Towards automatic diagnosis of rheumatic heart disease on echocardiographic exams through video-based deep learning - 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
. 2021 Aug 13;28(9):1834-1842.
doi: 10.1093/jamia/ocab061.

Towards automatic diagnosis of rheumatic heart disease on echocardiographic exams through video-based deep learning

João Francisco B S Martins  1 , Erickson R Nascimento  1 , Bruno R Nascimento  2 , Craig A Sable  3 , Andrea Z Beaton  4 , Antônio L Ribeiro  2 , Wagner Meira  1 , Gisele L Pappa  1
Affiliations

Towards automatic diagnosis of rheumatic heart disease on echocardiographic exams through video-based deep learning

João Francisco B S Martins et al. J Am Med Inform Assoc. .

Abstract

Objective: Rheumatic heart disease (RHD) affects an estimated 39 million people worldwide and is the most common acquired heart disease in children and young adults. Echocardiograms are the gold standard for diagnosis of RHD, but there is a shortage of skilled experts to allow widespread screenings for early detection and prevention of the disease progress. We propose an automated RHD diagnosis system that can help bridge this gap.

Materials and methods: Experiments were conducted on a dataset with 11 646 echocardiography videos from 912 exams, obtained during screenings in underdeveloped areas of Brazil and Uganda. We address the challenges of RHD identification with a 3D convolutional neural network (C3D), comparing its performance with a 2D convolutional neural network (VGG16) that is commonly used in the echocardiogram literature. We also propose a supervised aggregation technique to combine video predictions into a single exam diagnosis.

Results: The proposed approach obtained an accuracy of 72.77% for exam diagnosis. The results for the C3D were significantly better than the ones obtained by the VGG16 network for videos, showing the importance of considering the temporal information during the diagnostic. The proposed aggregation model showed significantly better accuracy than the majority voting strategy and also appears to be capable of capturing underlying biases in the neural network output distribution, balancing them for a more correct diagnosis.

Conclusion: Automatic diagnosis of echo-detected RHD is feasible and, with further research, has the potential to reduce the workload of experts, enabling the implementation of more widespread screening programs worldwide.

Keywords: deep learning; echocardiography; low-cost imaging; meta-learning; rheumatic heart disease; screening.

PubMed Disclaimer

Figures

Figure 1.
Figure 1.
Echocardiographic images from an 18-year-old boy with definite RHD. A shows a > 2 cm jet of mitral regurgitation in parasternal long axis Doppler view; B and C show a > 2 cm jet of aortic insufficiency (yellow arrows) in parasternal long axis Doppler and apical 4 chamber Doppler views. Ao, aorta; LA, left atrium; LV, left ventricle.
Figure 2.
Figure 2.
Examples of echocardiogram viewpoints present in our dataset. Images were sampled from different videos of a single exam. (a) Parasternal Long Axis; (b) Parasternal Long Axis with Doppler on the Mitral Valve Level; (c) Parasternal Long Axis with Doppler on the Aortic Valve Level; (d) Apical 4 Chambers; (e) Apical 4 Chambers with Doppler; (f) Apical 5 Chambers; (g) Apical 5 Chambers with Doppler.
Figure 3.
Figure 3.
C3D network architecture for video classification.
Figure 4.
Figure 4.
Proposed supervised meta-classifier for result aggregation toward exam classification.
Figure 5.
Figure 5.
Resulting confusion matrices for each method on RHD classification of echocardiographic exams. (a) VGG16 with Majority Vote; (b) C3D with Majority Vote ;(c) C3D with Meta-Classifier.
Figure 6.
Figure 6.
Examples of frames extracted from 4 videos where the model made the predictions with high confidence. Videos are from different exams, and we consider their predictions when in the test set. (a) RHD negative misclassified as RHD positive; (b) RHD positive misclassified as RHD negative; (c) RHD negative correctly classified; (d) RHD positive correctly classified.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Roth GA, Abate D, Abate KH, et al.Global, regional, and national age-sex-specific mortality for 282 causes of death in 195 countries and territories, 1980–2017: a systematic analysis for the global burden of disease study 2017. Lancet 2018; 392 (10159): 1736–88. - PMC - PubMed
    1. Davis AM, Vinci LM, Okwuosa TM, et al.Cardiovascular health disparities. Med Care Res Rev 2007; 64 (5 Suppl): 29S–100S. - PMC - PubMed
    1. Cohen MG, Fonarow GC, Peterson ED, et al.Racial and ethnic differences in the treatment of acute myocardial infarction. Circulation 2010; 121 (21): 2294–301. - PubMed
    1. James SL, Abate D, Abate KH, et al.Global, regional, and national incidence, prevalence, and years lived with disability for 354 diseases and injuries for 195 countries and territories, 1990–2017: a systematic analysis for the global burden of disease study 2017. Lancet 2018; 392 (10159): 1789–858. - PMC - PubMed
    1. Mondo C, Musoke C, Kayima J, et al.Presenting features of newly diagnosed rheumatic heart disease patients in Mulago Hospital: a pilot study. Cardiovasc J Afr 2013; 24 (2): 28–33. - PMC - PubMed

Publication types