Spatiotemporal Adaptive Eigenspectrum Remapping in Transcranial Ultrasound Localization Microscopy

Abstract

Clutter suppression is a critical preprocessing step in ultrasound blood flow imaging, yet conventional singular value decomposition (SVD) filtering depends on data-specific eigenvalue cutoffs that can remove slow flow along with tissue clutter. Here, we introduce adaptive eigenspectrum remapping (AER), a general framework in which clutter is suppressed by continuously remapping, rather than truncating, the eigenvalue spectrum of the spatiotemporal covariance, and we investigate zero-phase component analysis (ZCA) matrix whitening as an instance of AER for transcranial ultrasound localization microscopy (ULM). Rather than discarding eigenspaces, ZCA-based AER whitens the eigenvalue spectrum, reducing the dominance of coherent clutter, including skull reverberation, while retaining weak flow signals. We evaluate AER on both temporal and spatial covariance matrices in mouse brain datasets acquired with long temporal ensembles. Compared with SVD filtering, ZCA improves microbubble contrast and vascular image quality, increases trajectory density across low- and high-concentration regimes, and reveals more low-velocity trajectories in ULM velocity maps. Spatial AER further maintains reconstruction quality across extended ensembles and enables fuller single capillary reporter (SCaRe) reconstructions, including connected artery-to-capillary-to-vein tracks. These results support AER, here instantiated as ZCA whitening, as a data-adaptive alternative to manually thresholded SVD filtering for high-resolution microvascular imaging.