A Comparative Study of Contrast Enhanced Ultrasound Imaging Using Deep Learning vs. Amplitude Modulation: An in-Vivo Investigation

In contrast enhanced ultrasound (CEUS) imaging, nonlinear extraction of microbubble signals typically involves multi-pulse acquisitions, such as amplitude modulation (AM), where two consecutive pulses are required to enhance nonlinear echo components. For this, the received signals are combined linearly to suppress the linear echo content. However, these methods are often hindered by their susceptibility to motion artifacts, an increased noise level and a reduced frame rate compared to single pulse acquisition sequences. Using in vivo data from a commercially available preclinical ultrasound system, we demonstrate that our novel signal-based deep learning approach achieves single-shot nonlinear CEUS imaging with comparable contrast metrics to the built-in AM technique while reducing noise, motion artifacts and approximately doubling the frame rate.