Enhancing Fetal Cardiac Ultrasound Diagnosis: A Multi-Task Hybrid Attention Model for Accurate Standard Plane Detection

Abstract

Fetal heart health is a critical part of diagnosis and treatment, and one of the methods is fetal cardiac ultrasound. A key aspect of the process is the detection of standard ultrasound slices, which is essential for accurate diagnosis. The effectiveness of diagnosis relies heavily on the clinical experience and expertise of the ultrasound physician. To improve detection efficiency and minimize misdiagnosis, we developed a single-stage detection model for fetal cardiac ultrasound standard planes (FCUM) that uses multi-task learning and hybrid attention mechanisms to support the ultrasound physician’s diagnostic work.