Feature Extraction for Classification of Caenorhabditis Elegans Behavioural Phenotypes

Abstract

Caenorhabditis (C.) elegans is often used in genetic analysis in neuroscience because it has simple model organisms; an adult hermaphrodite contains only 302 neurons. We use an automated tracking system, which makes it possible to measure the rate and direction of movement for each worm and to compute the frequency of reversals in direction. In this paper, we propose new preprocessing method using hole detection, and then we describe how to extract features that are very useful for classification of C. elegans behavioural phenotypes. We use 3 kinds of features (Large-scale movement, body size, and body posture). For the experiments, we classify 9 mutant types of worms and analyze their behavioural characteristics.