Bivariate kernel smoothers. Applications in thoracic aorta pathology.
Aneurysms are the most common pathology that affects the thoracic aorta. They often require surgery due to high risk of rupture, with the consequent death of the patient. Surgical intervention is indicated when maximum diameters exceed 6 cm in the ascending aorta or 7 cm in the descending aorta (Elefteriades, 2002). Hence, the precise description of the thoracic aorta is crucial for the diagnosis of this type of lesion. The aim of the study is to develop robust method for measuring the calibre of the thoracic aorta to detect the presence of abnormalities. For this purpose, a database of 2206 CT (Computed tomography) images of 5 patients from the Department of Radiology of the University Hospital of Santiago de Compostela, were employed.The method involves several steps such as the automatic segmentation of the aorta from CT slices, the calculation of the centre line of the vessel to determine the normal planes of the structure, the smoothness of the data volume to improve the accuracy of the method and the calculus of the diameters. In order to obtain an accurate reconstruction of the whole volume of the aorta, we propose an adapation of the bivariate kernel smoothers (Ruppert and Wand, 1994). Such nonparametric regression techniques allow for a more flexible fit of real data than do the parametric regression techniques usually used.
keywords:
Publication: Congress
1624015028644
June 18, 2021
/research/publications/bivariate-kernel-smoothers-applications-in-thoracic-aorta-pathology
Aneurysms are the most common pathology that affects the thoracic aorta. They often require surgery due to high risk of rupture, with the consequent death of the patient. Surgical intervention is indicated when maximum diameters exceed 6 cm in the ascending aorta or 7 cm in the descending aorta (Elefteriades, 2002). Hence, the precise description of the thoracic aorta is crucial for the diagnosis of this type of lesion. The aim of the study is to develop robust method for measuring the calibre of the thoracic aorta to detect the presence of abnormalities. For this purpose, a database of 2206 CT (Computed tomography) images of 5 patients from the Department of Radiology of the University Hospital of Santiago de Compostela, were employed.The method involves several steps such as the automatic segmentation of the aorta from CT slices, the calculation of the centre line of the vessel to determine the normal planes of the structure, the smoothness of the data volume to improve the accuracy of the method and the calculus of the diameters. In order to obtain an accurate reconstruction of the whole volume of the aorta, we propose an adapation of the bivariate kernel smoothers (Ruppert and Wand, 1994). Such nonparametric regression techniques allow for a more flexible fit of real data than do the parametric regression techniques usually used. - Javier Roca-Pardiñas, Francisco de Asis López-Alvarez, Pablo G. Tahoces, Juan Antonio Martínez-Mera, José M.Carreira
publications_en