Microbubble Identification Based on Decision Theory for Ultrasound Localization Microscopy

Abstract

Ultrasound localization microscopy (ULM) enables the evaluation of the vascular microstructure by detecting, localizing, and tracking microbubbles (MBs) in the vascular network. ULM provides a vascular map of the network with improved spatial resolution but with an acquisition time of several minutes. Thus, it is of great importance to increase the number of MBs detected in order to limit the acquisition time. The standard MB detection method in ULM assumes that the contrast agents are the highest-intensity structures on the ultrasound images. However, in vivo data show that MB intensity may be lower than residual tissue or even noise. Thus, to facilitate the detection of these MBs, an MB detector based on decision theory is proposed in this paper. In this study, the proposed method based on the Neyman–Pearson criterion is compared with the standard intensity-based and the normalized cross-correlation detection methods on simulated and in vivo rat brain and kidney data. The new detection method makes it possible to control the false positive detection rate without degrading the MB detection rate on simulated data, to enhance the ULM vessel map resolution on in vivo brain data and to detect more vessels on in vivo kidney data.