Application of genetic optimization to medical image segmentation

A number of important problems in medical imaging can be classified as segmentation problems. These segmentation problems can be formulated as configurational optimization problems by representing the configurations of interest in an image as unique subsets of the complete image. An effective segmentation optimization algorithm must determine the specific image subset that best exhibits an a priori set of quantitative characteristics. Here, a genetic optimization algorithm was used to produce a population of individual sub-images that were tested via a quantitative objective function, ranked using a linear fitness and decrement scheme, and modified using a genetic cross-over operator. The algorithm was found to converge within 25 to 50 generations to a good fit to the targeted configuration in a robust and efficient manner.<>