Lightweight image segmentation for echocardiography

2025-09-03

AI Summary

To address the insufficient real-time performance of echocardiographic left ventricular (LV) segmentation, this paper proposes a lightweight U-Net architecture with only 2.0 million parameters, integrating deep supervision, affine augmentation, Dice loss, and post-processing optimization. Compared to nnU-Net, the proposed model reduces parameter count by 16× and achieves 4× faster single-frame inference (1.35 ms), while maintaining comparable accuracy on the CAMUS dataset (LV Dice score: 0.93) and demonstrating strong cross-dataset generalization. Ablation studies reveal that simple geometric augmentations and layer-wise supervision provide critical gains for cardiac segmentation. The model enables fully automatic LV volume and ejection fraction quantification, establishing an efficient and robust segmentation foundation for clinical real-time decision support.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Calibration & Uncertainty Quantification
• Search and Optimization: Learning to Search

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Accurate segmentation of the left ventricle in echocardiography can enable fully automatic extraction of clinical measurements such as volumes and ejection fraction. While models configured by nnU-Net perform well, they are large and slow, thus limiting real-time use. We identified the most effective components of nnU-Net for cardiac segmentation through an ablation study, incrementally evaluating data augmentation schemes, architectural modifications, loss functions, and post-processing techniques. Our analysis revealed that simple affine augmentations and deep supervision drive performance, while complex augmentations and large model capacity offer diminishing returns. Based on these insights, we developed a lightweight U-Net (2M vs 33M parameters) that achieves statistically equivalent performance to nnU-Net on CAMUS (N=500) with Dice scores of 0.93/0.85/0.89 vs 0.93/0.86/0.89 for LV/MYO/LA ($p>0.05$), while being 16 times smaller and 4 times faster (1.35ms vs 5.40ms per frame) than the default nnU-Net configuration. Cross-dataset evaluation on an internal dataset (N=311) confirms comparable generalization.

Problem

• Lightweight segmentation for echocardiography real-time use
• Reducing model size while maintaining segmentation accuracy
• Optimizing nnU-Net components for efficient cardiac analysis

Innovation

• Lightweight U-Net with 2M parameters
• Affine augmentations and deep supervision
• 16 times smaller and 4 times faster