PROJECT SUMMARY
Cardiovascular disease is a leading cause of mortality and morbidity worldwide. Currently, echocardiography
(i.e., ultrasound imaging of the heart) is the standard-of-care for optimal assessment and measurement of patient
heart function. Both transthoracic and transesophageal echocardiography are commonly used in hospitalized
patients, but they have limitations. Transthoracic echocardiography is non-invasive but data is only collected
while the probe is manually positioned on the patient's chest, which may result in missing important changes in
a patient's clinical course
and windows for potential intervention. Transesophageal echocardiography provides
real-time measurement of different indices of cardiac function, but requires general anesthesia or deep sedation.
To address these limitations, this proposed project aims to develop a wearable ultrasonic imager for continuous,
non-invasive monitoring of cardiac functions in real-time via a transthoracic approach. Based on our previous
work in wearable ultrasound blood pressure and flow sensors, the proposed device will utilize a stretchable
ultrasonic device with reduced transducer element pitch, enabled by a multilayered, highly stretchable and highly
integrated liquid metal electrode. The small pitch allows high transducer frequency and thus enhanced imaging
performance. The device will be designed to monitor the heart in two orthogonal planes simultaneously and will
use associated algorithms for B-mode imaging. To map the transducer locations on the dynamic human body,
we will use shape sensing multicore optical fibers to acquire the exact coordinates of each transducer and use
these location data for beamforming
. We will use the wide-beam compounding transmission strategy, which can
enhance the signal-to-noise ratio of ultrasound signals and therefore overall imaging performance of the device.
The results will be benchmarked against those from a clinical ultrasound probe. We will construct a deep learning
model based on Fully Convolutional Networks to process the acquired cardiac images automatically. We will
extract the left ventricular wall motion waveforms and left ventricular volume, from which critical indices, such as
stroke volume, ejection fraction, and cardiac output, can be generated continuously. The performance of the
device will be evaluated on a commercial phantom and tested on human subjects. After verifying the safety of
the proposed device, we will use it to collect the cardiac images from 138 healthy and patient subjects. The
accuracy of the results collected by the proposed device will be compared and analyzed statistically with that
obtained by a clinical-grade ultrasound machine. Our team is composed of highly interdisciplinary expertise that
is required for the success of the proposed research. If successful, this technology has the potential to
significantly improve cardiovascular monitoring and open new use cases for monitoring other critical visceral
organ, which will exert a sustained and powerful impact on many NIH-related research areas s.