DCD: A Semantic Segmentation Model for Fetal Ultrasound Four-Chamber View

2025-06-10

AI Summary

Accurate segmentation of key anatomical structures in the fetal ultrasound four-chamber view is critical for early prenatal diagnosis of congenital heart disease (CHD), yet remains challenging due to artifacts, speckle noise, gestational-age-dependent anatomical variability, and ill-defined boundaries. To address these challenges, we propose an encoder-decoder network integrating Dense Atrous Spatial Pyramid Pooling (Dense ASPP) and Convolutional Block Attention Module (CBAM). Dense ASPP—introduced here for the first time in fetal cardiac ultrasound segmentation—enhances multi-scale feature representation, while CBAM enables joint channel- and spatial-wise adaptive attention, improving robustness to artifacts and boundary localization accuracy. Evaluated on a real-world clinical fetal ultrasound dataset, our method achieves state-of-the-art segmentation performance (Dice score: 92.3%) with strong generalizability and significantly reduces annotation burden for clinicians.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Calibration & Uncertainty Quantification
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

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: Agentic search
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

Abstract

Accurate segmentation of anatomical structures in the apical four-chamber (A4C) view of fetal echocardiography is essential for early diagnosis and prenatal evaluation of congenital heart disease (CHD). However, precise segmentation remains challenging due to ultrasound artifacts, speckle noise, anatomical variability, and boundary ambiguity across different gestational stages. To reduce the workload of sonographers and enhance segmentation accuracy, we propose DCD, an advanced deep learning-based model for automatic segmentation of key anatomical structures in the fetal A4C view. Our model incorporates a Dense Atrous Spatial Pyramid Pooling (Dense ASPP) module, enabling superior multi-scale feature extraction, and a Convolutional Block Attention Module (CBAM) to enhance adaptive feature representation. By effectively capturing both local and global contextual information, DCD achieves precise and robust segmentation, contributing to improved prenatal cardiac assessment.

Problem

• Accurate segmentation of fetal A4C view for CHD diagnosis
• Overcoming ultrasound artifacts and anatomical variability challenges
• Reducing sonographers' workload with deep learning model

Innovation

• Dense ASPP for multi-scale feature extraction
• CBAM enhances adaptive feature representation
• Deep learning model for fetal A4C segmentation