LABORAS: automated behaviour classification of laboratory animals

R.J.A. Bulthuis1, A.F. Bergman1, F. Schlingmann2, J. Tolboom2, R. Remie2, H.A. van de Weerd3, P.L.P. van Loo3, V. Baumans3 and L.F.M. van Zutphen3

1 Metris System Engineering, Hoofddorp, The Netherlands
2 Department of Laboratory Animal Science, Solvay Pharmaceuticals b.v., Weesp, The Netherlands
3 Department of Laboratory Animal Science, Utrecht University, Utrecht, The Netherlands

 

Introduction

At present, the analysis of laboratory animal behaviour is based mainly on video recording followed by visual observation or computerised screening of the recordings. Visual observation is time-consuming and scan sampling of a proportion of frames gives only a rough indication of the actual behaviour of the animal. Metris has developed a system named LABORAS (Laboratory Animal Behaviour Observation, Registration and Analysis System) for automated classification of behavioural patterns by analysis of sensor signals induced by the animal's movements. Each of the behavioural categories locomotion, sleeping, climbing, grooming, drinking and eating is characterised by a specific and unique movement pattern which is detected by the sensors and then processed by computer into the behavioural categories. Validation experiments have indicated that the technique is accurate and the results reproducible [1, 2, 3]. The system has several advantages, e.g. no visual observation needed, reduction of time needed for behavioural research and a flexible time window. The system is suitable for mice and rats.

System description

Basically, LABORAS consists of a sensing platform, a signal conditioning and data acquisition device and software for administration, control and behaviour classification.

Sensing platform
The sensing platform (Fig. 1) is constructed of low-mass material and positioned on two orthogonally-placed sensors and a third, fixed point. The cage with the laboratory animal is placed at a marked position . The platform is designed to accommodate Macrolon types II and III cages for tests with mice and rats, respectively. The sensors are placed so that all movements are optimally detected and the approximate location of the animal can also be determined. To reduce the inertia of the platform and to enable the detection of climbing behaviour, the wire top and food hopper is mechanically separated from the lower part of the cage.

Figure 1. Diagram of the triangular sensing platform and a cage.

Signal conditioning and data acquisition device
Each sensor transforms the mechanical vibration into electrical signals which are filtered and amplified. Subsequently, the signals are sampled and stored on the computer's hard disk for later processing. Up to eight platforms can operate simultaneously.

Software for administration, control and classification
There are three software modules:

The analysis software has a modular design to enable addition of new specific behavioural categories or more detailed behavioural patterns.

Validation

Several LABORAS systems have accumulated an amount of data which is equivalent to more than 2500 h of experiment time. For validation purposes a setup with 4 platforms was used, each housing one animal. Two tests were performed, one with mice and one with rats, each lasting 48 h. To evaluate the performance of LABORAS, different strains (C57BL and BALB/c mice; Wistar and Sprague Dawley rats) and sexes were used. Both tests were recorded using a video camera and a time-lapse video recorder for each platform. Each videotape was scored by three different observers independently. The scoring method applied was scan sampling with a sampling rate of once every 15 s.

Results and conclusions

The tests with mice show good results. For all four mice and four rats in the test LABORAS differentiated between climbing (mice only), sleeping (resting), locomotion, grooming, eating and drinking. On average the correlation between human observers and LABORAS appeared to be better than 85% [2, 3]. It can be concluded that LABORAS offers a new efficient and reliable method for automated behaviour classification and position tracking of small rodents.

References

  1. Baumans, V.; Schlingmann, F.; van de Weerd, H.A.; Remie, R.; van Zutphen, L.F.M. (1996). Development of a balance system for analysis of rodent behaviour. Measuring Behavior ’96 (Utrecht, 16-18 October 1996), 29.
  2. Schlingmann, F.; van de Weerd, H.A.; Blom, H.J.M.; Baumans, V.; van Zutphen, L.F.M. (1994). Behavioural differentiation of mice housed on different cage floors. In: Welfare and Science, Proceedings of the 5th FELASA Symposium (J. Bunyan, ed), pp. 355-357. London: Royal Society of Medicine Press.
  3. Schlingmann, F.; van de Weerd, H.A.; Baumans, V.; Remie, R.; van Zutphen, L.F.M. (1998). A balance device for the analysis of behavioural patterns of the mouse. Animal Welfare, 7, 177-188.

Poster presented at Measuring Behavior '98, 2nd International Conference on Methods and Techniques in Behavioral Research, 18-21 August 1998, Groningen, The Netherlands

© 1998 Noldus Information Technology b.v.