Automatic segmentation of mouse behavior during video track-ing in home cages

A.B. Cherepov, T.V. Mukhina and K.V. Anokhin

Department of Systemogenesis, P.K. Anokhin Institute of Normal Physiology, Moscow, Russia

A common approach to analysis of behavior recorded by video tracking systems employs characteristics like animal speed, moved distance, latencies to enter certain zones, times spent there etc. These parameters only indirectly reflect the organization of complex behavioral patterns and their dynamics during learning. To overcome this limitation we developed a novel approach to quantitative segmentation of animal behavior by automatic dissection of behavioral continuum into meaningful behavioral units.

Our approach is based on the theory of functional systems (P.K. Anokhin 1935-1973) which segments behavior into similarly organized elementary functional units – ‘behavioral quanta’. As a first approximation to such segmentation we searched for ‘main’ breaks of movement continuum (complete stops in animal motion), using the running median algorithm introduced by D. Drai et al. However, analysis of video records of animal behavior after such coarse segmentation reveals ‘minor’ behavioral acts (e.g. turning, running, and rearing) that are often located within the obtained large segments. Therefore we introduced additional criteria for automatic recognition of behavioral breakpoints:

  • Episodes of deceleration followed by acceleration, while the speed of animal movement is low.
  • Change of angle between speed vectors for a pair of trajectory points.Employing these parameters we analyzed the behavior of C57Bl/6 and BALB/C mouse strains in response to a novel environment. Mouse activity was recorded in the home or new cages with a video tracking system. Natural structure of behavior was discriminated with segmentation software that was based on the described parameters. Our results demonstrate that the suggested parameters allow to approximate the automatic segmentation of behavior to the expert recognition of breakpoints in the mouse behavioral continuum.

Paper presented at Measuring Behavior 2005 , 5th International Conference on Methods and Techniques in Behavioral Research, 30 August - 2 September 2005, Wageningen, The Netherlands.

© 2005 Noldus Information Technology bv