Clustering 3D Tracking in Zebrafish Laboratory Trials

Due to the significant genetic similarities between zebrafish (Danio rerio) and humans, researchers are increasingly utilizing this species to develop novel therapies for various health conditions by analysing their behaviour across healthy and treated subjects. The current workflow in laboratory trials typically involves a small number of fish subjects, leading to simplified behavioural models where One-way ANOV A is often sufficient to meet statistical requirements. In this study, we propose an alternative approach to understanding fish behaviour. Our methodology begins with two datasets of 3D video recordings (95 zebrafish and 80 goldfish samples), followed by the application of the KMeans clustering and HMM (Hidden Markov Models) classification algorithms to time-series data derived from 3D tracking features.