Abstract

Neuroscientists are more and more interested by the neuro-biological mecanisms involved in the memory and/or the impact of stress. Video tracking is commonly used to study transgenic mouse behaviors, namely it's reaction to central object exploration in an open field arena. Automatic tracking systems requires less effort than manual tracking, but actual systems are based on center of mass. These tracking cannot be used to identify whether the mouse has touhced the novel object.
The main goal of this project is to develop an application capable of tracking mouse's muzzle precisely.

The main approach is to fit an ellipse in the mouse's outline and calculate the closest point to the muzzle. Sample points close to that are taken from the mouse's outline and a fast and reliable algorithm is implemented to find the tip of the muzzle. Algorithm finds the point for which (d1+d2) is maximum (see figure below). Tracking results on an image sequence have proven the reliability of this process. The next step of this project is to optimize the global processing, to take into account time space and to deal directly with videos.

In collaboration with Dr. Rime Madani, Laboratory of Behavioral Genetics, Life Sciences, EPFL.