Characterizing the Spatiotemporal Boundaries of Functional Ultrasound for Brain Decoding

Abstract

Functional ultrasound imaging (fUS) measures task-associated changes in cerebral blood volume (CBV) with fine spatial sampling and rapid acquisition, supporting functional mapping and brain-signal decoding. As an indirect measure of neural activity, fUS-CBV signals are shaped by neurovascular coupling and vascular dynamics, which influence the spatial and temporal stimulus information available for decoding. Consequently, the functional distinctions accessible in fUS-CBV signals require characterization under controlled experimental conditions. Here, controlled whisker stimulation in awake, head-fixed mice was used to characterize these decoding-relevant response properties. Whisker identity and number, stimulation duration, and inter-stimulus interval were systematically varied. All-whisker stimulation produced contralaterally dominant responses in the primary somatosensory barrel field and ventral posteromedial thalamic nucleus. Different whisker inputs were associated with distinguishable cortical response distributions, while stimulation involving more whiskers produced broader representative activation patterns. Across durations of 0.5 s to10 s, brief stimulation generated measurable stimulus-associated signals, and response magnitude, persistence, and spatial distribution varied with stimulus duration. For paired 2-s stimuli, trial-level separability was already high at the shortest tested 1-s gap, corresponding to a 3-s onset-to-onset spacing, and reached 100% at gaps of 3 s or longer. Building on the experimentally characterized spatiotemporal response properties of fUS-CBV signals, a proof-of-concept modeling and decoding framework was developed, comprising cross-animal spatial prediction, classification of short versus long stimulation from S1BF {Delta}CBV time courses with known stimulus onset, and modeling of paired-event separability. Together, the results show that fUS-CBV signals retain structured spatial and temporal information about controlled sensory input after neurovascular transformation. These measured response properties characterize the spatiotemporal boundaries within which neural activity can be reliably distinguished using fUS-CBV signals, providing constraints for the development of fUS-based decoding models and the design of stimulation paradigms suitable for reliable fUS decoding.