Improving the Quality of Fetal Heart Ultrasound Imaging With Multihead Enhanced Self-Attention and Contrastive Learning

Abstract

Fetal congenital heart disease (FCHD) is a common, serious birth defect affecting ~1% of newborns annually. Fetal echocardiography is the most effective and important technique for prenatal FCHD diagnosis. The prerequisites for accurate ultrasound FCHD diagnosis are accurate view recognition and high-quality diagnostic view extraction. However, these manual clinical procedures have drawbacks such as, varying technical capabilities and inefficiency. Therefore, the automatic identification of highquality multiview fetal heart scan images is highly desirable to improve prenatal diagnosis efficiency and accuracy of FCHD. Here, we present a framework for multiview fetal heart ultrasound image recognition and quality assessment that comprises two parts: a multiview classification and localization network (MCLN) and an improved contrastive learning network (ICLN). In the MCLN, a multihead enhanced self-attention mechanism is applied to construct the classification network and identify six accurate and interpretable views of the fetal heart. In the ICLN, anatomical structure standardization and image clarity are considered. With contrastive learning, the absolute loss, feature relative loss and predicted value relative loss are combined to achieve favorable quality assessment results. Experiments show that the MCLN outperforms other stateof-the-art networks by 1.52-13.61% when determining the F1 score in six standard view recognition tasks, and the ICLN is comparable to the performance of expert cardiologists in the quality assessment of fetal heart ultrasound images, reaching 97% on a test set within 2 points for the four-chamber view task. Thus, our architecture offers great potential in helping cardiologists improve quality control for fetal echocardiographic images in clinical practice.

A deep learning framework for identifying and segmenting three vessels in fetal heart ultrasound images

“…Deep learning has been applied in the field of fetal echocardiography for various tasks, including standard plane identification from a sequence of fetal heart images [ 23 – 25 ], detection of abnormal structures [ 26 – 29 ], and segmentation of cardiac structures [ 30 – 33 ]. Most studies designed for fetal cardiac structure segmentation have focused on the four-chamber view of the fetal heart [ 30 , 31 , 33 ].…”

Fetal Heart Disease Detection Via Deep Reg Network Based on Ultrasound Images

“…Surgery or other practices are necessary in the initial stage of their life for babies with critical CHD. Others may cause some serious clinical signs in a baby that include blue-colored nails or lips, difficulty breathing, fatigue while feeding and sleepiness (Zhang et al, 2023a). To effectively enhance the infant's diagnosis, rapid recognition of CHDs is important because it can require anomalous fetal heart structures; thus, allowing clinical treatment as soon as possible (usually within a week after birth) is required (Xiao et al, 2023;Prayer et al, 2023).…”