Elsevier

Medical Image Analysis

Volume 10, Issue 4, August 2006, Pages 657-670
Medical Image Analysis

A comprehensive shape model of the heart

https://doi.org/10.1016/j.media.2006.03.004Get rights and content

Abstract

Domain knowledge about the geometrical properties of cardiac structures is an important ingredient for the segmentation of these structures in medical images or for the simulation of cardiac physiology. So far, a strong focus was put on the left ventricle due to its importance for the general pumping performance of the heart and related functional indices. However, other cardiac structures are of similar importance, e.g., the coronary arteries with respect to diagnosis and treatment of arteriosclerosis or the left atrium with respect to the treatment of atrial fibrillation. In this paper we describe the generation of a geometric cardiac model including the four cardiac chambers and the trunks of the connected vasculature, as well as the coronary arteries and a set of cardiac landmarks. A mean geometric model for the end-diastolic heart has been built based on 27 cardiac CT datasets and has been evaluated with respect to its capability to estimate the position of cardiac structures. Allowing a similarity transformation to adapt the model to image data, cardiac surface positions can be predicted with an accuracy of below 5 mm.

Introduction

Diagnosis and therapy of cardiac diseases is one of the major issues of today’s medicine. Imaging of the cardiac anatomy is addressed by virtually all medical imaging modalities and a considerable portion of clinical interventions concern the heart. From this context arises firstly a demand for preferably non-invasive, accurate diagnosis procedures and secondly a demand for preferably minimally invasive therapeutic procedures. Limited health-care budgets call in both fields for efficient and as much as possible automated procedures. One attempt to facilitate the aforementioned requirements is the intense use of cardiac anatomical domain knowledge within the related computerized procedures. In this paper we describe the construction of a comprehensive end-diastolic cardiac shape model comprising the surfaces of the heart chambers and the respective vascular connection trunks, the coronary arteries, and a set of cardiac landmarks as reference structure as already partially described in Lorenz and von Berg (2005b). The Basis of the model is a set of high-resolution multi-slice CT datasets for surfaces and landmarks, and the geometry of the coronary artery tree as derived by Dodge based on bi-planar X-ray images (Dodge et al., 1988). Landmarks and the three main coronary arteries have been marked manually in the CT dataset. The surface model has been constructed by a semi-automated free-form surface to image adaptation without the need for manual delineation. The registration of the Dodge coronary artery model to the surface and landmark model was realized based on a match of the Dodge model to the delineated coronary arteries. The remainder of the text is organized as follows: Section 2 reviews recent activities of how to extract geometric properties from cardiac images and how to apply them on further image analysis. Section 3 briefly highlights what this paper adds to them. While Section 4 briefly introduces the three components of the model, Section 5 explains in detail the several steps and iterations on model construction, adaptation to images, and statistical analysis. Section 6 reports the variations to be observed in the patient sample set for each of the three model components, and it gives results of cross-validation between them.

Section snippets

Related work

The use of geometric models for cardiac diagnosis started some 40 years ago in the context of functional analysis of the left ventricle from X-ray angiography images. Since then, model based cardiac evaluation procedures have been described for all 2D, 3D or 4D imaging modalities that are capable of imaging the heart, such as echocardiography, magnetic resonance imaging (MRI), single-photon emission computed tomography (SPECT), positron emission tomography (PET), and X-ray computed tomography

Contributions

This paper presents to our knowledge for the first time a comprehensive cardiac shape model comprising all four heart chambers, the trunks of the connected vasculature, the coronary arteries, and a set of cardiac landmarks. The information combined in the model allows exploiting efficiently position information of known structures in order to estimate the position and orientation of unknown structures. For example it enables a landmark based estimation of the location of coronary arteries or

Model overview

The model is intended to serve as a general geometrical reference supporting various tasks associated with cardiac image processing. In order to be able to exploit arbitrary geometrical relationships, we attempted to include into the model all cardiac structures of importance, including a set of cardiac surfaces, the coronary artery tree, and a set of cardiac landmarks.

The surfaces of the blood volume of the left and right ventricle are the basis for functional heart analysis, the surfaces of

Source data

The model described in this paper is based on cardiac Multi-slice computed tomography (MSCT) data. It provides the coverage of the complete heart with high spatial resolution (typically almost isotropic sampling with 0.5 mm voxel size) and sufficient temporal resolution (typically 10 images per cardiac cycle). Coronary arteries can also be assessed well with MSCT (Hoffmann et al., 2004), however, planar fluoroscopy (Movassaghi et al., 2004) is still the method of choice for this task as it

Statistical properties of the model and model verification

The learning set from which the model was built consist of 27 samples. 10 of these samples have been checked and if necessary corrected by a clinical expert, 17 of these samples have been checked and corrected only by the authors. Consequently, we generated three different models: One is based on the 10 expert samples, another one on the 17 non-expert samples and one on all available samples. However, all sample meshes and all triangular surface mesh models are topologically identical. Below,

Conclusions and future work

The generation of a comprehensive geometrical model of the human heart has been described. Currently available is a mean model of the cardiac structures comprising the surfaces of the cardiac chambers and trunks of the connected vasculature, the coronary arteries, and a set of 25 landmarks. The model is based on published data on the coronary arteries and on 27 multi-slice CT datasets. Statistical properties of the model and model based localization of cardiac structures have been evaluated. In

Acknowledgements

We thank our colleagues from PMS-CT Cleveland (USA) and PMS-CT Haifa (Israel) for the abundance of cardiac MSCT image data and for the expert delineations of cardiac structures. We thank them and also our colleagues from PMS-MIT, PMS-MR, and PMS-XRD in Best (The Netherlands) for many fruitful discussions. We also thank our colleagues from the Philips Research Laboratory in Aachen (Germany) for using and improving our cardiac model and for in-depth discussion of modeling and model based

References (48)

  • I.L. Dryden et al.

    Statistical Shape Analysis

    (1998)
  • O. Ecabert et al.

    Towards automatic full heart segmentation in CT Images

  • A.F. Frangi et al.

    Three-dimensional modeling for functional analysis of cardiac images: a review

    IEEE Trans. Med. Imag.

    (2001)
  • A.F. Frangi et al.

    Automatic construction of multiple-object three-dimensional statistical shape models: application to cardiac modeling

    IEEE Trans. Med. Imag.

    (2002)
  • K.D. Fritscher et al.

    Automatic cardiac 4d segmentation using level sets

  • P.E. Gill et al.

    Practical Optimization

    (1997)
  • G. Golub et al.

    Matrix Computation

    (1996)
  • R. Haddad et al.

    A realistic anthropomorphic numerical model of the beating heart

  • C.M. Hoffmann

    Geometric and Solid Modeling

    (1989)
  • M.H.K. Hoffmann et al.

    Noninvasive coronary imaging with MDCT in comparison to invasive conventional coronary angiography: a fast-developing technology

    Am. J. Roentgenol.

    (2004)
  • B.K.P. Horn

    Closed-form solution of absolute orientation using unit quaternions

    J. Opt. Soc. Am.

    (1987)
  • M.R. Kaus et al.

    Automated segmentation of the left ventricle in cardiac MRI

  • M. Kaus et al.

    Automated 3D PDM construction from segmented images using deformable models

    IEEE Trans. Med. Imag.

    (2003)
  • S. Köhn et al.

    Ves: virtual echocardiography system

  • Cited by (114)

    • The evolution of 3D modeling in cardiac disease

      2019, 3-Dimensional Modeling in Cardiovascular Disease
    View all citing articles on Scopus
    View full text