21 March 2014 Automatic lobar segmentation for diseased lungs using an anatomy-based priority knowledge in low-dose CT images
Author Affiliations +
Abstract
Lung lobar segmentation in CT images is a challenging tasks because of the limitations in image quality inherent to CT image acquisition, especially low-dose CT for clinical routine environment. Besides, complex anatomy and abnormal lesions in the lung parenchyma makes segmentation difficult because contrast in CT images are determined by the differential absorption of X-rays by neighboring structures, such as tissue, vessel or several pathological conditions. Thus, we attempted to develop a robust segmentation technique for normal and diseased lung parenchyma. The images were obtained with low-dose chest CT using soft reconstruction kernel (Sensation 16, Siemens, Germany). Our PC-based in-house software segmented bronchial trees and lungs with intensity adaptive region-growing technique. Then the horizontal and oblique fissures were detected by using eigenvalues-ratio of the Hessian matrix in the lung regions which were excluded from airways and vessels. To enhance and recover the faithful 3-D fissure plane, our proposed fissure enhancing scheme were applied to the images. After finishing above steps, for careful smoothening of fissure planes, 3-D rolling-ball algorithm in xyz planes were performed. Results show that success rate of our proposed scheme was achieved up to 89.5% in the diseased lung parenchyma.
© (2014) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Sang Joon Park, Jung Im Kim, Jin Mo Goo, Doohee Lee, "Automatic lobar segmentation for diseased lungs using an anatomy-based priority knowledge in low-dose CT images", Proc. SPIE 9034, Medical Imaging 2014: Image Processing, 903430 (21 March 2014); doi: 10.1117/12.2043353; https://doi.org/10.1117/12.2043353
PROCEEDINGS
7 PAGES


SHARE
Back to Top