Development of a new 3D tracking system for multiple marmosets under free-moving conditions

Abstract

Abstract Accurate assessment of behavioral changes and social interactions in mammalian models is necessary to elucidate the pathogenesis of neurological disorders. The common marmoset (Callithrix jacchus) is a useful model in this regard. However, behavioral measurements, including assessments of sociality of free-moving group animals, have not been conducted in marmosets. Here, we developed a new behavioral analysis system for three-dimensional (3D) trajectories of independently free-moving multiple individuals by combining video tracking, 3D coordinates detected using light detection and ranging (Lidar), and facial recognition. Each marmoset was identified using deep learning facial recognition (accuracy ≥ 97%). Location preferences and distances between individuals were calculated using 3D trajectories, and grooming was detected using deep learning. This system will allow quantification of individual captive group animals, facilitating automatic measurement of social behavior. Furthermore, the behavioral changes observed might be directly extrapolated to humans and contribute to better understanding of the mechanisms underlying neurodegenerative disorders.