Convolutional Neural Network Analysis of Social Novelty Preference using DeepLabCut

• DOI: 10.1101/736983
• Publication Date: 2019-08-16T03:44:08Z
• Authors: Nicholas B. Worley; Anthony Djerdjaj; John P. Christianson
• Keywords: Identification

Abstract The description and quantification of social behavior in laboratory rodents is central to basic translational research. Conventional ethological approaches are fraught with challenges including bias, significant human effort temporal accuracy. Here we show proof principle that machine learning can be applied tests decision making. Rats underwent novelty preference which were scored both by hand again a convolutional neural network generated the DeepLabCut computer vision package Mathis colleagues. CNN temporally (30Hz) locally (<5pixels) accurate identification rat nose, eye ear positions then used compute interaction topography heat maps. In sum, hand- computer-scoring strongly correlated, each identified preferences interact novel conspecifics sets stage for applying analysis other types future.