Abstract: ¹⁸F-fluorodeoxyglucose integrated positron emission tomography and computed tomography (PET/CT) have been extensively used for evaluation of lung tumor. Traditional lung segmentation algorithm have not considered the detection of juxta-pleural nodules or large mass sufficiently and have segmented lung inaccurately. In this study, we developed a novel fully automated lung segmentation and lung tumor extraction algorithm for ¹⁸F-FDG PET/CT images in patients with lung cancer. The algorithm consisted of initial lung segmentation, adaptive maximum intensity projection, tumor extraction, and optimal lung segmentation. The validation of the algorithm was accomplished by comparing automated analysis results to the manual analysis results. The dice similarity coefficient, Jaccard index and accuracy of lung segmentation performance were 98.2 %, 96.6 % and 99 %, respectively. In extraction of lung tumor, sensitivity, specificity and accuracy were 87.9 %, 89.9 % and 88.5 %. Our automated algorithm for ¹⁸F-FDG PET/CT images performed lung segmentation and lung tumor extraction effectively in patients with lung cancer.

Keywords: Lung segmentation, Lung tumor extraction, Adaptive maximum intensity projection, Seed-based region growing.

_________________________________________________________________________________________