We developed a toolbox for videographic processing of head-fixed rodent behaviors. It extracts the principal components of the animal’s behavior, either in a single movie, or across movies recorded simultaneously. Several regions of interest in the movie can be processed simultaneously (such as the whisker pad or the nose). We found that the behavioral components from the full face of the mouse predicted up to half of the explainable neural activity across the brain.
In addition to extracting movement components, it can compute the pupil area of the rodent using a center-of-mass estimation. Also, in experiments in which the mouse sits on a surface with texture, the software can estimate the running speed of the mouse.