Automatic Analysis Of Rat Locomotor Behavior

The behavior analysis of the laboratory animal is an essential part of the research of central nervous functions. It is not only widely applied in neuropsychology and neuropharmacology but also extended to the interdisciplinary studies, such as the artificial intelligent, bionics and brain computer interface. The parameterized description of locomotor behavior become the basis of the behavior analysis, so to set up the method of automatic analysis of the animal behavior is significant for the development of the related science.According to the requirement of the locomotor behavior analysis and the characteristic of rat's behavior, the Gaussian mixture model tokenizer - behavior models was carried out to analysis the rat locomotor behavior in the open field test automatically.Firstly, digital image processing was used to automatically record the location of the rat's barycenter in open field test. And the robust Locally Weighted Scatter Plot Smoothing(LOWESS) and Repeated Running Median(RRM) were combined to extract features out of the track and recognize arrests from time series of track. Then the track was separated into two sub-behaviors (arrest and locomotion).Secondly, Gaussian mixture model tokenization was used to recognize the basic behavior units (progression, lingering episode and darting) from the track. In succession the new parameters (the spatial spread of lingering episode, number of darting) were gained and used to estimate the open field test.Finally, hidden Markov models was implemented to construct the models of the rat's locomotor behavior for describing the behavior in different conditions.An experiment to analysis rat's behavior in the condition of sleep deprivation was conducted using the automatic analysis method mentioned above. And the results suggest that the spatial spread of lingering episode and the median have more significant differences between the control group and every SD groups than thedistance moved. So that they perform better in reflecting the rat's locomotor and excited capability. Meanwhile the number of darting is confirmed to be a good index of measuring the rat's locomotor behavior in different conditions. In behavior description, the correct reorganization rate of normal model is higher than the SD model, and the rates are 100% and 66.7% respectively. So it comes to the conlusion that the Gaussian mixture models tokenizer - behavior models is efficacious and practical in automatic analysis the locomotor behavior of the laboratory rat in openfield test.