AbstractsBiology & Animal Science

Automatic 3D extraction of pleural plaques and diffuse pleural thickening from lung MDCT images

by Banafsheh Pazokifard

Institution: University of New South Wales
Year: 2016
Posted: 02/05/2017
Record ID: 2073507
Full text PDF: http://handle.unsw.edu.au/1959.4/55858


Pleural plaques (PPs) and diffuse pleural thickening (DPT) are very common asbestos related pleural diseases (ARPD). They are currently identified non-invasively using medical imaging techniques. A fully automatic algorithm for 3D detection of calcified pleura in the diaphragmatic area and thickened pleura on the costal surfaces from multi detector computed tomography (MDCT) images has been developed and tested. The algorithm for detecting diaphragmatic pleura includes estimation of the diaphragm top surface in 3D and identifying those voxels at a certain vertical distance from the estimated diaphragm, and with intensities close to that of bone, as calcified pleura. The algorithm for detecting thickened pleura on the costal surfaces includes: estimation of the pleural costal surface in 3D, estimation of the centrelines of ribs and costal cartilages and the surfaces that they lie on, calculating the mean distance between the two surfaces, and identifying any space between the two surfaces whose distance exceeds the mean distance as thickened pleura. The accuracy and performance of the proposed algorithm was tested on 20 MDCT datasets from patients diagnosed with existing PPs and/or DPT and the results were compared against the ground truth provided by an experienced radiologist. Several metrics were employed and evaluations indicate high performance of both calcified pleura detection in the diaphragmatic area and thickened pleura on the costal surfaces. This work has made significant contributions to both medical image analysis and medicine. For the first time in medical image analysis, the approach uses other stable organs such as the ribs and costal cartilage, besides the lungs themselves, for referencing and landmarking in 3D. It also estimates fat thickness between the rib surface and pleura (which is usually very thin) and excludes it from the detected areas, when identifying the thickened pleura. It also distinguishes the calcified pleura attached to the rib(s), separates them in 3D and detects calcified pleura on the lung diaphragmatic surfaces. The key contribution to medicine is effective detection of pleural thickening of any size and recognition of any changes, however small. This could have a significant impact on managing patient risks. Advisors/Committee Members: Sowmya, Arcot, Computer Science & Engineering, Faculty of Engineering, UNSW, Compton, Paul, Computer Science & Engineering, Faculty of Engineering, UNSW.