A survey of shaped-based registration and segmentation techniques for cardiac images

https://doi.org/10.1016/j.cviu.2012.11.017Get rights and content

Highlights

  • This article presents a review summary of the shape modeling applications to cardiac image.

  • The covered modalities are MRI, CT, echocardiography, PET and SPECT images.

  • The methods are classified based on their properties.

  • The article covered methods published in journals within the last 10 years.

Abstract

Heart disease is the leading cause of death in the modern world. Cardiac imaging is routinely applied for assessment and diagnosis of cardiac diseases. Computerized image analysis methods are now widely applied to cardiac segmentation and registration in order to extract the anatomy and contractile function of the heart. The vast number of recent papers on this topic point to the need for an up to date survey in order to summarize and classify the published literature. This paper presents a survey of shape modeling applications to cardiac image analysis from MRI, CT, echocardiography, PET, and SPECT and aims to (1) introduce new methodologies in this field, (2) classify major contributions in image-based cardiac modeling, (3) provide a tutorial to beginners to initiate their own studies, and (4) introduce the major challenges of registration and segmentation and provide practical examples. The techniques surveyed include statistical models, deformable models/level sets, biophysical models, and non-rigid registration using basis functions. About 130 journal articles are categorized based on methodology, output, imaging system, modality, and validations. The advantages and disadvantages of the registration and validation techniques are discussed as appropriate in each section.

Introduction

The heart is the most energetic organ in our body. Beating about every second, it continuously supplies the body with vital oxygen-carrying blood. Heart disease is the leading cause of death in modern countries [1], [2], [3]. The mortality rate of CVD is estimated to be 17 million in 2005 and thus is ranked as the top killer worldwide [3], [4], [5]. According to the AHA update of 2009, CVD is the cause of 10% of days of lost productivity in low- and middle-income countries, and 18% of days of lost productivity in high income countries. CVD morbidity rates are estimated to rise from around 47 million days globally in 1990 to 82 million days in 2020 [4], [5], [6].

Analysis of the cardiac function using imaging instruments has shown to be effective in reducing the mortality and morbidity of CVD. Myocardial motion analysis is time consuming and suffers from inter and intra-observer variability. Computerized analysis can help clinicians to interpret the medical conditions objectively [1], [7], [8], [9]. Cardiac image processing techniques, mainly categorized as segmentation and registration, have been used widely to assess the functionality of the heart [10], [11], [12], [13], [14]. Cardiac image segmentation provides us with high quality structural information of the heart while registration techniques calculate the local functional analysis helpful in diagnosis and treatment planning of patients. Modeling of the cardiac shape, motion and physical structure have played a major role in the development of the image analysis algorithms. Previously, there have been some review papers in the field of computer analysis of cardiac imaging [15], [16]. Some review papers have focused on echocardiography segmentation [17], cine MR segmentation [18] or Tagged MRI [19], [20]. Frangi et al. [16] classified cardiac modeling techniques to three classes: surface models, volume models, and deformable models. The review focused on 3D cardiac modeling techniques based on different modalities namely angiography, cardiac US, isotope imaging, cardiac CT, and MRI. With the increasing number of the cardiac modeling techniques, a review article to summarize recent efforts is timely.

This survey aims to (1) classify major contributions in the field of cardiac modeling, (2) introduce the methodologies in this field, (3) tutor beginners to initiate their research, and (4) introduce the major challenges of registration and segmentation with examples. The techniques developed in the last 10 years are classified as: statistical models, deformable models/level sets, biophysical models, and non-rigid registration methods. The articles are classified in different tables describing the method, database, and output of the technique. The novelties of the methods are described in relevant sections and are compared to alternative algorithms. The advantages and disadvantages of each category are discussed as well and different approaches to validation are classified.

The surveyed articles in this review are from major journals such as IEEE Transaction on Medical Imaging, Medical Image Analysis, IEEE Transaction on Image Processing, IEEE Transaction on Information on Biomedicine, Ultrasound in Medicine and Biology, International Journal of Computer Vision, Computer Vision and Image Understanding, IEEE Transaction on Biomedical Engineering, Cardiovascular Magnetic Resonance, and Journal of Magnetic Resonance in Medicine.

The heart is composed of a muscular contractile organ (myocardium) surrounded by two layers of connective tissue inside and outside called endocardium and epicardium, respectively. The heart has four chambers and four major valves (Fig. 1). LV, the prominent chamber of the heart, is the major contractile chamber, and maintains the systemic circulation. Myocardial contraction is maintained by a circulatory system of coronary arteries that supplies the muscle with oxygenized hemoglobin and nutrients. Coronary arteries (right and left) are two branches of the aorta and supply the myocardium through smaller branches such as LCX, LAD and diagonal arteries [2].

Due to atherosclerosis, the coronary arteries may gradually become occluded and end in CAD (Coronary Artery Disease). Coronary occlusion leads to disturbance in the cardiac contractility and causes global or regional dysfunction in the heart and may be diagnosed using state-of-the-art medical imaging techniques such as echocardiography, MRI, CT, and nuclear medicine [2]. In studying ventricular motion, physicians typically assign a subjective segmental function score to different segments of the ventricles:

  • Normokinesia (0): The myocardial motion and thickening is normal.

  • Hypokinesisa (1): The affected segment moves slower and thickens less than normal.

  • Akinesia (2): The infarcted region has totally lost its ability to contract in the systolic phase and moves passively along with its surrounding myocardial tissue.

  • Dyskinesisa (3): The infarcted region moves paradoxically and bulges out during systole due to the ventricular blood pressure.

  • Aneurysm (4): The infarcted region undergoes remodeling, becomes thin, bulging outwards during the systolic phase like a balloon, leading to rupture and death [9].

The cardiac blood circulation is an alternation of two phases: diastole (relaxation phase) and systole (contraction phase). Normally 70% of the whole LV blood in end diastole is ejected out during systole. The Ejection Fraction (EF) ratio is an index of global LV function. EF is calculated as (EDV–ESV)/EDV where EDV is the volume of the LV at end-diastole and ESV is the volume of the LV during end-systole. Ventricular walls thicken during systole – this is typically referred to as wall thickening and has been proven to be a very reliable index of regional myocardial function. Heart failure is characterized by a significant decrease in the EF. An additional index of cardiac performance is myocardial mass which can be determined from myocardial volume, assuming the myocardium to have uniform density [9].

Section snippets

Cardiac imaging modalities

There are several cardiac imaging modalities that are in widespread use. These include MRI, CT, echocardiography, and nuclear medicine. Each method has advantages and draw backs that are discussed in this section. MRI, CT, and Echocardiography are amenable to computer analysis and much effort has been devoted to automated processing of images from these modalities. In comparison, nuclear medicine has seen less effort devoted to computerized analysis.

Methodological classification

Calculation of important cardiac indices such as cardiac volumes and Ejection Fraction are based on accurate segmentation of the heart chambers while computation of the regional displacements and mechanical indices of cardiac function are related to temporal registration of the imaging data. Segmentation and registration can be performed via techniques that will be discussed in this section.

Bottom up methods: Such as thresholding, morphological, pixel classification and edge-based techniques

Validations

Manual tracking: Manual tracking of the markers has been utilized for the validation of motion detection algorithms. However manual validation is trickier in registration with respect to segmentation. Manual landmark detection is cumbersome in 3D and marker displacements do not represent sub-pixel motion. Finally prominent landmarks are not a good sample of the general tissue displacement because they usually represent edges, corners or higher amount of contrast.

Simulation and phantoms: Several

Conclusions

This article presented a review summary of the shape modeling applications to cardiac image analysis in MRI, CT, echocardiography, PET and SPECT. The main classes surveyed were statistical models, deformable models/level set, biophysical models, and non-rigid registration techniques using basis function methods. The article covered image analysis of cardiac images published in journals within the last 10 years, classified in a number of tables based on various properties such as method,

References (182)

  • T. Makela et al.

    A 3-D model-based registration approach for the PET, MR and MCG cardiac data fusion

    Med. Image Anal.

    (2003)
  • M. Ma et al.

    Model driven quantification of left ventricular function from sparse single-beat 3D echocardiography

    Med. Image Anal.

    (2010)
  • L. Cordero-Grande et al.

    Unsupervised 4D myocardium segmentation with a Markov Random Field based deformable model

    Med. Image Anal.

    (2011)
  • T. Dietenbeck et al.

    Detection of the whole myocardium in 2D-echocardiography for multiple orientations using a geometrically constrained level-set

    Med. Image Anal.

    (2012)
  • S. Grbic et al.

    Complete valvular heart apparatus model from 4D cardiac CT

    Med. Image Anal.

    (2012)
  • Y. Chenoune et al.

    Segmentation of cardiac cine-MR images and myocardial deformation assessment using level set methods

    Comput. Med. Imaging Graph.

    (2005)
  • A. Andreopoulos et al.

    Efficient and generalizable statistical models of shape and appearance for analysis of cardiac MRI

    Med. Image Anal.

    (2008)
  • N. Duchateau et al.

    A spatiotemporal statistical atlas of motion for the quantification of abnormal myocardial tissue velocities

    Med. Image Anal.

    (2011)
  • I. Dydenko et al.

    A level set framework with a shape and motion prior for segmentation and region tracking in echocardiography

    Med. Image Anal.

    (2006)
  • T. Rohlfing et al.

    Evaluation of atlas selection strategies for atlas-based image segmentation with application to confocal microscopy images of bee brains

    NeuroImage

    (2004)
  • R.A. Heckemann et al.

    Automatic anatomical brain MRI segmentation combining label propagation and decision fusion

    NeuroImage

    (2006)
  • B.K.P. Horn et al.

    Determining optical flow

    Artif. Intell.

    (1981)
  • J. Lötjönen et al.

    Statistical shape model of ventricles and epicardium from short- and long-axis MR images

    Med. Image Anal.

    (2004)
  • M. Hernandez et al.

    Registration of anatomical images using paths of diffeomorphisms parameterized with stationary vector field flows

    Int. J. Comput. Vision

    (2009)
  • J. Ashburner

    A fast diffeomorphic image registration algorithm

    Neuroimage

    (2007)
  • T.F. Cootes et al.

    Combining point distribution models with shape models based on finite-element analysis

    Image Vis. Comput.

    (1995)
  • T.F. Cootes et al.

    Active shape models – their training and application

    Comput. Vis. Image Und.

    (1995)
  • D. Cristinacce et al.

    Automatic feature localization with constrained local models

    Pattern Recogn.

    (2008)
  • M.B. Stegmann et al.

    Multi-band modeling of appearance

    Image Vis. Comput.

    (2003)
  • D.L. Kasper et al.

    Harrison’s Principles of Internal Medicine

    (2008)
  • V. Fuster et al.

    Hurst’s the Heart

    (2007)
  • American Heart Association, Heart Disease and Stroke Statistics, Update (at-a-glance version), 2009....
  • Centers for Medicare and Medicaid Services, Office of the Actuary, National Health Statistics Group....
  • Centers for Disease Control and Prevention. Chronic Disease Overview: Costs of Chronic Disease. Centers for Disease...
  • R.A. Rettig

    Medical innovation duels cost containment

    Health Aff.

    (1994)
  • A. Webb

    Introduction to Biomedical Imaging

    (2003)
  • W.R. Hedrick et al.

    Ultrasound Physics and Instrumentation

    (2004)
  • O. Catherine

    Textbook of Clinical Echocardiography

    (2009)
  • D. Sutton et al.

    A Textbook of Radiology

    (2002)
  • D. Fred et al.

    Essentials of Nuclear Medicine Imaging

    (2005)
  • J.A. Sorenson et al.

    Physics in Nuclear Medicine

    (1987)
  • T.L. Faber et al.

    Left ventricular function and perfusion from gated SPECT perfusion images: an integrated method

    J. Nucl. Med.

    (1999)
  • T. Makela et al.

    A review of cardiac image registration methods

    IEEE Trans. Med. Imag.

    (2002)
  • A.F. Frangi et al.

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

    IEEE Trans. Med. Imag.

    (2001)
  • J.A. Noble et al.

    Ultrasound image segmentation: a survey

    IEEE Trans. Med. Imag.

    (2006)
  • E.R. McVeigh et al.

    Imaging myocardial strain

    IEEE Signal Process. Mag.

    (2001)
  • T.F. Cootes et al.

    Active appearance models

    IEEE Trans. Pattern Anal. Mach. Intel.

    (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)
  • J. Koikkalainen et al.

    Methods of artificial enlargement of the training set for statistical shape models

    IEEE Trans. Med. Imag.

    (2008)
  • A. Suinesiaputra et al.

    Automated detection of regional wall motion abnormalities based on a statistical model applied to multislice short-axis cardiac MR images

    IEEE Trans. Med. Imag.

    (2009)
  • Cited by (133)

    • An 8-layer residual U-Net with deep supervision for segmentation of the left ventricle in cardiac CT angiography

      2021, Computer Methods and Programs in Biomedicine
      Citation Excerpt :

      Therefore, the study on the LV segmentation method has important significance and value. Level set and deformable models are widely used for LV segmentation [2]. Vikram et al. [7] proposed a coarse-to-fine system consisting of global localization and local deformations for segmentation of the LV from cardiac CT images.

    View all citing articles on Scopus
    1

    Address: Room 409, Lutz Hall, University of Louisville, Louisville, KY 40292, United States. Fax: +1 502 8764534.

    View full text