Automatic Echocardiography Segmentation via Transition Probability Correlation for Stable Semantic Extraction

2026-07-08

AI Summary

This study addresses the challenges of semantic ambiguity and fragmented boundaries in echocardiographic images caused by speckle noise and low signal-to-noise ratio. To this end, the authors propose a convolutional network that integrates local inductive bias with long-range dependencies. The core innovations include a Semantic-Texture Local Similarity Fusion (STLSF) module, which models local transition probability correlations for semantic refinement and employs semantic guidance to enhance texture details, and a frequency-aware denoising pretraining strategy that enables the encoder to better adapt to ultrasound imaging priors. Evaluated on the CAMUS and EchoNet-Dynamic datasets, the method achieves Dice scores of 93.87% and 92.62%, with Hausdorff distances (HD95) of 3.29 mm and 2.73 mm, respectively, outperforming current state-of-the-art approaches.

Abstract

While echocardiography is essential for cardiovascular diagnosis, inherent speckle noise and low signal-to-noise ratio often lead to ambiguous semantic features and fragmented boundaries. These limitations significantly hinder the segmentation accuracy of deep learning models in complex clinical cases. Moreover, temporal motion of the heart plays a critical role in recognizing anatomical structures. To address these challenges, we designed a STLSF module which comprises a window-matching-based semantic correction component and a semantics-guided texture enhancement component. By leveraging local transition probability correlations to correct semantics and employing semantics-guided texture enhancement, the STLSF module effectively mitigates texture instability and ambiguous semantic interpretations caused by disadvantaged echocardiography quality. Additionally, to facilitate the encoder's adaptation to the intrinsic priors of ultrasound-specific imaging patterns, we propose a frequency-aware denoising pre-training method. The entire work builds a convolution-based network with locality inductive bias and long-range dependencies. Extensive experiments confirm our SOTA performance, achieving 93.87\% Dice on CAMUS and 92.62\% on EchoNet-Dynamic, with respective HD95 values of 3.29mm and 2.73mm.

Problem

• echocardiography segmentation
• speckle noise
• semantic ambiguity
• temporal motion
• low signal-to-noise ratio

Innovation

• transition probability correlation
• semantic correction
• texture enhancement
• frequency-aware denoising
• echocardiography segmentation