Statistical Dependency Guided Contrastive Learning for Multiple Labeling in Prenatal Ultrasound

TL;DR

This paper introduces a multi-label learning framework with statistical dependency guidance and contrastive learning for improved prenatal ultrasound plane recognition, achieving high accuracy and broad applicability.

Contribution

It presents a novel multi-label learning scheme using word embeddings, graph convolutional networks, and contrastive learning for fetal ultrasound image classification.

Findings

Achieved 90.25% accuracy in standard plane labeling

Reported 85.59% accuracy in planes and structures labeling

Obtained 94.63% mAP on large dataset

Abstract

Standard plane recognition plays an important role in prenatal ultrasound (US) screening. Automatically recognizing the standard plane along with the corresponding anatomical structures in US image can not only facilitate US image interpretation but also improve diagnostic efficiency. In this study, we build a novel multi-label learning (MLL) scheme to identify multiple standard planes and corresponding anatomical structures of fetus simultaneously. Our contribution is three-fold. First, we represent the class correlation by word embeddings to capture the fine-grained semantic and latent statistical concurrency. Second, we equip the MLL with a graph convolutional network to explore the inner and outer relationship among categories. Third, we propose a novel cluster relabel-based contrastive learning algorithm to encourage the divergence among ambiguous classes. Extensive validation was performed on our large in-house dataset. Our approach reports the highest accuracy as 90.25% for standard planes labeling, 85.59% for planes and structures labeling and mAP as 94.63%. The proposed MLL scheme provides a novel perspective for standard plane recognition and can be easily extended to other medical image classification tasks.