ESTIMATING OROFACIAL MOVEMENT IN HEAD-FIXED MICE USING DENSE OPTICAL FLOW

In head-fixed mice, fast orofacial movements such as whisking, breathing, and sniffing strongly modulate neural dynamics across cortical and subcortical circuits. Yet, commonly used behavioral descriptors often reduce movement to scalar measures, discarding critical information such as directionality, phase, and spatiotemporal structure. As a result, the coupling between fine-grained orofacial behavior and neural activity remains difficult to interpret.
Dense optical flow provides a principled framework to capture continuous, structured motion fields from video, but different algorithms can yield markedly different motion estimates when applied to subtle orofacial movements. Here, we systematically benchmark dense optical flow approaches for quantifying orofacial dynamics in head-fixed mice using facial video recordings from multiple laboratories. We evaluate how different algorithms capture region-specific motion patterns, with particular emphasis on fast dynamics that extend beyond the capabilities of keypoint-based tracking and motion energy measures.
We further assess generalizability, computational efficiency, and practical usability across experimental setups. Optical flow-derived motion features are validated against synchronously recorded analogue sniff sensor signals, revealing close correspondence in phase and frequency and supporting their physiological relevance. To examine biological interpretability at the circuit level, we apply dense optical flow to publicly available Brain-Wide Map video data and analyze relationships between orofacial motion features and neural activity across brain regions.
Together, this work establishes dense optical flow as a powerful and interpretable tool for capturing fast, structured orofacial dynamics and provides practical guidance for algorithm selection when linking fine-grained behavior to large-scale neural recordings.