CORTICAL INTEGRATION OF BODY POSTURE AND WHISKER DYNAMICS IN FREELY MOVING RATS

By studying unrestrained, spontaneous behavior in rats, our group has previously demonstrated robust posture representations in sensory, motor and integrative cortices (Mimica et al., 2018, Mimica, Tombaz et al., 2023). Based on these findings, we hypothesized that the posterior parietal cortex (PPC) might serve as a high-level integration hub where whisker dynamics and body posture information are combined into a unified egocentric action space. To explore this hypothesis, we combine head-mounted cameras for facial tracking with 3D posture tracking and Neuropixels recordings. We first present different approaches for integrating facial tracking data into a behavioral segmentation pipeline and explore how including whisker data affects the segmentation of behavioral motifs. The first pipeline is based on methods previously utilized in the lab (adapted from Berman et al., 2014) and markerless whisker keypoints extraction using the DeepLabCut platform (Mathis et al., 2018). We also present a newly developed and detailed NeuroAI pipeline that embeds whisker videos into lower dimensional representations. The pipeline extracts information from video frames automatically using a deep network specifically tailored to detect whisker shapes, with subsequent embedding of the extracted data via a convolutional auto-encoder and dimensionality reduction with UMAP. We can then decode the dynamic state of the whiskers in low-dimensional UMAP space using population vector activity from the PPC and S1. Our preliminary results show how, across animals, neural representations effectively encode whisker configurations, and that this activity can be decoded with high accuracy.