Super-Resolution Ultrasound Imaging by Sparse Bayesian Learning Method

Super-resolution ultrasound (SR-US) imaging technique overcomes the acoustic diffraction limit and greatly improves the spatial resolution. Furthermore, by exploiting temporal fluctuations in microbubbles, a super-resolution fluctuation imaging (SOFI) method has also been used to improve the temporal resolution of SR-US. However, this is at the expense of the reduction of spatial resolution. To address this problem, in this paper, a temporally correlated multiple sparse Bayesian learning (TMSBL) method is introduced to SR-US. Since TMSBL takes advantage of both the temporal correlation between successive frames and sparsity priors of microbubbles, it provides the possibility to obtain an enhancement in spatial resolution compared to SOFI (only considering the temporal fluctuations). To evaluate the performance of the proposed method, two types of numerical simulation and one physical phantom experiment were performed. Especially, we also investigate the effect of US imaging modes on the performance of SR-US implemented by TMSBL, where US data were acquired from the plane wave (the high imaging speed and the low spatial resolution) and synthetic transmit aperture (the low imaging speed and the high spatial resolution) scan, respectively. The experimental results show that compared to SOFI method when using the proposed TMSBL method, the imaging spatial resolution of SR-US can be improved. In addition, the imaging performance obtained by TMSBL can be hardly affected by US imaging modes. As a result, by combining with plane wave scan, TMSBL provides the potential in enhancing the spatial resolution of SR-US while maintaining a desired temporal resolution.