Unveiling hidden features of acoustic radiation forces in soft tissues via physics-informed neural network-based full shear wave inversion

The therapeutic efficacy employing mechanical effect of focused ultrasound (FUS) largely depends on precise control of the key features of acoustic radiation force (ARF) including spatial localization, magnitude distribution, and force field geometry. However, the heterogeneous nature of biological tissues poses persistent challenges in quantitative ARF characterization. Here, we report a novel methodology for quantifying focused ARF features by leveraging its mechanical consequences, specifically the shear waves generated by ARF in soft tissues. In our method, full shear wave inversion (FSWI) relying on a deep neural network is performed to reconstruct the otherwise inaccessible shear wave motions when the ARF is active. By integrating physical constraints from wave equations into the deep neural network, our method demonstrates remarkable robustness against noise and superior generalization capabilities in inferring the features of focused ARF. Numerical simulations and tissue-mimicking phantom experiments have been performed to validate this method. The results demonstrate that our approach enables reliable assessment of the ARF focal position, precise spatial mapping of the focal zone geometry, and reasonable quantification of ARF magnitude, which were not achievable with previous methods. Our method enhances precision in treatment planning while enabling dynamic intraoperative therapy tracking, thereby may promote the use of FUS across diverse clinical settings, including transcranial ultrasound (TUS) neuromodulation and the stimulation of endogenous immune responses.