An Information-Theoretical Approach to Medical Image Segmentation

In this note, we present a new method that allows us to determine threshold values for separating presence and absence of proteins in a stack of fluorescence images describing a spatial distribution of proteins across a biological object (like a slice of nervous tissue, a sample of blood cells etc.). This method is based on the so-called Multi-Information Function which is closely related to the Mutual-Information Function and the Kullback-Leibler distance. We apply this method to stacks of fluorescence images and find that the resulting threshold values are almost identical with threshold values found using completely independent methods based on technological and biological aspects of the images in question.