Randomized channel subsampling method for efficient ultrafast ultrasound imaging

Abstract

Emerging as a novel modality with high spatiotemporal sensitivity, ultrafast ultrasound imaging has been well established and incorporated into clinical apparatus. However, such a mode typically requires large data acquisitions, with more than thousands of frames per second within a relatively long period, which poses high demands on real-time storage and processing hardware. Herein, an image reconstruction approach based on randomized channel subsampling is presented, which significantly reduces the amount of data acquisition while maintaining imaging performance. In vivo datasets from rats were used to evaluate the performance of the proposed method in B-mode imaging, ultrafast Doppler imaging, and super-resolution ultrasound localization microscopy (ULM) under different subsampling conditions. Following a thorough comparison of the contrast-to-noise ratio, signal-to-noise ratio, and visibility of the ultrafast Doppler based small-vessel imaging, the resolution and saturation of ULM imaging were also investigated under different subsampling conditions. The feasibility of the proposed method was demonstrated in task-based functional ultrasound (fUS) imaging with whisker stimulation. A comparison between the results of fUS based on introduced randomized channel subsampling and conventional fUS was also conducted to evaluate the subsampling effects on reconstruction accuracy and cerebral change detection sensitivity.