Topology-Aware Brain Vessel Segmentation in Ultrafast Doppler Imaging

Changes in the morphological features of cerebral microvasculature are closely related to various cerebrovascular diseases. Quantification and assessment of cerebrovascular morphology require vessel segmentation in vascular imaging. In recent years, the signal-to-noise ratio and spatial resolution of ultrasound vascular images have been greatly improved by ultrafast Doppler imaging. However, brain vessel segmentation in ultrafast Doppler images remains challenging due to vessel discontinuity as well as non-uniform intensities and blur. In this work, we present a topology-aware segmentation (TAS) method for brain vessel segmentation in ultrafast Doppler imaging. TAS first binarizes the images using local thresholds derived from an estimated vessel orientation field and then reconstructs the vessel region under a global connectivity constraint. Experimental results on imaging data of a rat brain suggest that TAS can achieve higher topological fidelity and provide a more precise estimation of vessel size. Moreover, we successfully reconstruct the vasculature of a whole rat brain from imaging data collected in 80 seconds via TAS, demonstrating its great potential in facilitating convenient and reliable assessment of cerebral microvasculature.