High-contrast ultrafast ultrasound power Doppler Imaging using subarray adaptive temporal autocorrelation with phase compensation

Compared to the conventional zero-lag autocorrelation (R(0)) estimation which directly accumulates the power of Doppler ensembles for ultrafast ultrasound power Doppler (PD) imaging, first-lag autocorrelation (R(1)) enhances the noise suppression in PD imaging by exploiting the temporal coherence among ensembles. The temporal-multiply-and-sum (TMAS) algorithm can further reduce the noise level by multiplying the correlation pairs in R(1) estimation before summation. Nonetheless, TMAS estimation often suffers from noticeable signal loss due to blood flow decorrelation in the presence of temporal flow variation. To overcome this, we propose a method to integrate phase compensation with subarray adaptive TMAS (SA-TMAS) algorithm to restore the signal of blood flow in R(1) PD images. In phase compensation, the phase shift of blood flow signal among different plane-wave transmit angles is corrected to improve the angular coherence within each ensemble. Moreover, the subarray method helps to preserve the flow signal by using a shorter temporal window for R(1) estimation. Correlation between the complementary transmit subset (CST) is also used for adaptive weighting of the subarray PD image and for adaptive modulation of the temporal coherence to avoid excessive suppression of blood flow. The proposed SA-TMAS with phase compensation is validated using simulation and in-vivo datasets. Results show that the proposed method achieves noise suppression comparable to the original full array TMAS while effectively preserving vascular signals due to reduced flow decorrelation. In the simulations, the phase compensation alone improves the contrast-to-noise ratio (CNR) of R(0) image by 2~8 dB depending on how dominant the axial flow is in the vessel. Using the R(0) image as the benchmark, the proposed R(1) SA-TMAS image additionally provides CNR improvement of 4 dB in the in-vivo dataset while maintaining the corresponding generalized CNR (gCNR) and thus enables the high-contrast vascular PD images.