Abstract

Images of the olfactory system, part of the brain responsible for the perception of odours, were provided by the Brain and Mind Institute (Swiss Federal Institute of Technology Lausanne). The aim of the project was to develop a method performing an automatic identification of glomeruli, spherical structures of the olfactory bulb, appearing on the stack of images in the presence of odour stimulation.
The identification is primarily based on the temporal redundancy between images in the same stack, however it also takes advantage of the spatial distribution in individual images. In fact, the intensity of a glomerulus develops exponentially in time and its spatial distribution is gaussian. Due to a diffusion of the exponential activity, a method based merely on the temporal redundancy wouldnÕt be sufficient. A convolution with the difference of two Gaussians and a search for local maximums allow us to find the center of all spherical structures in the image. Finally an exponential function is fitted onto every pixel found by the convolution and the search for maximums. A simple threshold on the amplitude and the time constant of the approximated exponential makes it possible to distinguish whether the spherical structure is a glomerulus or not.