Quantifizierung der Morphologie von menschlichen Gefäßen aus 3D tomographischen Bilddaten
Zusammenfassung der Projektergebnisse
Accurate quantification of blood vessels is an important task in medical applications. However, a main difficulty is to cope with the complex and curved 3D anatomy. In this research project, we have developed a new model-based approach for accurate segmentation and quantification of the 3D morphology of the aortic arch and the main bifurcating branches from 3D tomographic image data. The approach combines novel 3D analytic models of the shape and intensity structure of vessels with mathematically well-founded sequential estimation methods. Main advantages of this approach are that more realistic 3D intensity models of vessels are used compared to previous work and that the full image information is exploited. The main application domain is the quantification of the morphology of the aortic arch and the supra-aortic branches in 3D tomographic images for thoracic endovascular aortic repair (TEVAR). Accurate quantification of these vessel structures (e.g., width, curvature, contour lengths) is important for operation planning and treatment of aneurysms using stent grafts with the goal to reduce complications and reinterventions. We have developed different 3D intensity models of blood vessels and have combined them with linear as well as nonlinear sequential filtering approaches (Kalman filter, particle filter). In addition, we have developed algorithms for automatic initialization of the model-based approach as well as for branch detection. This enabled for the first time fully automatic segmentation and quantification of the aortic arch and the supra-aortic branches. We carried out extensive quantitative evaluations using 3D synthetic data, 3D phantom images, and clinical 3D CTA and 3D MRA images, and showed that our approach is robust and quantifies the morphology of vessels with higher accuracy than previous approaches. We have also performed theoretical studies on the performance which support the experimental results. Based on the segmentation and quantification results we have generated a 3D statistical model of the aortic arch and its supra-aortic branches. We have also extended our approach for analysis of dynamic 4D images to allow quantification of the anatomy over time. We considered changes of the morphology of the aortic arch during the heart cycle. Our approach combines spatial and temporal tracking, intensity-based matching, and fitting of 3D intensity models. Compared to previous work, the approach can quantify the orthogonal and longitudinal vessel motion as well as the local vessel diameters. The approach was successfully applied to clinical 4D CTA image sequences. From the results we found that the aortic motion changes in orientation and amplitude along the aorta. The amplitude of the motion decreases with increasing distance from the heart and the aorta expresses a highly twisted motion. To improve the accuracy in the case of pathologies, we have also developed a joint segmentation and registration approach. This approach combines 3D model-based segmentation with elastic image registration and benefits from the robustness of model-based segmentation and the accuracy of elastic registration. The experimental results showed that the approach can cope with a large spectrum of vessel shapes and particularly with pathological shapes. Our approach is the first joint segmentation and registration approach that has been introduced for vessels. An overview of the project was given in Management & Krankenhaus and distributed at the Deutsche R¨ontgenkongreß 2012. Moreover, the project was presented at an exhibition of the DFG in the Deutsche Bundestag (German Federal Parliament), Berlin, in March 2012. Our project was selected as one of ten outstanding DFG projects from about 20,000 projects from all scientific fields. The DFG exhibition has also been presented in other German cities. The work in this project is a significant extension of existing work in medical image analysis and the developed model-based approach should be interesting for medical imaging companies.
Projektbezogene Publikationen (Auswahl)
-
“Segmentation of 3D tubular structures based on 3D intensity models and particle filter tracking”, in J.P.W. Pluim and B.M. Dawant, (eds.), Proc. SPIE Medical Imaging 2009: Image Processing (MI’09), Lake Buena Vista, FL/USA, Feb. 2009, SPIE
S. Wörz, W.J. Godinez, and K. Rohr
-
“3D Quantification of the Aortic Arch Morphology in 3D CTA Data for Endovascular Aortic Repair”, IEEE Transactions on Biomedical Engineering, 57:10, 2359–2368, 2010
S. Wörz, H. von Tengg-Kobligk, V. Henninger, F. Rengier, H. Schumacher, D. Böckler, H.- U. Kauczor, and K. Rohr
-
“Combined Model-Based Segmentation and Elastic Registration for Accurate Quantification of the Aortic Arch”, in T. Jiang, A. Colchester, J. Duncan, M. Viergever, N. Navab, and J. Pluim, Proc. Thirteenth Internat. Conf. on Medical Image Computing and Computer-Assisted Intervention (MICCAI’10), Beijing, China, Sept. 2010, Lecture Notes in Computer Science, Springer Berlin Heidelberg, pages 444-451, 2010
A. Biesdorf, K. Rohr, H. von Tengg-Kobligk, and S. Wörz
-
“Model-Based Segmentation and Motion Analysis of the Thoracic Aorta from 4D ECG-Gated CTA Images”, in G. Fichtinger, A. Martel, and T. Peters, Proc. 14th Internat. Conf. on Medical Image Computing and Computer-Assisted Intervention (MICCAI’11), Toronto, Canada, Sept. 2011, Lecture Notes in Computer Science, Springer Berlin Heidelberg, pages 589-596, 2011
A. Biesdorf, S. Wörz, T. Müller, T.F. Weber, T. Heye, W. Hosch, H. von Tengg-Kobligk, and K. Rohr
-
“Segmentation and Quantification of the Aortic Arch using Joint 3D Model- Based Segmentation and Elastic Image Registration”, Medical Image Analysis, 16, 1187-1201, 2012
A. Biesdorf, K. Rohr, D. Feng, H. von Tengg-Kobligk, F. Rengier, D. Böckler, H.-U. Kauczor, and S. Wörz
-
“True four-dimensional analysis of thoracic aortic displacement and distension using model-based segmentation of computed tomography angiography”, Int. Journal Cardiovascular Imaging, 30:1, 185–194, 2014
T.F. Weber, T. Müller, A. Biesdorf, S. Wörz, F. Rengier, T. Heye, T. Holland-Letz, K. Rohr , H.-U. Kauczor, H. von Tengg-Kobligk
