An automatic lung cancer detection from X-ray images obtained through yearly serial mass survey

Abstract

A fully automatic method of detecting new lung nodules based on the subtraction method between two serial mass chest radiographs is proposed. Nodule detection is carried out following the three steps: pre-processing, subtraction and post-processing. In the pre-processing step, the outline of the bilateral lungs is extracted and modified in order to perform good subtraction between the two images to be compared. In the post-processing step, limitation and localization of candidate nodules of lung cancer are carried out. In these two steps, an evaluator function called "simpleness" of a single connection region is applied to identify outline of the lung and to select candidates of cancer(s).