Assessment of pain in mouse facial images

Analysing mouse behavior in medical experiments to determine adverse effects of medical drugs requires special expertise and it is a time consuming tedious task. Automatic scaling of facial pain mimics in mice are important for a fast and objective labeling. Although there exists a manual procedure for scaling mouse facial pain expression, a full automatic method does not exist yet. In this paper, a computational method is proposed for assessment of pain through facial expressions of mouses in experiments where pain paradigms are applied. For this purpose, mouse face regions in videos were extracted manually frame by frame and also their pain scales were labeled by experts in order to construct a data set. Then, this data set were used for training a neural network using deep learning. The results obtained in this preliminary study, where a limited dataset of still images was used, are quite encouraging. Our studies in order to make our results more reliable and to devolope a fully automatic approach for scaling mouse grimace in videos are still going on.