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

A Pipeline for Semiautomatic Segmentation, Modelling and Shape Analysis for Cardiac Motion of an Individual Heart

Meeting Abstract (gmds2004)

  • corresponding author presenting/speaker Roland Pilgram - Institute for Medical Knowledge Representation and Visualization, University for Health Sciences, Medical Informatics and Technology Tyrol (UMIT), Innsbruck, Österreich
  • Karl David Fritscher - Institute for Medical Knowledge Representation and Visualization, University for Health Sciences, Medical Informatics and Technology Tyrol (UMIT), Innsbruck, Österreich
  • Michael Franz Schocke - Department of Radiology I, University Hospital of Innsbruck, Innsbruck, Österreich
  • Otmar Pachinger - Department of Cardiology, University Hospital of Innsbruck, Innsbruck, Österreich
  • Rainer Schubert - Institute for Medical Knowledge Representation and Visualization, University for Health Sciences, Medical Informatics and Technology Tyrol (UMIT), 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. Doc04gmds090

Die elektronische Version dieses Artikels ist vollständig und ist verfügbar unter: http://www.egms.de/de/meetings/gmds2004/04gmds090.shtml

Veröffentlicht: 14. September 2004

© 2004 Pilgram 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 vervielfältigt, verbreitet und öffentlich zugänglich gemacht werden, vorausgesetzt dass Autor und Quelle genannt werden.


Gliederung

Text

Introduction

Objective analysis of shape began to emerge as a useful method with the potential to improve the accuracy of medical diagnosis as well as the understanding of processes behind growth and disease. Therefore, over the last years a variety of methods for segmentation and object representations have been suggested for 3D shape analysis.

The first step towards shape analysis of populations of individual objects is the segmentation. The second step is to find an adequate model for shape analysis, and for quantitative statistical analysis. A particular challenging anatomical structure for shape analysis and model based segmentation is the human heart due to its complex structure and cardiac motion, both typically affected by pathologies.

The presented work aims to find an objective way for segmentation and model building to analyze shape and shape variations during the cardiac cycle of the object ensemble consisting of the left ventricle, left atrium, right ventricle, right atrium, and the radices of the aorta and pulmonary trunk of one human heart.

Methods

The analysis is based on a time series of eleven MRI short-axis scans of the heart of one volunteer (male, 22a). These eleven scans started with the r-peak (0 ms) of the electrocardiogram, with a time span of 50 ms to each other, representing a full cardiac cycle.

For the segmentation of the heart we used software developed by our group, providing level-set segmentation algorithms based on filters of the Insight Segmentation and Registration Toolkit (ITK) [1]. The first point in time (0 ms) of the cardiac cycle has been segmented by setting user defined seed points and using geodesic active contour level sets [2] to get the final shape of the chambers of the heart. After minimal manual refinement we used these labels as templates for level-set segmentation of the following points in time of the cardiac cycle.

These labeled data sets for each time step were used for the generation of medial based representations (m-reps) [3], [4]. The concept, application and the key idea of establishing geometric correspondence between different time steps of m-reps is described in detail in [5], [6], [7]. For doing statistical analysis using m-reps the classical principal component analysis (PCA) has been extended by Fletcher et al. [8] to principal geodesic analysis to be valid in figural space.

Results

Level set segmentation has decreased the need for manual user interaction during the segmentation process to a minimum. Therefore the objectiveness of segmentation and the comparability of the labels during the cardiac cycles has been improved.

In overall the PCA distribution represents shape changes over the cardiac cycle. By following consecutive time phase values - characteristic states along the cardiac cycle - a characteristic closed loop curve is revealed. The first 2 main components of the PCA cover a shape space of more than 80% of the variation, the first 6 a shape space of 98%.

Figure 1 [Fig. 1] illustrates the shape variability along the cardiac cycle for six different consecutive time steps (0 ms, 100 ms, 200 ms, 300 ms, 400 ms, 500 ms), regarding the value of the first two principal components. A negative 1st component represents a small atrium and a big ventricle, corresponding to the start of the systole, in this case 0 ms. In contrast a positive 1st component describes a filled -big- atrium, and an emptied -small- ventricle, roughly corresponding to the beginning of the diastole (about 300 ms). In overall the first main component describes mainly the size differences of the two cavities during the cardiac cycle. The 2nd component describes a combination of a twisting and valve plane shifting, which assists the passive refilling process of the chambers. Both of these components describe the main changes of the shape during the cardiac cycle.

Discussion

This combination of semiautomatic segmentation and modeling method using PCA turns out to be an objective powerful pipeline obtaining a fast objective heart model to analyze motion dependent shape changes.

The cardiac cycle, represented by the shape changes obtained from the component analysis, can be described by a quantitative characteristic component trajectory. This spatial trajectory may be an additional valuable description form, besides the well known electrocardiogram for the interpretation of the cardiac cycle.

One of our next steps is to enhance this pipeline and validate the results on a certain number of healthy subjects, probably using different statistical methods to get additional knowledge about growth and disease of the human heart.


References

1.
ITK Homepage; http://www.itk.org.
2.
R. Kimmel, V. Caselles, and G. Sapiro, "Geodesic active contours," IJCV, vol. 22, pp. 61-97, 1997.
3.
S. Joshi, S. Pizer, P. T. Fletcher, P. Yushkevich, A. Thall, and J. S. Marron, "Multiscale deformable model segmentation and statistical shape analysis using medial descriptions," IEEE Trans Med Imaging, vol. 21, pp. 538-50, 2002
4.
S. M. Pizer, D. S. Fritsch, P. A. Yushkevich, V. E. Johnson, and E. L. Chaney, "Segmentation, registration, and measurement of shape variation via image object shape," IEEE Trans Med Imaging, vol. 18, pp. 851-65, 1999
5.
S. M. Pizer, P. T. Fletcher, S. Joshi, A. Thall, J. Z. Chen, Y. Fridman, D. S. Fritsch, G. Gash, J. M. Glotzer, M. R. Jiroutek, C. Lu, K. E. Muller, G. Tracton, P. A. Yushkevich, and E. L. Chaney, "Deformable M-Reps for 3D Medical Image Segmentation," IJCV, vol. 55, 2003.
6.
S. M. Pizer, P. T. Fletcher, S. C. Joshi, G. Gash, J. Stough, A. Thall, G. Tracton, and E. L. Chaney, "A Method & Software for Segmentation of Anatomic Object Ensembles by Deformable M-Reps," submitted for publication, and available via Bibliography/Object Geometry, Statistics, and Segmentation at website http://midag.cs.unc.edu, 2004.
7.
R. Pilgram, K. D. Fritscher, P. T. Fletcher, and R. Schubert, "Shape Modeling of the multiobject organ heart," presented at BioMED, Innsbruck, 2004.
8.
P. T. Fletcher, S. Joshi, C. Lu, and S. Pizer, "Gaussian Distribution on Lie Groups and their application to Statistical Shape Analysis," presented at IPMI, 2003