ProtoEFNet: Dynamic Prototype Learning for Inherently Interpretable Ejection Fraction Estimation in Echocardiography

2025-12-03

AI Summary

Traditional ejection fraction (EF) assessment relies on manual endocardial contour delineation, which is time-consuming and exhibits substantial inter-observer variability; existing deep learning methods are predominantly black-box models, and their post-hoc explanations lack clinical interpretability for decision support. To address this, we propose a dynamic spatiotemporal prototype learning framework that employs a Prototype-Aware Separation loss (PAS) to enable inherently interpretable modeling of continuous EF values—directly linking model predictions to clinically meaningful cardiac motion patterns. Our method operates end-to-end on echocardiographic video inputs with fully differentiable training. On the Echonet-Dynamic dataset, it achieves state-of-the-art accuracy comparable to non-interpretable models (F1 = 79.64 ± 2.10), while providing intuitive, verifiable prototype-level explanations. This substantially enhances clinical trustworthiness and adoption potential.

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Calibration & Uncertainty Quantification
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Ejection fraction (EF) is a crucial metric for assessing cardiac function and diagnosing conditions such as heart failure. Traditionally, EF estimation requires manual tracing and domain expertise, making the process time-consuming and subject to interobserver variability. Most current deep learning methods for EF prediction are black-box models with limited transparency, which reduces clinical trust. Some post-hoc explainability methods have been proposed to interpret the decision-making process after the prediction is made. However, these explanations do not guide the model's internal reasoning and therefore offer limited reliability in clinical applications. To address this, we introduce ProtoEFNet, a novel video-based prototype learning model for continuous EF regression. The model learns dynamic spatiotemporal prototypes that capture clinically meaningful cardiac motion patterns. Additionally, the proposed Prototype Angular Separation (PAS) loss enforces discriminative representations across the continuous EF spectrum. Our experiments on the EchonetDynamic dataset show that ProtoEFNet can achieve accuracy on par with its non-interpretable counterpart while providing clinically relevant insight. The ablation study shows that the proposed loss boosts performance with a 2% increase in F1 score from 77.67$pm$2.68 to 79.64$pm$2.10. Our source code is available at: https://github.com/DeepRCL/ProtoEF

Problem

• Automates ejection fraction estimation from echocardiography videos.
• Enhances model interpretability with dynamic cardiac motion prototypes.
• Improves clinical trust through inherent explainability in EF regression.

Innovation

• Dynamic spatiotemporal prototypes capture cardiac motion patterns
• Prototype Angular Separation loss enforces discriminative continuous representations
• Video-based prototype learning achieves accuracy with clinical interpretability