Fast channel-domain denoising for ultrafast Doppler imaging

Ultrafast Power Doppler (UPD) imaging is a well-established method for visualizing the microvasculature. Before computing a UPD image, a critical step is to filter out tissue clutter. Currently, the most common approach is singular value decomposition (SVD), as it allows blood and tissue subspaces to be separated in ultrasound images with great sensitivity. However, UPD images are also corrupted by depth-dependent electronic noise, which is present in the same subspaces as blood and therefore cannot be fully filtered out using SVD. Here we propose the Coherence Factor Mask (CFM) filter, a strategy for denoising UPD images in the channel domain. The CFM filter combines channel-domain SVD and coherence-based pixel weighting. In addition, to cope with the high computational load associated with channel processing, we combine this filter with power SVD (pSVD), an accelerated implementation of SVD for power Doppler imaging. Results from an in vivo experiment demonstrate improvements in signal and contrast-to-noise ratio in UPD images of the order of 15 and 13 dB, respectively, when using CFM. In addition, the combination of CFM and pSVD reduces the running time of the method by a factor of 12 without impacting its efficiency.