Initialization Method for Lung CT Segmentation

Among all cancers, lung cancer (LC) is one of the most common tumors, presenting an increase of 2% per year on its worldwide incidence. In Brazil, INCA (Instituto Nacional de Câncer Ministerio da Saude (2018). http://www2.inca.gov.br/wps/wcm/connect/tiposdecancer/site/home/pulmao, [1]) estimates the occurrence of 31,270 new LC cases, being 18,740 among men and 12,530 among women, in 2018. In this work, we propose a method for seeds initialization for lung segmentation. The method goal is to spread seeds through the lung in its base, hilo and apex. The purpose of this technique is to increase the segmentation method speed and decrease the amount of needed iterations. The results were obtained by using 40, 60, 80 or 100 beams with lengths varying from 40, 60, 80 to 100 voxels. The beam length was also varied with 40, 60, 80 and 100 voxels. The difference of the proposed work to the methods in the literature is the different subregions. The proposed method will create different seeds over the three regions of apex, hilo and base for each lung.