第24回小型魚類研究会
Peng Yang1, Riki Kajiwara1, Hiro Takahashi2, Motoyuki Itoh1
1 Graduate School of Pharmaceutical Science, Chiba University
2 Graduate School of Medical Sciences, Kanazawa University
Abstract
Zebrafish (Danio rerio) is an ideal model animal to monitor three-dimensional (3D) behavior. Because zebrafish shows complex actions to respond to a variety of stimuli or conditions, behavior analysis of zebrafish is popular in recent years, and there is a high demand on the identification of the previously unidentified behavioral patterns from complex 3D coordinates data.
Currently, the available software can analyze behaviors as numerical data like the coordinates, speed, and rotation angle etc. Since they can only deal with the quantification of primitive behavior features, it is difficult to use for finding new behavior features.
In this work, we aim to construct a new behavior analysis method that solves these problems with machine learning. Machine learning is a type of artificial intelligence that provides computers with the ability to learn without being explicitly programmed and enables the analysis correlating complicated several parameters and the feature extraction from the data automatically.
We used two cameras to capture 3D tracking data of zebrafish and identify specific behavioral patterns by analyzing 3D tracking data of zebrafish using FuzzyART which is a type of machine learning algorithm.
We performed an experiment to deliver electric shocks to zebrafish and tracked the zebrafish swimming in 3D simultaneously. By processing the obtained data with FuzzyART, we discovered a distinguishing behavior when the electric shock was applied but not seen when zebrafish were placed in a normal environment. And then we developed a web application ShinyR-3D-zebrafish for analyzing and visualizing zebrafish behavior pattern which recognized by machine learning. ShinyR-3D-zebrafish decreases the complexity and time required to visualize and analyze the data, producing informative data tables and interactable 3D-tracking plot and animation. Moreover, for the behavior pattern which newly defined by three-dimensional tracking analysis above, we also try to develop a tool to accept user-supplied data and detected and quantitate the behavior pattern. Our program has an intuitive graphical user interface that enables novice users to quickly perform complex analyses. And this technique could be applied to the discovery of a new behavior pattern link to mutant zebrafish, drug administration screening, and cognitive ability test of zebrafish in the future.
you can access demo of ShinyR-3D-zebrafish by click here.
