Spatiotemporal Clustering of Functional Ultrasound Signals at the Single-Voxel Level.

Abstract

Functional ultrasound (fUS) imaging is a well-established neuroimaging technology that offers high spatiotemporal resolution and a large field of view. Typical strategies for analyzing fUS data comprise either region-based averaging, typically based on reference atlases, or correlation with experimental events. Nevertheless, these methodologies possess several inherent limitations, including a restricted utilization of the spatial dimension and a pronounced bias influenced by preconceived notions about the recorded activity. In this study, we put forth single-voxel clustering as a third method to address these issues. A comparison was conducted between the three strategies on a typical dataset comprising visually evoked activity in the superior colliculus in awake mice. The application of single-voxel clustering yielded the generation of detailed activity maps, which revealed a consistent layout of activity and a clear separation between hemodynamic responses. This method is best considered as a complement to region-based averaging and correlation. It has direct applicability to challenging contexts, such as paradigm-free analysis on behaving subjects and brain decoding.