Improving Automated Behaviour Analysis in Zebrafish Laboratory Trials

Abstract

Studies on zebrafish are carried out to learn how specific health conditions might be resolved. The fact that the genomes of this species and humans are so similar makes this possible. These trials allow behaviour comparisons between healthy, sick, and cured subjects. When the behaviours of those healthy and healed match or show a high degree of similarity, researchers frequently find a new medicine. However, there are obstacles that researchers must overcome, such as pricey equipment, time-consuming analysis techniques, and unsuitable settings for analysis devices. To enhance our comprehension of the activities of the fish subjects, we created the Framework for Activity Real-Time Monitoring. This application makes real-time processing of video records possible, as well as simple calibration and reconfiguration. It also attempts to characterize the behaviour to facilitate comparisons across a broad spectrum of behaviours. In this study, we provide prospective methods for using k-means clustering applied to a set of 85 features from motion tracking routes to categorize fish behaviours. To assess our solution, we additionally create a public dataset (based on 95 zebrafish and 80 goldfish video recordings) and compare the behaviour of the two fish species.