Real-time Echocardiography Video Segmentation via Slot Propagation, Spatiotemporal Feature Fusion, and Frequency-phase Enhancement

ICLR 2026 Conference Submission8219 Authors

Keywords: echocardiography segmentation, slot propagation, semi-supervised learning

Abstract:

Accurate and real-time segmentation of cardiac structures in echocardiography videos is crucial for the diagnosis and treatment of heart disease. However, it is a very challenging task due to low imaging quality, speckle noise, ambiguous boundaries, and incomplete annotations.The real-time demand in clinical settings make this task even more difficult.
In this paper, we propose an novel echocardiography video segmentation model to more comprehensively address these challenges than existing solutions.The core innovative techniques in our model include (1) a slot propagation mechanism to capture target-specific information to improve our model's ability to distinguish our targets from noisy background, (2) a spatiotemporal feature fusion algorithm to tackle dramatic shape changes across frames, and (3) a frequency-phase enhancement module to extract more independent and distinct semantic patterns amid severe speckle noise and artifacts.
We extensively evaluate our method on two representative datasets and the results demonstrate that our model achieves better segmentation precision than existing models by a considerable margin, while maintaining real-time performance.
Codes will be publicly available upon publication.

Primary Area: applications to computer vision, audio, language, and other modalities

Submission Number: 8219