Application of neural-network-based deep-learning analysis to reveal changes of the movement parameters in rat neuromuscular disorder models

Abstract

Convolutional neural networks (CNNs) used for solving computer vision tasks are now able to detect objects and features with human-level accuracy. Deep learning algorithms often require large training datasets, which limit their use in everyday laboratory applications, however recently a new generation of CNNs emerged. For assessing the movement of the selected model animals in the 3D space with high spatial and temporal accuracy, we developed a method with combining the accurate pose estimation of the DeepLabCut CNN with RGB-D depth sensing device.