gms | German Medical Science

49. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie (gmds)
19. Jahrestagung der Schweizerischen Gesellschaft für Medizinische Informatik (SGMI)
Jahrestagung 2004 des Arbeitskreises Medizinische Informatik (ÖAKMI)

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie
Schweizerische Gesellschaft für Medizinische Informatik (SGMI)

26. bis 30.09.2004, Innsbruck/Tirol

Atrial and ventricular myocardium extraction using model based techniques

Meeting Abstract (gmds2004)

  • corresponding author presenting/speaker Bernhard Pfeifer - Institut für Biomedizinische Signalverarbeitung und Bildgebung, UMIT - Private Universität für Gesundheitswissenschaften, Medizinische Informatik und Technik, Innsbruck, Österreich
  • Friedrich Hanser - Institut für Biomedizinische Signalverarbeitung und Bildgebung, UMIT - Private Universität für Gesundheitswissenschaften, Medizinische Informatik und Technik, Innsbruck, Österreich
  • Christoph Hintermüller - Institut für Biomedizinische Signalverarbeitung und Bildgebung, UMIT - Private Universität für Gesundheitswissenschaften, Medizinische Informatik und Technik, Innsbruck, Österreich
  • Michael Seger - Institut für Biomedizinische Signalverarbeitung und Bildgebung, UMIT - Private Universität für Gesundheitswissenschaften, Medizinische Informatik und Technik, Innsbruck, Österreich
  • Gerald Fischer - Institut für Biomedizinische Signalverarbeitung und Bildgebung, UMIT - Private Universität für Gesundheitswissenschaften, Medizinische Informatik und Technik, Innsbruck, Österreich
  • Robert Modre - Institut für Biomedizinische Signalverarbeitung und Bildgebung, UMIT - Private Universität für Gesundheitswissenschaften, Medizinische Informatik und Technik, Innsbruck, Österreich
  • Bernhard Tilg - Institut für Biomedizinische Signalverarbeitung und Bildgebung, UMIT - Private Universität für Gesundheitswissenschaften, Medizinische Informatik und Technik, Innsbruck, Österreich

Kooperative Versorgung - Vernetzte Forschung - Ubiquitäre Information. 49. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie (gmds), 19. Jahrestagung der Schweizerischen Gesellschaft für Medizinische Informatik (SGMI) und Jahrestagung 2004 des Arbeitskreises Medizinische Informatik (ÖAKMI) der Österreichischen Computer Gesellschaft (OCG) und der Österreichischen Gesellschaft für Biomedizinische Technik (ÖGBMT). Innsbruck, 26.-30.09.2004. Düsseldorf, Köln: German Medical Science; 2004. Doc04gmds091

Die elektronische Version dieses Artikels ist vollständig und ist verfügbar unter: http://www.egms.de/de/meetings/gmds2004/04gmds091.shtml

Veröffentlicht: 14. September 2004

© 2004 Pfeifer et al.
Dieser Artikel ist ein Open Access-Artikel und steht unter den Creative Commons Lizenzbedingungen (http://creativecommons.org/licenses/by-nc-nd/3.0/deed.de). Er darf vervielf&aauml;ltigt, verbreitet und &oauml;ffentlich zug&aauml;nglich gemacht werden, vorausgesetzt dass Autor und Quelle genannt werden.


Gliederung

Text

Introduction

A geometrical model of the atrial and ventricular myocardium is of interest in many fields of biophysics. Atrial and ventricular myocardium contain the electrical sources responsible for the generation of the body-surface ECG [1], [2]. An accurate geometric knowledge of these sources is crucial when dealing with the electrocardiographic forward and inverse problem. In addition, myocardial electrical activity is followed by mechanical contraction of the heart.

Investigating the mechanical functioning is based on precise geometrical models as well as structural comparisons among a cohort of patients. Studying the blood flow in order to get deeper insights in the formation of embolisms requires a realistic heart model including crucial structures like the atrial and ventricular cavities, appendages and ridges. In addition, the formation of atrial fibrillation is influenced by the complex geometry mainly caused by the orifices of inferior and superior vena cava and the pulmonary veins as well as the tricuspid and mitral annuli [3]. The main problem of constructing a realistic heart model is that atrial and right ventricular myocardium can hardly be segmented in volume data because of the low sensitivity and resolution even for state-of-the art medical imaging modalities. The only structures which can be seen with suffcient accuracy are the associated blood masses.

In contrast to the atrial myocardium and the right ventricular myocardium the left ventricular myocardium can be seen pretty accurate. Hence, the extraction of the myocardium of the left ventricle can be done in the same way as used for the right ventricle and the atrium. The following section describes the approach for atrial and ventricular model extraction.

Methods

We have been following an indirect approach for constructing an entire model of a patient's atrial myocardium. The main problem of constructing a realistic atrial model is that atrial myocardium can hardly be segmented in volume data because of the low sensitivity and resolution even for state-of-the-art medical imaging modalities like MRI and CT. The only structure which can be seen with suffcient accuracy is the atrial blood mass. Although the myocardium, especially the left ventricular myocardium, can be seen very well, the indirect reconstruction of the myocardium seems to be a faster method than a direct myocardium extraction. Any technique that is capable of generating an atrial model will succeed also for the ventricle.

The approach is based on a segmentation of the blood masses. The atrial and ventricular myocardium is then constructed artificially by applying appropriate voxel manipulations. After blood mass segmentation the "Label-Voxel-Field" method involves a label-voxel-field manipulation by adding label voxels in the outward normal direction until a user defined wall thickness (normally four millimeters for the atrium and 8 to 10 millimeter for the ventricle) is reached. Therefore the algorithm uses virtual circles with a radius range from a defined minimum wall thikness up to a maximum wall thikness. These circles roll around the blood mass boundary in order to reconstruct the myocardial structure. If the algorithm is able to determine the myocardium using an adaptive threshold that probes all voxles that are element of the virtual circles, then the myocardium can be reconstructed exactly.

In spite of the fact that this method of the algorithm yields good results, the marching cubes algorithm, that is used for surface extraction, prone to produce holes when triangulating thin wall structures. This also occures when the segmentation of the blood masses differs too much between adjacent slides. To overcome this problem, the so called "3D variant" takes one slide above and one slide below the initial labelset slide into account and tries to similarify the adjacent image slides.

Results

We reconstructed the atria and the ventricles of eight patients to benchmark the approach. Figure 1 [Fig. 1] shows the reconstructed myocardial structure of the left and the right atrium.

The segmented labelset is triangulated with a marching cubes algorithm followed by a remeshing process guaranteeing quality standards (equilaterality of triangles) that qualify for a FEM/BEM formulation for solving the electrocardiographic inverse problem. Figure 2 [Fig. 2] shows a triangulated model of a patients heart. Left and middle figure show the atrial myocardium in different perspectives, while the right figure shows the ventricular myocardium including the blood masses.

Discussion

The big defiance to develop the label-voxel-field approach was the need of a semi-automatic segmentation pipeline for enabling noninvasive analysis of cardiac arrhythmias in clinical application.

Due to the nature of the segmentation process, the label-voxel-field manipulation method uses only the volumedata and a labelset stack. This method can, therefore, be directly inserted into a segmentation pipeline for (semi)-automatic volumeconductor modeling. The feature extraction of the label-voxel-field manipulation variant into a 3D variant was necessary in order to prevent the triangulated model from producing holes. This occurs when the segmentation process of the blood masses in adjacent layers yields results that differ too much. The 3D variant of the algorithm tries, for this reason, to similarify the segmented blood masses in adjacent layers by enlarging the smaller segmented blood masses. The big advantage of this procedure is that the triangulation of the myocardium with a marching cubes algorithm produces barely holes than using the 2D version for myocardium extraction alone, notwithstanding that the 3D variant has limitations. One problem is that the segmented blood masses in some layers are adapted by growing the blood masses regardless of the gray-values in the associated image slide. Therefore, it may occur that the expanded labelset grows into the myocardial structure. This problem especially arises in case the quality of the slides does not permit to extract the blood masses exactly. Especially vein-related and other relatively fine structures and special artefacts are involved in producing holes during triangulation.

Acknowledgment

This research study has been funded by the START Y144 program granted by the Austrian Federal Ministry of Education, Science and Culture (bm:bwk) in collaboration with the Austrian Science Fund (FWF).


References

1.
R. Modre, B. Tilg, G. Fischer, and P.Wach, "An iterative algorithm for myocardial activation time imaging," Comput. Methods Programs Biomed. 64, pp. 1-7, 2001.
2.
F. Greensite, "The mathematical basis for imaging cardiac electrical function," Crit. Rev. Biomed. Eng. 22, pp. 347-399, 1994.
3.
E. J. Vigmond, R. Ruckdeschel, and N. A. Trayanova, "Reentry in a morphologically realistic atria," J Cardiovasc Electrophysiol 12(9), pp. 1046-1054, 2001.