Edinburgh Research Explorer

Sex, Flies and no Videotape

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Related Edinburgh Organisations

Original languageEnglish
Title of host publication5th International Conference on Methods and Techniques in Behavioral Research
Publication statusPublished - 2005

Abstract

The courtship ritual for the small fly Drosophila melanogaster has been discussed in the academic literature for almost 100 years. It is an exemplary behavior for the functional dissection of the nervous system as it is a natural behaviour that requires the animal to integrate and learn from multiple sensory modalities. Thus courtship activity can be used as a very finely tuned assay for general CNS function, olfactory capabilities or as a learning and memory assay by modifying the simple conditions of the basic experiment. Despite its many benefits, it is labour intensive to analyse quantitatively and typically involves around 100-200 minutes of manually annotated video footage per experimental data point. Therefore the genetic screens that Drosophila is rightly famous for are prohibitively expensive to perform using this as an assay. Although an obvious candidate for computational analysis, even the latest research papers use manual methods.

We have developed a hybrid system that tracks the animals in real time from a live video feed or from file. The tracking system has been developed to deal with 3D occlusion problems by a combination of segmentation optimisation and temporal analysis of moving objects pre and post occlusion. From the tracked objects we find that we can reliably measure the intensity of courtship activity between two flies using a range of parameters over time that include orientation and distance between the objects. These parameters are learned from expertly annotated video files. Although designed with this very specific experimental aim in mind, the methods are implemented in such a way that the numbers of objects in the environment is not predetermined by the user and are learned from the video. Further, the behavioral classifier is not specific and is trained by learning parameters that fit expertly annotated video footage. Hence our system should generalise beyond the exciting, but miniature, world of insect mating.

ID: 4869931