A dual‐mechanism framework for imbalanced right ventricular ejection fraction estimation from 2D echocardiography
Medical Physics · 2026-12 · 53 (12)
DOI: 10.1002/mp.70759 ↗At a glance
Journal article in Medical Physics, published online on 2026-11-19. Written by Suyu Dong and 6 co-authors. 44 references. Funded by: National Key Research and Development Program of China, National Natural Science Foundation of China.
Abstract (original language)
Background
Right ventricular ejection fraction (RVEF) is a clinically essential indicator of right ventricular systolic function and a well‐established predictor of outcomes in cardiovascular disease. Reduced RVEF is strongly correlated with adverse clinical events such as heart failure progression, recurrent hospitalization, and increased mortality. Accurate and reliable estimation of RVEF is therefore crucial for risk assessment and guidance of treatment strategies.
Purpose
Addressing the problem of imbalanced regression in RVEF estimation, we introduce a dual‐mechanism framework that improves both robustness and generalization when learning from two‐dimensional echocardiography (2DE) videos.
Methods
Our approach integrates Gaussian Kernel Constrained Mixup, which expands and smooths the training distribution via structured sample synthesis, and Feature Space Calibration, which imposes ordinal constraints on the latent representation using weighted contrastive learning. This combined strategy enhances both data efficiency and feature‐label consistency across the entire value range. The well‐trained model can be seamlessly integrated into clinical workflows to assist physicians in rapid patient screening. It enables efficient, high‐throughput analysis of echocardiographic videos, providing reliable RVEF estimates that help identify high‐risk patients at an early stage and support large‐scale preliminary screening in resource‐limited settings.
Results
On the RVENet dataset comprising 3583 echocardiographic videos, our method achieved state‐of‐the‐art performance with a mean absolute error (MAE) of and an of 0.67. Evaluation on an independent external test cohort of 128 videos demonstrated generalization, with an MAE of . Stratified analysis confirmed consistent performance across diverse RVEF ranges, indicating robustness to imbalanced data distribution.
Conclusions
The proposed framework effectively tackles data imbalance and regression bias in RVEF estimation through dual mechanisms that improve both sample diversity and feature discriminability. It exhibits good accuracy, generalizability, and balanced performance across the RVEF spectrum. The implementation of our approach is publicly available at https://github.com/Wnssl/DIRVEF .
Funding
- National Key Research and Development Program of China
- National Natural Science Foundation of China
Cite
APA
Dong, S., Wang, N., Sun, Y., Hu, X., Zhao, Y., Wang, G., & Luo, G. (2026). A dual‐mechanism framework for imbalanced right ventricular ejection fraction estimation from 2D echocardiography. Medical Physics, 53(12). https://doi.org/10.1002/mp.70759
Vancouver
Dong S, Wang N, Sun Y, Hu X, Zhao Y, Wang G, et al. A dual‐mechanism framework for imbalanced right ventricular ejection fraction estimation from 2D echocardiography. Medical Physics. 2026;53(12). doi:10.1002/mp.70759
BibTeX
@article{Dong2026,
author = {Dong, Suyu and Wang, Nan and Sun, Yixin and Hu, Xiaoxiao and Zhao, Yuming and Wang, Guohua and Luo, Gongning},
title = {A dual‐mechanism framework for imbalanced right ventricular ejection fraction estimation from 2D echocardiography},
journal = {Medical Physics},
year = {2026},
volume = {53},
number = {12},
doi = {10.1002/mp.70759}
}
EUVIMED