Video-Based Temporal Attention Network for Automated Analysis of Left Ventricular Ejection Fraction

Q: What are the key findings of this study?

The Temporal Attention Network estimated left ventricular ejection fraction with a mean absolute error of 4.12%, representing an 11.3% improvement over previous state-of-the-art approaches.

Q: What is the clinical evidence from this study?

Population: Echocardiography for LVEF assessment (n=10030). Intervention: Temporal Attention Network (TAN) vs. Previous state-of-the-art approaches. Primary outcome: Mean absolute error (MAE) of LVEF estimation (MAE 4.12%, 95% CI 3.89%-4.35%).

Q: What does this research mean for the field?

A video-based temporal attention network enables automated analysis and estimation of left ventricular ejection fraction from echocardiogram videos. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Fazliddin Makhmudov; Jamshid Khamzaev; Jakhongir Karimberdiyev; Mekhriddin Rakhimov; Nigora Djurayeva; Islambek Saymanov; Abdulla Almuradov; Alpamis Kutlimuratov

doi:10.32604/cmes.2026.085457

January 1, 2026Computer Modeling in Engineering & SciencesOpen Access

Video-Based Temporal Attention Network for Automated Analysis of Left Ventricular Ejection Fraction

View Full Paper

Ask AI

Bookmark

Share

Key result

Temporal Attention Network improves LVEF estimation accuracy by ~11% over previous state-of-the-art models.

MAE 4.12%
95% CI 3.89%-4.35%
n=10,030

Why the study?

Does a Temporal Attention Network improve the precision and efficiency of automated LVEF estimation from echocardiography videos compared to previous methods?

Key Points

To develop and evaluate a deep learning framework combining spatial feature extraction, temporal modeling, and attention mechanisms for automated estimation of left ventricular ejection fraction from echocardiography videos.
Constructed a Temporal Attention Network integrating a ResNet-50 convolutional neural network, a bidirectional long short-term memory network with 512 hidden units per direction, and a Convolutional Block Attention Module.
Trained the network end-to-end on 10,030 apical four-chamber echocardiography videos from the EchoNet-Dynamic dataset using the Adam optimizer with cosine annealing.
Evaluated performance on an independent test set of 1,277 unseen echocardiography videos across diverse demographics and cardiac pathologies.
The model achieved a mean absolute error of 4.12% (95% CI: 3.89%–4.35%), a root mean squared error of 5.67%, and an R2 of 0.847 on 1,277 test videos, representing an 11.3% improvement over prior state-of-the-art methods.
For detecting heart failure with reduced ejection fraction (LVEF < 40%), the model attained 94.7% accuracy, 92.8% sensitivity, 95.6% specificity, and an AUROC of 0.977.
Ablation experiments showed temporal modeling reduced errors by 25% compared to frame-based methods and attention mechanisms provided an additional 16% improvement, with consistent accuracy across sexes (males: 4.18%, females: 4.05%, p = 0.51).

Structured PICO

Does a Temporal Attention Network improve the precision and efficiency of automated LVEF estimation from echocardiography videos compared to previous methods?

P

Population

10,030 patients aged 18 to 95 years undergoing clinical echocardiography, providing apical four-chamber videos for automated left ventricular ejection fraction assessment.

I

Intervention

Temporal Attention Network (TAN) integrating a ResNet-50-based convolutional neural network, a bidirectional long short-term memory (BiLSTM) network, and a Convolutional Block Attention Module (CBAM)

C

Comparator

Previous state-of-the-art approaches and frame-based methods

O

Outcome

Mean absolute error (MAE), root mean squared error (RMSE), and R2 of LVEF estimationsurrogate

Main Result

Effect estimate: MAE 4.12% (95% CI 3.89%-4.35%)

Absolute Event Rate: 4.12% vs 4.65%

The Temporal Attention Network provides highly accurate, automated LVEF estimation from echocardiography videos, achieving performance comparable to human experts and improving upon previous state-of-the-art AI methods.

Limitations

Performance degradation under poor image quality (MAE increases to 6.73%)
Relies on a single dataset (EchoNet-Dynamic) from one institution
Degradation of performance under poor image quality

Implication

May enhance automated LVEF estimation from echo videos; extends prior AI methods yet leaves open prospective validation before clinical use.

Abstract

: The left ventricular ejection fraction (LVEF) is one of the most important quantitative indicators in cardiovascular medicine. It serves as a key measure for diagnosing, assessing risk, and guiding treatment decisions in conditions such as heart failure, cardiomyopathies, and coronary artery disease. However, obtaining accurate LVEF values using echocardiography remains a major challenge. The process is often affected by differences between observers (8%–15%), time-consuming manual measurements that take several minutes per case, dependence on the operator’s skill

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Makhmudov et al. (2026) studied Echocardiography for LVEF assessment (n=10,030). Temporal Attention Network (TAN) vs. Previous state-of-the-art approaches was evaluated on Mean absolute error (MAE) of LVEF estimation (MAE 4.12%, 95% CI 3.89%-4.35%). The Temporal Attention Network estimated left ventricular ejection fraction with a mean absolute error of 4.12%, representing an 11.3% improvement over previous state-of-the-art approaches.

synapsesocial.com/papers/6abc576efbfca106a1247ec2 https://doi.org/10.32604/cmes.2026.085457

Topics

Artificial intelligence in cardiology

View Full Paper

Ask AI

Bookmark

Share

Video-Based Temporal Attention Network for Automated Analysis of Left Ventricular Ejection Fraction

Key Points

Structured PICO

Main Result

Limitations

Implication

Abstract

Discussion

Cite This Study

Topics

Video-Based Temporal Attention Network for Automated Analysis of Left Ventricular Ejection Fraction

Key Points

Structured PICO

Main Result

Limitations

Implication

Abstract

Discussion

Cite This Study

Keep reading

More related papers

More on Artificial intelligence in cardiology

More from Computer Modeling in Engineering & Sciences

Topics

Keep reading

More related papers

More on Artificial intelligence in cardiology

More from Computer Modeling in Engineering & Sciences