![]() ![]() This results in high variance and limited precision with inter-observer variation ranging from 7.6% to 13.9% 4, 12- 15. ![]() While the American Society of Echocardiography and the European Association of Cardiovascular Imaging guidelines recommend tracing and averaging up to 5 consecutive cardiac cycles if variation is identified, EF is often evaluated from tracings of only one representative beat or visually approximated if a tracing is deemed inaccurate 5, 15. Human assessment of EF has variance in part due to the common finding of irregularity in the heart rate and the laborious nature of a calculation that requires manual tracing of ventricle size to quantify every beat 4, 5. However, echocardiography is associated with significant inter-observer variability as well as inter-modality discordance based on methodology and modality 2, 4, 5, 11- 14. In particular, left ventricular ejection fraction (EF), the ratio of change in left ventricular end systolic and end diastolic volume, is one of the most important metrics of cardiac function, as it identifies patients who are eligible for life prolonging therapies 7, 11. A variety of methodologies have been used to quantify cardiac function and diagnose dysfunction. Impairment of cardiac function is described as “cardiomyopathy” or “heart failure” and is a leading cause of hospitalization in the United States and a growing global health issue 1, 9, 10. As a new resource to promote further innovation, we also make publicly available the largest medical video dataset of 10,030 annotated echocardiogram videos.Ĭardiac function is essential for maintaining normal systemic tissue perfusion with cardiac dysfunction manifesting as dyspnea, fatigue, exercise intolerance, fluid retention and mortality 1, 2, 3, 5- 8. By leveraging information across multiple cardiac cycles, our model can rapidly identify subtle changes in ejection fraction, is more reproducible than human evaluation, and lays the foundation for precise diagnosis of cardiovascular disease in real-time. Prospective evaluation with repeated human measurements confirms that the model has comparable or less variance than human experts. In an external dataset from another healthcare system, EchoNet-Dynamic predicts ejection fraction with mean absolute error of 6.0% and classifies heart failure with reduced ejection fraction with an AUC of 0.96. Trained on echocardiogram videos, our model accurately segments the left ventricle with a Dice Similarity Coefficient of 0.92, predicts ejection fraction with mean absolute error of 4.1%, and reliably classifies heart failure with reduced ejection fraction (AUC of 0.97). To overcome this challenge, we present the first video-based deep learning algorithm, EchoNet-Dynamic, that surpasses human expert performance in the critical tasks of segmenting the left ventricle, estimating ejection fraction, and assessing cardiomyopathy. However human assessment of cardiac function focuses on a limited sampling of cardiac cycles and has significant inter-observer variability despite years of training 4, 5. ![]() Accurate assessment of cardiac function is crucial for diagnosing cardiovascular disease 1, screening for cardiotoxicity 2, and deciding clinical management in patients with critical illness 3. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |