23 February 2012 Lung lobe segmentation based on statistical atlas and graph cuts
Author Affiliations +
This paper presents a novel method that can extract lung lobes by utilizing probability atlas and multilabel graph cuts. Information about pulmonary structures plays very important role for decision of the treatment strategy and surgical planning. The human lungs are divided into five anatomical regions, the lung lobes. Precise segmentation and recognition of lung lobes are indispensable tasks in computer aided diagnosis systems and computer aided surgery systems. A lot of methods for lung lobe segmentation are proposed. However, these methods only target the normal cases. Therefore, these methods cannot extract the lung lobes in abnormal cases, such as COPD cases. To extract lung lobes in abnormal cases, this paper propose a lung lobe segmentation method based on probability atlas of lobe location and multilabel graph cuts. The process consists of three components; normalization based on the patient's physique, probability atlas generation, and segmentation based on graph cuts. We apply this method to six cases of chest CT images including COPD cases. Jaccard index was 79.1%.
© (2012) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Yukitaka Nimura, Takayuki Kitasaka, Hirotoshi Honma, Hirotsugu Takabatake, Masaki Mori, Hiroshi Natori, and Kensaku Mori "Lung lobe segmentation based on statistical atlas and graph cuts", Proc. SPIE 8315, Medical Imaging 2012: Computer-Aided Diagnosis, 83153B (23 February 2012); doi: 10.1117/12.911254; https://doi.org/10.1117/12.911254

Back to Top