Modeling behavior to disentangle motion-related effects in functional ultrasound imaging in awake, head-fixed mice

Abstract Functional ultrasound imaging (fUSI) enables brain-wide mapping of hemodynamic activity in awake rodents, offering a powerful bridge between systems neuroscience in animals and human neuroimaging. However, extending fUSI beyond tightly controlled sensory paradigms to more naturalistic behaviors is limited by motion-related confounds arising from locomotion, physiological arousal, and movement-linked neural activity. Here, we introduce a behavior-informed modeling framework that explicitly incorporates continuous measurements of running speed and head motion into a general linear model to account for motion-related variance while preserving interpretable task-related signals. We validate this approach in two head-fixed paradigms with distinct motion profiles: a visual stimulation task with minimal stimulus-driven movement, and a noxious stimulation task in which locomotion and arousal are intrinsic to the behavioral response. In both cases, explicit behavioral modeling recovers neural response patterns that more closely resemble low-motion reference conditions than blind, model-free denoising approaches. Critically, during noxious stimulation, behavior-informed modeling preserves shock-intensity-dependent activity in the primary somatosensory cortex that is attenuated by global component removal. These findings demonstrate that explicit modeling of behavior enables reliable interpretation of brain-wide fUSI signals during naturalistic, high-motion conditions, opening the door to studying affective and cognitive processes that were previously difficult to access with functional ultrasound imaging.