Evaluation of an improved automated analysis of freezing behaviour in rats and its use in trace fear conditioning.

Pavlovian fear conditioning in rodents often involves the assessment of freezing behaviour. Since human-based scoring is labour-intensive and subject to biases, several automated methods have been proposed, but few of them have been assessed for temporal definition. An accurate method for the automated analysis of freezing in rats is presented. It is based on a comparison of contrast from successive video images (Learn. Mem. 7 (2000) 58). A major improvement involves subtracting a previously recorded image of the empty background from each frame before processing. This simple operation greatly improves reliability, since it cancels the influence of video noise outside the region of interest in the image (the rat), and can also take advantage of the movements of a patterned animal over a textured background. The technique is compared with manual scoring and with the Anagnostaras method in a trace fear conditioning experiment. The method is appropriate to monitor the time course of freezing behaviour. The same system with a different threshold can also score locomotor activity. A quantitative evaluation of the residual errors inherent to automated methods is attempted.