A tool to classify motor behavior from video recordings of the mouse face. See our paper on bioRxiv and 2020-03-08_Master_Thesis_Wiessalla.pdf for details. All steps from the watershed clustering algorithm are taken from Berman et al. 2014 “Mapping the stereotyped behaviour of freely-moving fruit flies”: Repository, Publication. The HOG generation parameters & code logic follows Dolensek et al. 2020 “Facial expressions of emotion states and their neuronal correlates in mice“: Publication. Their GitHub repository can be found here, but full code is only on zenodo.
Simply clone the repository to a destination of your choice and run it from there. You require MATLAB2018b to run. The code relies onthe following packages: Image Processing Toolbox 11.1, Statistics and Machine LearningToolbox 11.7, Computer Vision Toolbox 9.2, Parallel Computing Toolbox 7.2, MATLABParallel Server 7.2. Tested file type for the videos is *.mp4.
Add your file paths to the fileNames input cell when calling processWrapper. You will be prompted with a GUI loading the first frame of your video. Draw a ROI for the area of the face you wish to analyze. The second ROI was designed to specify the eye. Originally this is done for synchronisation, as our 2-P-IR laser beam will shine outwards through the eyes. As our original setup uses an IR camera it’s an easy way to determine the start of a recording. If you don’t need that functionality specify 'noDrop=true' when calling processWrapper. Then just draw your ROI anywhere and confirm with a double-click.
This was written as part of Tristan Wiessalla’s master thesis at our group at EMBL. In-detail explanation of the code can be found in his thesis, especially in Appendix A. A more broad explanation can be found in chapter 6.