This is a summary of: Hu, H. et al. A wearable cardiac ultrasound imager. Nature 613667–675 (2023).
Cardiovascular diseases are often associated with changes in the pumping capabilities of the heart that can be detected by heart imaging. Non-invasive, continuous imaging of the heart would therefore be valuable for the early detection and surveillance of cardiovascular conditions. Existing methods for evaluating cardiac function are capable of either imaging or continuous measurement, but not both. Non-imaging methods contain less information than images, but imaging methods can capture only brief snapshots, involve cumbersome apparatus, and require highly trained technicians to manipulate the probe. Insufficient monitoring periods can result in failure to detect symptoms that appear only transiently, such as sinus arrhythmia (when the heartbeat is out of rhythm). Furthermore, the existing bulky imaging modalities cannot image the heart of an individual who is moving — an important limitation given that some disease-related responses appear only during exercise-induced stress1. Moreover, any quantitative measurements derived from cardiac imaging must currently be analyzed manually by cardiologists in a process that is arduous, lengthy and subject to inter-observer variability2.
We developed a wearable ultrasound patch that enables continuous imaging of the heart, even when the wearer is moving. The imager is small and can adhere to the human body without being held or requiring cumbersome fixation. It has similar elasticity to the skin and so provides minimal constraint to the user’s daily activities.
The imager is composed of two linear arrays of ultrasound transducers arranged in a cross shape, which enables imaging from two orthogonal views without the need for repositioning. The dense elements in the ultrasound patch provide an imaging quality that is similar to that of a commercial imager (Fig. 1a). Furthermore, we customized a neural network to automatically process the images to extract key metrics of cardiac performance, such as stroke volume (the volume of blood pumped per contraction), cardiac output (the volume of blood pumped per minute) and ejection fraction (the volumetric fraction of blood in one heart chamber, the left ventricle, ejected with each contraction), thereby eliminating inter-observer variability (Fig. 1b). To test the system, we monitored the heart’s response to stress before, during and after exercise. Using ultrasound to test cardiac anatomy and function during exercise is unprecedented.
It is said that a picture is worth a thousand words. The hands-free continuous operation of the wearable ultrasound imager provides many possibilities for improving both diagnostic accuracy and the patient’s experience.
Wearable ultrasound technology opens up a new dimension for deep-tissue sensing3–5. The technology could be extended to image various deep tissues and central organs, not just the heart. Continuous ultrasound images that capture the anatomy and dynamic functions of internal tissues and organs can provide an unprecedented amount of information about health status and fitness. The algorithm we developed, which uses an artificial-intelligence technique called deep learning, can provide objective, actionable information for clinicians, helping to inform their therapeutic decisions. Deep-tissue signals provide more-accurate information about disease status and prognosis than do the surface signals measured by conventional wearable devices. The continuous stream of information could also have substantial implications for predicting acute adverse events and enabling timely intervention, potentially saving many lives.
The wearable imaging patch in its present form is not perfect. Its spatial resolution, particularly in the elevational direction, must be enhanced. To adapt the device to different static and dynamic curved surfaces, the compensation of the distorted signals caused by deformation of the transducer array needs to be optimized by advanced imaging algorithms or geometry-sensing technologies. In addition, the deep-learning model should be generalized to make it applicable to a larger population. The device is currently tethered to a back end by flexible cables for data and power transmission; the back end controls the ultrasound and processes the images, and provides a user interface and image display screen. Further work is needed to develop a fully integrated, wearable ultrasound system with soft miniaturized driving electronics and a built-in power source. — Hongjie Hu and Hao Huang are at the University of California, San Diego, La Jolla, California, USA.
Behind the paper
This is a four-year project that started in early 2019. At the beginning, owing to a lack of medical knowledge, our team of engineers struggled to find an appropriate application for a wearable ultrasound imager. After learning in a casual chat that one of our colleagues’ relatives had died from a sudden heart attack, we realized that our technology could have a real impact on people’s lives by continuously monitoring cardiac function to reduce the mortality rate. This was the ‘bingo!’ moment that prompted us to approach cardiologists on the campus. The cardiologists told us they had never seen the heart monitored in motion and were greatly interested in our project. This is how we found the unique value of the wearable ultrasound imager through serendipity, and were able to develop it to fit the purpose. — H. Huang
Figure 1 is published under a CC BY 4.0 license.