Automated Boundary Delimitation Methods in the Image Analysis of Aortic Dissections

Aortic dissection is a rare and dangerous condition associated with high morbidity and mortality, it is very important to reduce the treatment time. This research takes advantage of all sorts of different image analysis techniques, applies them to the aortic dissection images from computer tomography and uses the computer to identify the tissue damage area automatically in order to replace manual segmentation. In this study, five different edge detection algorithms (namely Roberts algorithm, Sobel algorithm, Laplacian algorithm, Marr-Hildreth algorithm and Canny algorithm) have been applied as preprocessor for aortic dissection edge detection. The gold standard and the area obtained by five different edge detection algorithms are analyzed using T test. The results of T test for Roberts, Sobel, Laplacian, Marr-Hildreth and Canny are 0.704, 0.741, 0.771, 0.752 and 0.811, respectively. Average error rate of Roberts, Sobel, Laplacian, Marr-Hildreth and Canny are 13.14%, 13.60%, 12.75%, 13.49% and 10.17%, respectively. The Canny algorithm generates results that are most comparable to manual segmentation. It is a reliable method during research. This information can provide valuable reference for aortic dissection consultation and treatment basis.