Artificial Intelligence-Enabled Echocardiographic Assessment of Right Ventricular Function

100%

0bo3-2026-01-18_12_45_43-tokodi-jimg112925-1937-t.pdf

Artificial Intelligence-Enabled Echocardiographic Assessment of Right Ventricular Function Running Title: AI-Enabled Echocardiographic Assessment of RV Function

Background: Right ventricular function is an important predictor of morbidity and mortality in various cardiovascular conditions. Nevertheless, its echocardiographic assessment is challenging due to its complex anatomy and location in the chest, resulting in limited inter-observer reproducibility.

Objectives: We aimed to develop a novel deep learning model - EchoNet-RV - to segment the RV in apical four-chamber view echocardiographic videos and estimate RV fractional area change.

Methods: For training EchoNet-RV, seven thousand one hundred sixty-nine expert-annotated apical four-chamber view echocardiographic videos were used. The model's performance was evaluated on a held-out internal test set of one thousand three hundred twenty apical four-chamber view videos and two international external test sets of three thousand one hundred seven and one thousand seventy-seven apical four-chamber view videos from two separate centers. Additionally, the associations between the predicted RV fractional area change values and the composite endpoint of heart failure hospitalization or all-cause death were also analyzed in the first external test set.

Results: EchoNet-RV segmented the RV with Dice coefficients of zero point eight nine three, zero point eight nine one to zero point eight nine five, and zero point seven nine seven to zero point seven nine eight and predicted RV fractional area change with mean absolute errors of five point seven nine five, five point five six zero to six point zero three one, five point eight three zero, five point six nine two to five point nine seven zero, and six point three six two, six point zero six four to six point six six zero percentage points in the held-out test set and the two external test sets, respectively. In five hundred randomly selected videos from the external test sets, EchoNet-RV's prediction error was significantly lower than the inter-observer variability. Moreover, it identified RV fractional area change less than thirty-five percent with areas under the receiver operating characteristic curve of zero point eight five nine, zero point eight four three to zero point eight seven six, zero point seven two five, zero point seven one zero to zero point seven four zero, and zero point six eight four, zero point six five three to zero point seven one three in the three test sets. EchoNet-RV also outperformed two multi-task models, EchoPrime and PanEcho, in estimating RV fractional area change and identifying RV dysfunction in the external test sets. In the first external test set, predicted RV fractional area change values were inversely associated with the composite endpoint, adjusted hazard ratio: zero point nine four eight, zero point nine one seven to zero point nine seven nine, independent of age, sex, cardiovascular risk factors, and left ventricular systolic function.

Conclusions: EchoNet-RV enables the rapid and automated assessment of RV fractional area change, with strong potential to become a valuable tool for the echocardiographic evaluation of RV function and disease surveillance.

CONDENSED ABSTRACT

In this study, we developed EchoNet-RV, an echocardiography-based deep learning model for automated right ventricle segmentation and right ventricular fractional area change estimation, and evaluated its performance on two international external datasets. EchoNet-RV demonstrated robust performance in right ventricle segmentation, right ventricular fractional area change estimation, and right ventricle dysfunction detection, with prediction errors significantly lower than inter-observer variability. It also outperformed two multi-task models, EchoPrime and PanEcho, in estimating right ventricular fractional area change and identifying right ventricle dysfunction. Moreover, the model's predictions were also associated with adverse clinical outcomes. EchoNet-RV enables rapid and automated right ventricular fractional area change assessment, with strong potential to become a valuable tool for the echocardiographic evaluation of right ventricle function and disease surveillance.

One hundred thirteen ABBREVIATIONS

INTRODUCTION

METHODS

Stanford dataset

Semmelweis dataset

MacKay Memorial Hospital dataset

Video annotation

Video preprocessing

Architecture and development of EchoNet-RV

Comparing prediction error with beat-to-beat and inter-observer variability

Benchmarking against multi-task models

Analyzing associations with outcomes

Performance metrics and statistical analysis

Ethical approval

RESULTS

Performance in RVFAC prediction

Benchmarking against multi-task models

Associations with outcomes

DISCUSSION

Limitations

CONCLUSIONS

Overview

The study presents EchoNet-RV, an AI-based model that automates the assessment of right ventricular function through echocardiography. It demonstrates superior performance compared to traditional methods, showing promise for improved clinical evaluation and monitoring of heart conditions.

Key Points

1EchoNet-RV automates right ventricular segmentation and fractional area change estimation
2The model outperformed traditional methods in multiple test sets
3It showed a strong correlation between predicted values and adverse clinical outcomes
4EchoNet-RV reduces inter-observer variability, enhancing clinical reliability
5The study supports AI's role in improving echocardiographic evaluations for heart health.

Details

Authors: Márton Tokodi, MD, PhD, Bryan He, BSc, PhD, Ádám Szijártó, MSc, Andrea Ferencz, MD, Kai Shiida, MD, Máté Tolvaj, MD, Alexandra Fábián, MD, PhD, Milos Vukadinovic, BSc, Béla Merkely, MD, PhD, DSc, Susan Cheng, MD, MMSc, MPH, Chung-Lieh Hung, MD, MSc, PhD, Attila Kovács, MD, PhD, DSc, David Ouyang, MD
Category: Health and Medicine

PDF
Endocrine Disorders
This document provides an overview of the endocrine system, detailing the anatomy and functions of various glands, hormones released, and diagnostic tests related to endocrine disorders.
PDF
NCM 116: CARE OF CLIENTS WITH NEUROLOGIC DISORDERS (ALTERATION IN PERCEPTION)
This document serves as an outline for a course on the care of clients with neurologic disorders, focusing on the nervous system's structure, functions, and endocrine disorders as they relate to patient care.
PDF
Fish Diseases and Their Management in Aquaculture
This document provides comprehensive insights into various fish diseases, their environmental conditions, and management strategies relevant to aquaculture. It covers issues related to temperature stratification, oxygen depletion, and different viral infections affecting fish health.
PDF
Anesthesia for Vascular Surgery
This document discusses the role of anesthesia in vascular surgery, particularly focusing on the challenges presented by atherosclerosis and its implications for perioperative care. It highlights the importance of preoperative evaluation and pharmacologic management to optimize outcomes for patients undergoing surgery for peripheral vascular disease.
PDF
Management of Patient with Neurologic Dysfunction
This document provides comprehensive guidelines on the management of patients with altered levels of consciousness, detailing definitions, causes, clinical manifestations, and nursing interventions required for care. It incorporates assessment protocols, diagnostic procedures, and potential complications associated with neurologic dysfunction.