Echo-Robot: Semi-autonomous cardiac ultrasound image acquisition using AI and robotics

Abstract

Echocardiography is a critical tool for diagnosing cardiovascular diseases, offering detailed insights into heart functions. However, its accessibility is currently limited by a shortage of trained sonographers, specific skill requirements, and the physical strain imposed on professionals during repetitive procedures. This article introduces a new robotic system designed to automate the acquisition of transthoracic echocardiography (TTE) images. The system autonomously adjusts the position and orientation of the ultrasound transducer based on analysing realtime ultrasound images, without relying on tomographic data or depth sensors. Initially, the transducer is manually placed on the subjects skin, and the system uses a deep learning approach to grade the quality of ultrasound images captured at each position. The robot then adjusts its position by spiralling outwards from the starting point, moving to the location with the highest image quality score. Next, the system fine-tunes the transducers orientation in 5-degree increments along all three axes of rotation, informed by another deep learning module to identify the field of view. The robotic system was tested using a cardiac simulator, achieving approximately 80 view when the probe was initially positioned randomly in a 6 by 6 cm area beneath the left nipple. The impact of this work would be rapid diagnostics in the Emergency Departments to reduce the length of stay in hospitals, a reduction of hospital admissions related to heart disease by accessing local healthcare communities, acceleration of clearing the post-Covid backlog, and improved quality of life and longevity of patients.