mofrad MR. US hospital use of echocardiography: Insights from the nationwide inpatient sample. J. Am. Coll. Cardiol. 2016;67:502–511. - PubMed
    1. Zhang J, et al. Fully automated echocardiogram interpretation in clinical practice: feasibility and diagnostic accuracy. Circulation. 2018;138:1623–1635. doi: 10.1161/CIRCULATIONAHA.118.034338. - DOI - PMC - PubMed