Accurate automatic segmentation and measurement of aortic aneurysm in CT angiography
Détails
ID Serval
serval:BIB_9643E071E50C
Type
Actes de conférence (partie): contribution originale à la littérature scientifique, publiée à l'occasion de conférences scientifiques, dans un ouvrage de compte-rendu (proceedings), ou dans l'édition spéciale d'un journal reconnu (conference proceedings).
Sous-type
Abstract (résumé de présentation): article court qui reprend les éléments essentiels présentés à l'occasion d'une conférence scientifique dans un poster ou lors d'une intervention orale.
Collection
Publications
Institution
Titre
Accurate automatic segmentation and measurement of aortic aneurysm in CT angiography
Titre de la conférence
RSNA 2007, Radiological Society of North America, 93rd Scientific Assembly and Annual Meeting
Adresse
Chicago, Illinois, November 25-30, 2007
Statut éditorial
Publié
Date de publication
2007
Langue
anglais
Résumé
PURPOSE: To evaluate the performance of a new Computer-Aided Detection (CAD) system for AUTOMATIC segmentation and measurement of Aortic Aneurysm (AA).
METHOD AND MATERIALS: Fifty CT Angiography (CTA) data, from which 30 indicated abdominal AA, were used to develop and evaluate the CAD system. Three Rdaiologists manually identified and segmented all AA areas independently using in-house software. For each of the three manual segmentations, an "average" AA region was obtained by taking the overlapping part of the three manual segmentations and the result were used as our "gold standard" for development and evaluation of the AUTOMATIC method. The proposed method first AUTOMATICally identifies and segments the aorta lumen. Once the Aorta and its centreline was found, a series of images perpendicular to its centerline was reconstructed to measure the diameters of aorta. A method was then developed to examine all objects attached to the segmented aorta and whose characteristics lie in a mumber of pre-defined criteria set by incorporating prior understanding of the normal expected variation of aorta, such as its diameters. These regions were considered as candidates for AA regions. The identified AA regions were used to initialise a deformable model-based method used to segment and quantify Thrombus area. The accuracy of the algorithm was validated by comparing the segmentation result against the average "manual segmentation" of three radiologists. Two metrics were calculated to assess the accuracy of the algorithm: the overlapping ratio (OR) and the coverage ratio (CR)
RESULTS: The average values for OR and CR was 0.95 (1 meaning 100% overlap), indicating that the outline segmentation of AA was very close to that defined manually by the radiologists.
CONCLUSION: The manual identification of the enlarged portions of the aorta on a number of cross-sectional images is extremely tedious, time-consuming and subjective process. By using the Computer Aided Detection more ACCURATE, cost effective plus faster result can be achieved.
CLINICAL RELEVANCE/APPLICATION: The progressive growth of AA may eventually cause rupture if not diagnosed or treated.
METHOD AND MATERIALS: Fifty CT Angiography (CTA) data, from which 30 indicated abdominal AA, were used to develop and evaluate the CAD system. Three Rdaiologists manually identified and segmented all AA areas independently using in-house software. For each of the three manual segmentations, an "average" AA region was obtained by taking the overlapping part of the three manual segmentations and the result were used as our "gold standard" for development and evaluation of the AUTOMATIC method. The proposed method first AUTOMATICally identifies and segments the aorta lumen. Once the Aorta and its centreline was found, a series of images perpendicular to its centerline was reconstructed to measure the diameters of aorta. A method was then developed to examine all objects attached to the segmented aorta and whose characteristics lie in a mumber of pre-defined criteria set by incorporating prior understanding of the normal expected variation of aorta, such as its diameters. These regions were considered as candidates for AA regions. The identified AA regions were used to initialise a deformable model-based method used to segment and quantify Thrombus area. The accuracy of the algorithm was validated by comparing the segmentation result against the average "manual segmentation" of three radiologists. Two metrics were calculated to assess the accuracy of the algorithm: the overlapping ratio (OR) and the coverage ratio (CR)
RESULTS: The average values for OR and CR was 0.95 (1 meaning 100% overlap), indicating that the outline segmentation of AA was very close to that defined manually by the radiologists.
CONCLUSION: The manual identification of the enlarged portions of the aorta on a number of cross-sectional images is extremely tedious, time-consuming and subjective process. By using the Computer Aided Detection more ACCURATE, cost effective plus faster result can be achieved.
CLINICAL RELEVANCE/APPLICATION: The progressive growth of AA may eventually cause rupture if not diagnosed or treated.
Création de la notice
09/04/2008 17:17
Dernière modification de la notice
20/08/2019 15:58
Données d'usage
