Real-World Visual Navigation for Cardiac Ultrasound View Planning

Echocardiography (ECHO) is commonly used to assist in the diagnosis of cardiovascular diseases (CVDs). However, manually conducting standardized ECHO view acquisitions by manipulating the probe demands significant experience and training for sonographers. In this work, we propose a visual navigation system for cardiac ultrasound view planning, designed to assist novice sonographers in accurately obtaining the required views for CVDs diagnosis. The system introduces a view-agnostic feature extractor to explore the spatial relationships between source frame views, learning the relative rotations among different frames for network regression, thereby facilitating transfer learning to improve the accuracy and robustness of identifying specific target planes. Additionally, we present a target consistency loss to ensure that frames within the same scan regress to the same target plane. The experimental results demonstrate that the average error in the apical four-chamber view (A4C) can be reduced to 7.055 \(^\circ \) . Moreover, results from practical clinical validation indicate that, with the guidance of the visual navigation system, the average time for acquiring A4C view can be reduced by at least 3.86 times, which is instructive for the clinical practice of novice sonographers.