Pleura detection in chest computed tomography with application for nodule detection.

ABSTRACT

This paper proposes an efficient algorithm for detecting pleural objects that come in contact with a nodule. To reduce complexity, the algorithm recursively performed a curve-fitting method on each slice of the volume of interest to locate the object between the parietal and visceral pleurae surfaces and measured the quality of the fitting curve. When a nodule contacted the surfaces of the chest wall or diaphragm, they were automatically separated using the fitting curve. The algorithm was performed on 864 slices of 40 nodules. The segmentation results were visually inspected by a consensus of attending physicians to search for any segmentation errors. The consensus accepted 93.6% of the segmentation results.