Season-Dependent Variations in Behavioral Patterns of Freely Moving Rats, Revealed by Computer Vision

In this study, we applied a self-developed computer vision-based neural network to investigate the effect of seasons on the behavior of Wistar rats in the Open Field Test. We compared two groups of 10 rats, born in March/September and tested 2 months later. Despite standardized laboratory conditions, significant behavioral differences were observed in motor activity, exploratory and comfort behaviors. Spring-born rats exhibited increased exploratory (e.g. climbing) and comfort behaviors (e.g. grooming), while autumn-born rats showed greater motor activity and thigmotaxis. These findings suggest that seasonality influences long-term behavioral organization, even in controlled environments. To achieve objectivity and reliability of the analysis, machine learning methods were used, including pre-trained neural networks and time-color encoding for automated behavioral analysis, reducing human bias. The methodology involved video processing, semantic segmentation, classification, and statistical analysis of 13 predefined behavioral patterns. The results highlight the importance of accounting for birth timing as a covariate in behavioral studies, offering insights into the role of seasonal cues in shaping rodent behavior under controlled conditions. Our neural network proved its high sensitivity and reliability by identifying behavioral differences between two similar groups of intact control rats. This methodology can be successfully applied in models of nervous diseases or screening of pharmacological substances.