Deep Learning Based Single-Shot Focused Tissue Harmonic Imaging: an in-Vivo Study

Pulse Inversion and Amplitude Modulation techniques are commonly used to extract the harmonic components in tissue harmonic imaging (THI). However, they are hindered by the reduction in frame rate and susceptibility to motion artefacts due to the necessity of firing two or more successive pulses. In our previous work, we showed the ability of deep-learning based techniques to achieve signal-based single-shot harmonic imaging in ex-vivo tissue. Here, we demonstrate the capability of our single-shot approach to achieve comparable results within in-vivo settings in contrast to conventional techniques. Specifically, the proposed approach yields harmonic images with comparable contrast metrics to the conventional checkerboard apertures technique, while achieving an approximate threefold increase in frame rate. Furthermore, we demonstrate the robustness of the approach across different human subjects and in-vivo tissues, alleviating the necessity for retraining. These findings further pave the way for a robust harmonic imaging technique achieving image quality that is on par with existing harmonic imaging methods but with a higher frame rate and fewer motion artifacts.