Microbubble detection using Neyman-Pearson theory for volumetric ultrasound localization microscopy

Microbubble (MB) identification is a crucial step in ultrasound localization microscopy (ULM) as it defines the patches of clutter-filtered brightness (B-mode) images where the sub-wavelength localization algorithm should be applied. Neyman-Pearson (NP) criterion has been applied as a method for MB detection in 2D in-vivo ULM. The principle of NP is to establish a pixel-specific threshold based on the pixel’s temporal statistics. NP increased the resolution compared to intensity based identification (INT) and had a more complete mapping of the vessels than cross-correlation based identification (COR). However, the previously proposed method was limited to 2D and suffered from the out-of-plane motion and from the inability to capture the 3D structure of vascularization. Herein, NP is first extended to volumetric imaging and then compared to 3D INT and 3D COR. Simulation results demonstrate that the proposed method is more effective in detecting the low signal-to-noise ratio MBs compared to INT. In vivo rat kidney data further reveal that NP produces fewer artifacts in ULM maps while achieving similar resolution to INT and COR. In the other hand, NP produced less blurry maps, which may indicate either that NP detects less noise or that it is less sensitive to MBs.