Spectral envelopes of rhythmic facial movements predict intention and motor cortical representations
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Animals, including humans, use coordinated facial movements to sample the environment, ingest nutrients, and communicate. To study these behaviors and the neural signals that underlie them, we introduce face-rhythm, a tool for quantitatively tracking, extracting, and interpreting facial movements. The approach utilizes markerless point tracking, spectral analysis, and tensor component analysis to extract demixed components of behavior from videos of facial movements. Face-rhythm is fully unsupervised and allows for the discovery of uninstructed behaviors; when applied to videos of facial behavior, face-rhythm identifies interpretable behaviors such as whisking, sniffing, and snout movements. Analysis of videos of mice in various behavioral conditions, including a classical conditioning protocol, a brain-machine interface (BMI) task, and natural behaviors outside of a task structure, revealed robust signatures of uninstructed facial movements in all regimes. The expression of these facial movements predicted internal belief states during classical conditioning and was correlated with instructed neural activity when the BMI was activated. Furthermore, facial behaviors identified by face-rhythm were highly represented in face-associated areas of primary motor cortex (M1). We found that M1 neural activity encodes a mixed representation of both the phase of facial movements as well as the phase-invariant spectral envelope of movement patterns, with higher-frequency facial movements being more likely to be represented as phase-invariant. Our results demonstrate that face-rhythm provides a novel and flexible approach for decomposing continuous face movements into natural behavioral motifs that are closely linked to neural activity patterns.