Automated detection of nodules in the ct lung images using multi-modal genetic algorithm
Jamshid Dehmeshki, Md Abu Bakker Siddique, Xinyu Lin, Mary E. Roddie, J. Costello · 2004
This study deploys a multi-modal genetic algorithm (GA) augmented by an island model cooperating with a speciation module to identify lung nodules in chest CT images. The genetic algorithm is a model of machine learning which derives its behaviour from the processes of evolution in nature. The island model based GA maintains diversity and converges towards different solutions hence capturing multiple peaks of the fitness function. The speciation module gathers the genetically similar individuals into one pool to avoid accumulation of several subpopulations around the same peak of the fitness function. The detection process comprises two stages. In the first stage, the template matching based GA was used to determine the target position in the observed image efficiently and to select an adequate template image from several reference patterns for quick template matching. The fitness of the individual (chromosome) was defined as the similarity calculated by the cross correlation coefficient between the target and template image as determined by the chromosome. In the second stage, the GA scheme was used to detect the regions with circular elements present in the segmented CT images as potential nodule. The fitness function of GA process was defined using three points, from the boundary of the object, given by the chromosome. The results show that the scheme can be efficiently applied for detection of isolated or attached circular regions present in the images.