Animal Movement Pattern Model Identification based on Detrended Fluctuation Analysis

Abstract

We propose an original approach to the analysis of animal movement patterns in an open field test. In our approach, the characteristic movement patterns are revealed by the Detrended Fluctuation Analysis (DFA). The animal movement trajectories during the open field test have been retrieved from the video recordings using the DeepLabCut framework. Linear projections of the movement trajectories are represented in either Cartesian or polar coordinate systems and analyzed by DFA. Our results indicate that characteristic animal movement patterns indicate two asymptotic regimes, with positive power-law correlated increments at small scales and negative power-law correlated increments at large scales, separated by a single crossover. The above results give rise to a model-parametric animal movement pattern representation, that is capable of revealing significant differences between different treatment scenarios. We believe that due to its simplicity and unambiguity, the proposed approach could be useful for the enhancement of the effectiveness and speed-up of the decision making procedures in various applications involving animal tests, such as applied pharmacology, including novel drugs screening.