Automated classification of sensorimotor deficit in stroke mice

this repository contains the data and evaluation results of the master thesis 'Automated classification of sensorimotor deficit in stroke mice' by Sebastian Kloubert. The thesis aimed to evaluate whether a combination of different camera angles and the application of the software package DeepLabCut (DLC) to video tracking data is appropriate to automatically detect events related to sensorimotor behavior of the behavior tests Grid Walk and Cylinder and simultaneously overcome the manual rater-based detection errors. The new test setup promises a better view of the behavior tests Grid Walk and Cylinder for manual analysis. The attempt to use these new camera angles to automate the behavior tests by applying the software package DeepLabCut failed for both the two-dimensional and the three-dimensional analysis approach and did not overcome manual rater-based detection errors. However, several possible improvements can be outlined, which should allow for quick automation in future work. The calculation of additional kinematic features promises to make the occurrence of the events related to sensorimotor behavior of the behavior tests more precise and perhaps in the future to allow a separation of stroke and healthy mice.