Computational analysis of movement behaviors of medaka (Oryzias latipes) after the treatments of copper by using fractal dimension and artificial neural networks

Abstract

Response behaviors of medaka were computationally analyzed before and after the treatments of copper at low concentration (1.0 mg/L). Parameters (e.g., speed, stop time, turning rate, etc) of the movement patterns were used as input for training the Multi-Layer Perception. Detection rates of the movement patterns such as 'Slow movement' and 'No movement' increased after the treatments. However, a higher degree of variation was observed in detection rates. Fractal dimension calculated from the movement data of individual specimens decreased consistently after the treatments. Higher consistency in fractal dimension was further achieved by using the data for collective rearing. Feasibility of behavioral monitoring was discussed in assessing toxic chemicals in environment.