Skip to main content

Über dieses Buch

This book constitutes the thoroughly refereed post-conference proceedings of the 5th International Workshop on Statistical Atlases and Computational Models of the Heart: Imaging and Modelling Challenges, STACOM 2014, held in conjunction with MICCAI 2014, in Boston, MA, USA, in September 2014. The 30 revised full papers were carefully reviewed and selected from numerous submissions. The papers cover a wide range of topics such as sections on cardiac image processing; atlas construction; statistical modelling of cardiac function across different patient populations; cardiac mapping; cardiac computational physiology; model customization; atlas based functional analysis; ontological schemata for data and results; integrated functional and structural analyses; as well as the pre-clinical and clinical applicability of these methods.



MoCo (Motion Correction) Challenge


Advanced Normalization Tools for Cardiac Motion Correction

We present our submission to the STACOM 2014 MoCo challenge for motion correction of dynamic contrast myocardial perfusion MRI. Our submission is based on the publicly available Advanced Normalization Tools (ANTs) specifically tailored for this problem domain. We provide a brief description with actual code calls to facilitate reproducibility. Time plots and \(K^{trans}\) values, based on the validation methodology of [11], are also provided to determine clinically relevant performance levels.
Nicholas J. Tustison, Yang Yang, Michael Salerno

Deformable Image Registration and Intensity Correction of Cardiac Perfusion MRI

Dynamic contrast Magnetic Resonance myocardial perfusion imaging has evolved into an accurate technique for the diagnosis of coronary artery disease. In this manuscript, we introduce and evaluate the performance of a non-rigid joint multi-level image registration and intensity correction algorithm on a common dataset. An objective functional is formed for which the corresponding Hessian and Jacobian is computed and employed in a multi-level Gauss-Newton minimization approach. In this paper, our experiments are based on elastic regularization on the transformation and total variation on the intensity correction. Our preliminary validations suggest that the registration scheme provides suitable motion correction if the parameters in the algorithm are properly tuned.
Mehran Ebrahimi, Sancgeetha Kulaseharan

Comparison of Linear and Non-linear 2D+T Registration Methods for DE-MRI Cardiac Perfusion Studies

A series of motion compensation algorithms is run on the challenge data including methods that optimize only a linear transformation, or a non-linear transformation, or both – first a linear and then a non-linear transformation. Methods that optimize a linear transformation run an initial segmentation of the area of interest around the left myocardium by means of an independent component analysis (ICA) (ICA-*). Methods that optimize non-linear transformations may run directly on the full images, or after linear registration. Non-linear motion compensation approaches applied include one method that only registers pairs of images in temporal succession (SERIAL), one method that registers all image to one common reference (AllToOne), one method that was designed to exploit quasi-periodicity in free breathing acquired image data and was adapted to also be usable to image data acquired with initial breath-hold (QUASI-P), a method that uses ICA to identify the motion and eliminate it (ICA-SP), and a method that relies on the estimation of a pseudo ground truth (PG) to guide the motion compensation.
Gert Wollny, María-Jesus Ledesma-Carbayo

Motion Correction for Dynamic Contrast-Enhanced CMR Perfusion Images Using a Consecutive Finite Element Model Warping

We present results of a non-rigid registration algorithm to correct breathing motion in cardiac MR perfusion sequences applied to the STACOM 2014 Motion Correction Challenge dataset. The algorithm is based on the finite element method whereby a 2D free form deformation model is deformed to match image features. Image warping is performed through global-to-local mapping of motion parameters. To overcome the contrast intensity problem in the perfusion images, the registration was applied consecutively between adjacent frames. Eleven cases were provided by the challenge: Ten cases were ECG-gated MR perfusion images with rest and adenosine-induced stress series, while the last case was an ungated MR perfusion stress acquisition. The algorithm achieved good results in terms of modified Hausdorff distance: \(1.31\pm 0.93\) pixels (max: 6.5 pixel), horizontal shifting \(< 4.5\) pixels, and vertical shifting: \(< 4\) pixels. Moderate Jaccard index: \(0.57\pm 0.14\) was achieved.
Nils Noorman, James Small, Avan Suinesiaputra, Brett Cowan, Alistair A. Young

Deformable and Rigid Model-Based Image Registration for Quantitative Cardiac Perfusion

Background: Inter-frame image registration is a major hurdle in accurate quantification of myocardial perfusion using MRI. The registration is not standard, in that changing contrast between frames makes it difficult to register the images automatically.
Methods: A multiple step approach was employed. First, a region around the heart was identified out automatically in order to focus the registration. Then we performed rigid shifts between frames with a cross correlation type of method, to obtain a coarse registration. Then we created model images from a two compartment model and an arterial input function from the RV blood pool of the images. These model images represent the uptake and washout of the contrast agent. However they do not contain any motion since the two compartment motion cannot explicitly model motion. These motion-free model images are used as reference images and each frame was registered to its associated model image. Rigid and deformable registration as implemented by ANTS. The entire process was automatic and required ~240 seconds.
This registration approach was tested on the 10 provided ECG-gated rest/stress datasets.
Conclusion: Rigid and deformable registration was performed on the provided datasets. The technique was found to perform better on datasets with higher signal to noise ratio and without sudden respiratory motions.
Devavrat Likhite, Ganesh Adluru, Edward DiBella

Automatic Perfusion Analysis Using Phase-Based Registration and Object-Based Image Analysis

MRI perfusion imaging enables the non-invasive assessment of myocardial blood supply. The purpose of the presented work is to enable a quantitative assessment of the image sequences for clinical application. To this end an automatic preprocessing including ROI detection and outlier removal has been combined with a phase-based registration approach and an object-based myocardium segmentation. The suggested processing pipeline has been tested with 21 image sequences provided by the STACOM motion correction challenge. The corrected image sequences have been assessed by comparison with gamma variate curves fitted to the voxels intensity curves. The automatic segmentation could be compared with expert segmentations provided by the challenge organizers. The results indicate an improvement through the motion correction and a good agreement with the reference segmentation in most cases.
Lennart Tautz, Teodora Chitiboi, Anja Hennemuth

LV mechanics Challenge


Left Ventricular Diastolic and Systolic Material Property Estimation from Image Data

Cardiovascular simulations using patient-specific geometries can help researchers understand the mechanical behavior of the heart under different loading or disease conditions. However, to replicate the regional mechanics of the heart accurately, both the nonlinear passive and active material properties must be estimated reliably. In this paper, automated methods were used to determine passive material properties while simultaneously computing the unloaded reference geometry of the ventricles for stress analysis. Two different approaches were used to model systole. In the first, a physiologically-based active contraction model [1] coupled to a hemodynamic three-element Windkessel model of the circulation was used to simulate ventricular ejection. In the second, developed active tension was directly adjusted to match ventricular volumes at end-systole while prescribing the known end-systolic pressure. These methods were tested in four normal dogs using the data provided for the LV mechanics challenge [2]. The resulting end-diastolic and end-systolic geometry from the simulation were compared with measured image data .
Adarsh Krishnamurthy, Christopher Villongco, Amanda Beck, Jeffrey Omens, Andrew McCulloch

Evaluation of Personalised Canine Electromechanical Models

Cardiac modelling aims at understanding cardiac diseases and predicting cardiac responses to therapies. By generating the electrical propagation, the contraction and the mechanical response, we are able to simulate cardiac motion from non-invasive imaging techniques. Four healthy canine clinical data (left ventricles) were provided by the STACOM’2014 challenge. Our study is based on Bestel-Clement-Sorine mechanical modelling, while the electrophysiological phenomena is driven by an Eikonal model. Our model has been calibrated by a quantitative sensitivity study as well as a personalized automatic calibration. Results and comparison with clinical measures are shown in terms of left ventricular volume, flow, pressure and ejection fraction.
Sophie Giffard-Roisin, Stéphanie Marchesseau, Loïc Le Folgoc, Hervé Delingette, Maxime Sermesant

Connection Forms for Beating the Heart

LV Mechanics Challenge (Methods)
We combine recent work on modeling cardiac mechanics using a finite volume method with the insight that heart wall myofiber orientations exhibit a particular volumetric geometry. In our finite volume mechanical simulation we use Maurer-Cartan one-forms to add a geometrical consistency term to control the rate at which myofiber orientation changes in the direction perpendicular to the heart wall. This allows us to estimate material properties related to both the passive and active parameters in our model. We have obtained preliminary results on the 4 canine datasets of the 2014 mechanics challenge using the FEBio software suite. In ongoing work we are validating and improving the model using rat heart (ex-vivo DTI and in-vivo tagging) MRI datasets, from which we have estimated strain tensors.
Arthur Mensch, Emmanuel Piuze, Lucas Lehnert, Adrianus J. Bakermans, Jon Sporring, Gustav J. Strijkers, Kaleem Siddiqi

Patient–Specific Parameter Estimation for a Transversely Isotropic Active Strain Model of Left Ventricular Mechanics

Computational models are valuable tools for understanding the mechanical function of the heart. In particular, the prospect of doing patient–specific simulations of heart function may have a significant impact on clinical practice. However, patient–specific simulations give rise to severe challenges related to model choices, parameter fitting and model validation. In this study we investigate parameter variability in a model of left ventricular mechanics applied to four different canine heart cases. The mechanics is modeled by a transversely isotropic active strain model, with two parameters adjusted to fit end diastolic and end systolic pressures and volumes. The chosen model is able to accurately reproduce these data, and enables very efficient parameter fitting. Visual inspection of the resulting deformed geometries also shows a reasonable match with the image based reference.
Sjur Gjerald, Johan Hake, Simone Pezzuto, Joakim Sundnes, Samuel T. Wall

Estimation of Diastolic Biomarkers: Sensitiviy to Fibre Orientation

An accurate estimation of myocardial stiffness and decaying active tension is critical for the characterization of the diastolic function of the heart. Computational cardiac models can be used to analyse deformation and pressure data from the left ventricle in order to estimate these diastolic metrics. The results of this methodology depend on several model assumptions. In this work we reveal a nominal impact of the choice of myocardial fibre orientation between a rule-based description and personalised approach based on diffusion-tensor magnetic resonance imaging. This result suggests the viability of simplified clinical imaging protocols for the model-based estimation of diastolic biomarkers.
Sander Land, Steve Niederer, Pablo Lamata

Fully-Coupled Electromechanical Simulations of the LV Dog Anatomy Using HPC: Model Testing and Verification

Verification of electro-mechanic models of the heart require a good amount of reliable, high resolution, thorough in-vivo measurements. The detail of the mathematical models used to create simulations of the heart beat vary greatly. Generally, the objective of the simulation determines the modeling approach. However, it is important to exactly quantify the amount of error between the various approaches that can be used to simulate a heart beat by comparing them to ground truth data. The more detailed the model is, the more computing power it requires, we therefore employ a high-performance computing solver throughout this study. We aim to compare models to data measured experimentally to identify the effect of using a mathematical model of fibre orientation versus the measured fibre orientations using DT-MRI. We also use simultaneous endocardial stimuli vs an instantaneous myocardial stimulation to trigger the mechanic contraction. Our results show that synchronisation of the electrical and mechanical events in the heart beat are necessary to create a physiological timing of hemodynamic events. Synchronous activation of all of the myocardium provides an unrealistic timing of hemodynamic events in the cardiac cycle. Results also show the need of establishing a protocol to quantify the zero-pressure configuration of the left ventricular geometry to initiate the simulation protocol; however, the predicted zero-pressure configuration of the same geometry was different, depending on the origin of the fibre field employed.
Jazmin Aguado-Sierra, Alfonso Santiago, Matias I. Rivero, Mariña López-Yunta, David Soto-Iglesias, Lydia Dux-Santoy, Oscar Camara, Mariano Vazquez

STACOM Challenge: Simulating Left Ventricular Mechanics in the Canine Heart

In this paper we outline our approach for creating subject-specific whole-cycle canine left-ventricular models, as part of the 2014 STACOM Challenge. Each canine heart was modeled using the principle of stationary potential energy, with the myocardium treated as a nearly incompressible hyperelastic material. Given incomplete data on the motion and behavior of each canine heart, we decreased model complexity by employing reduced–parameter constitutive laws. Additionally, base plane motion and left ventricular volume input data were integrated into the cardiac cycle model through the inclusion of novel external energy potentials (using Lagrange multipliers), which allow for relaxed adherence to the constraints and minimize spurious energy modes stemming from model simplification and data noise. Subsequently, using the available data we employ the reduced-order unscented Kalman filter (ROUKF) approach to estimate the myocardial passive parameters and active tension. Finally, along with model predictions for each canine, we assess the spatial convergence and robustness of our model.
Liya Asner, Myrianthi Hadjicharalambous, Jack Lee, David Nordsletten

Identifying Myocardial Mechanical Properties from MRI Using an Orthotropic Constitutive Model

This paper presents a method to characterise the passive orthotropic and contractile properties of left ventricular (LV) myocardial tissue using MRI data of cardiac anatomy, structure and function. Personalised anatomical LV models were fitted to image data from four canine hearts. Diffusion tensor MRI data from the same hearts were parameterised using finite element fitting to provide fibre angle fields that represent longitudinal axes of the myocytes. Fitted fibre angle fields were combined with laminar-sheet orientation data extracted from the Auckland dog heart model and embedded into the customised LV anatomical models. A modified Holzapfel-Ogden orthotropic constitutive relation was parameterised using published data from ex vivo shear tests on myocardial tissue blocks. This parameterised constitutive model was scaled for each case in the present study by fitting the individualised LV models to end-diastolic image data. Contractile tension was then estimated by comparing LV model predictions to the end-systolic image data. Personalised models of this kind can be used to predict the 3D deformation and regional stress distributions throughout the LV wall during the entire cardiac cycle.
Zhinuo J. Wang, Vicky Y. Wang, Sue-Mun Huang, Justyna A. Niestrawska, Alistair A. Young, Martyn P. Nash

Regular papers


Evaluating Local Contractions from Large Deformations Using Affine Invariant Spectral Geometry

We propose a geometric tool for quantifying dense local contractions of the left ventricle given its three dimensional segmented computed tomography (CT). Our approach is based on metric invariants in spectral geometry coupled to a non-rigid alignment algorithm, and can be implemented on data obtained by any modality as only a segmented surface is used. We assume local affine movement of the tissue, and generate a global piecewise constant affine invariant model to regularize the alignment. In contrast to traditional methods which seek diffeomorphic deformation, we show that an isomorphic paradigm can enhance alignment results. We show the superiority of utilizing the proposed metric as part of known non-rigid alignment algorithms on synthesized examples. We further analyzed local contractions and provide statistics for 9 healthy patients and demonstrate abnormal local contractions in atrial fibrillation.
Dan Raviv, Jon Lessick, Ramesh Raskar

Image-Based View-Angle Independent Cardiorespiratory Motion Gating for X-ray-Guided Interventional Electrophysiology Procedures

Cardiorespiratory phase determination has numerous applications during cardiac imaging. We propose a novel view-angle independent prospective cardiorespiratory motion gating technique for X-ray fluoroscopy images that are used to guide cardiac electrophysiology procedures. The method is based on learning coronary sinus catheter motion using principal component analysis and then applying the derived motion model to unseen images taken at arbitrary projections. We validated our technique on 7 sequential biplane sequences in normal and very low dose scenarios and on 5 rotational sequences in normal dose. For the normal dose images we established average systole, end-inspiration and end-expiration gating success rates of 100 %, 97.4 % and 95.2 %, respectively. For very low dose applications, the method was tested on images with added noise. Average gating success rates were 93.4 %, 90 % and 93.4 % even at the low SNR value of \(\sqrt{5}\), representing a dose reduction of more than 10 times. This technique can extract clinically useful motion information whilst minimising exposure to ionising radiation.
Maria Panayiotou, Andrew P. King, R. James Housden, YingLiang Ma, Michael Truong, Michael Cooklin, Mark O’Neill, Jaswinder Gill, C. Aldo Rinaldi, Kawal S. Rhode

Analysis of Mitral Valve Motion in 4D Transesophageal Echocardiography for Transcatheter Aortic Valve Implantation

Transcatheter aortic valve implantation (TAVI) is used to treat aortic stenosis in high-risk patients that cannot undergo cardiac surgery. Because it is minimally-invasive, it could be beneficial to treat patients in better conditions as well. Because their expected lifetime is much longer, the long-term benefit of the TAVI implant must be ensured. If the TAVI stent is placed too far into the left ventricular outflow tract it can impair movement of the anterior mitral leaftlet. Case reports demonstrated endocarditis and leaflet damage due to such friction.
To predict possible complications, we identified mitral valve, aortic valve, and left ventricular outflow tract in 4D transesophageal echocardiography series using model-based segmentation. The segmentation model was a combined structure of the left heart with dynamic valves that was adapted as a whole. Valve dynamics were modeled using shape modes. In a leave-one-patient-out validation of 16 datasets, the respective mean segmentation error for mitral and aortic valve was \(0.99\pm 1.16\,\mathrm {mm}\) and \(1.27\pm 1.68\,\mathrm {mm}\).
We further analyzed the overlap of the mitral leaflet trajectory with the target region for a possible TAVI implant in 18 patients. The overlap as a function of distance from the aortic annulus varied considerably with peak overlaps of 4.7 to 16.6 mm. Such information is potentially useful for procedure planning and device selection to avoid mitral valve impairment by TAVI.
Frank M. Weber, Thomas Stehle, Irina Waechter-Stehle, Michael Götz, Jochen Peters, Sabine Mollus, Jan Balzer, Malte Kelm, Juergen Weese

Structural Abnormality Detection of ARVC Patients via Localised Distance-to-Average Mapping

Many heart conditions result in irregular ventricular shape caused by, for example, increased ventricular pressure, regurgitated blood and poor electrical conduction, which affect the overall function of the heart. Structural abnormalities can be characteristic of a disease. Therefore, identifying structurally abnormal regions can give indicators for diagnosis and can provide useful information to guide long-term therapy planning. Given the difficulty in quantitatively measuring structural abnormalities in patients where the ventricular structure is significantly affected by the pathology, such as patients with arrhythmogenic right ventricular cardiomyopathy (ARVC), a method for computing the distance between a normal geometry and patient-specific geometries is presented. The proposed method involves computing distance maps that can visually emphasise regions with high variation from a normal geometry. A consistent parameterisation of the ventricular shape is imposed using an open-source implementation of the LDDMM algorithm on currents to deform patient-specific geometries to a mean surface, which is also computed using the LDDMM algorithm. The chosen shape parameterisation can be applied to meshes extracted from any segmentation algorithm, allowing a wide range of data to be analysed from different hospitals, different scanners and different imaging modalities. Given a consistent shape parameterisation of all meshes, distance maps can be generated by plotting the Euclidean distance point-wise on a triangulated mesh to visualise regions of high shape variability. The proposed method was applied to 10 ARVC patients to highlight patient-specific shape features.
Kristin McLeod, Marcus Noack, Jørg Saberniak, Kristina Haugaa

Joint Myocardial Motion and Contraction Phase Estimation from Cine MRI Using Variational Data Assimilation

We present a cardiac motion estimation method with variational data assimilation that combines image observations and a dynamic evolution model. The novelty of the model is that it embeds new parameters modeling heart contraction and relaxation. It was applied to a synthetic dataset with known ground truth motion and to 10 cine-MRI sequences of patients with normal or dyskinetic myocardial zones. It was compared to the inTag tagging tracking software for computing the radial motion component, and to the diagnosis for dyskinesia. We found that the new dynamic model performed better than the standard transport model, and the contraction parameters are promising features for diagnosing dyskinesia.
Viateur Tuyisenge, Laurent Sarry, Thomas Corpetti, Elisabeth Innorta-Coupez, Lemlih Ouchchane, Lucie Cassagnes

Segmentation of the Aortic Valve Apparatus in 3D Echocardiographic Images: Deformable Modeling of a Branching Medial Structure

3D echocardiographic (3DE) imaging is a useful tool for assessing the complex geometry of the aortic valve apparatus. Segmentation of this structure in 3DE images is a challenging task that benefits from shape-guided deformable modeling methods, which enable inter-subject statistical shape comparison. Prior work demonstrates the efficacy of using continuous medial representation (cm-rep) as a shape descriptor for valve leaflets. However, its application to the entire aortic valve apparatus is limited since the structure has a branching medial geometry that cannot be explicitly parameterized in the original cm-rep framework. In this work, we show that the aortic valve apparatus can be accurately segmented using a new branching medial modeling paradigm. The segmentation method achieves a mean boundary displacement of 0.6 ± 0.1 mm (approximately one voxel) relative to manual segmentation on 11 3DE images of normal open aortic valves. This study demonstrates a promising approach for quantitative 3DE analysis of aortic valve morphology.
Alison M. Pouch, Sijie Tian, Manabu Takabe, Hongzhi Wang, Jiefu Yuan, Albert T. Cheung, Benjamin M. Jackson, Joseph H. Gorman, Robert C. Gorman, Paul A. Yushkevich

Estimation of Regional Electrical Properties of the Heart from 12-Lead ECG and Images

Computational models of cardiac electrophysiology are being investigated for improved patient selection and planning of therapies like cardiac resynchronization therapy (CRT). However, their clinical applicability is limited unless their parameters are fitted to the physiology of an individual patient. In this paper, a method that estimates spatially-varying electrical diffusivities from routine ECG data and dynamic cardiac images is presented. Contrary to current methods based on invasive electrophysiology studies or body surface potential mapping, our approach relies on widely available 12-lead ECG and motion information obtained from clinical images. First, a map of mechanical activation time is derived from a cardiac strain map. Then, regional electrical diffusivities are personalized such that the computed cardiac depolarization matches both the mechanical activation map and measured ECG features. The fit between measured and computed electrocardiography data after model personalization is evaluated on 14 dilated cardiomyopathy patients, exhibiting low mean errors in terms of the diagnostic ECG features QRS duration (0.1 ms) and electrical axis (10.6\(^{\circ }\)). The proposed regional approach outperforms global personalization when 12-lead ECG is the only electrophysiology data available. Furthermore, promising results of a preliminary CRT study on one patient demonstrate the predictive power of the personalized model.
Philipp Seegerer, Tommaso Mansi, Marie-Pierre Jolly, Dominik Neumann, Bogdan Georgescu, Ali Kamen, Elham Kayvanpour, Ali Amr, Farbod Sedaghat-Hamedani, Jan Haas, Hugo Katus, Benjamin Meder, Dorin Comaniciu

Multi-source Motion Decoupling Ablation Catheter Guidance for Electrophysiology Procedures

Accurate and stable positioning of the ablation catheter tip during the delivery of radiofrequency impulses in cardiac electrophysiology remains a challenge due to the endocardium motion from multiple sources (cardiac cycle and respiration) and inevitable slippage of the catheter tip. This paper presents a novel ablation catheter guidance framework during electrophysiology procedures. Catheter tip electrode position readings from intraoperative electroanatomical data are used to decouple tip motion from different motion sources as part of the pre-ablation mapping. The resulting information is then used to determine if there is relative slippage between the catheter tip and endocardial surface and is shown as a probability map for online decision support of the ablation process. The proposed decomposition method and the slippage assessment were performed on a retrospective cohort of 19 patients treated for ventricular tachycardia (13 cases) or atrial fibrillation (6 cases) and were also validated on artificially generated signals.
Mihaela Constantinescu, Su-Lin Lee, Sabine Ernst, Guang-Zhong Yang

Statistical Model of Paroxysmal Atrial Fibrillation Catheter Ablation Targets for Pulmonary Vein Isolation

Atrial fibrillation (AF) is the most common cardiac arrhythmia. Pulmonary vein isolation (PVI) by catheter ablation is a cornerstone treatment of paroxysmal AF. Low success rates are mainly due to reconnecting tissue. Local myocardial wall-thickness (WT) information is missing; lesion transmurality is impossible to estimate. WT information can be obtained from pencil beam high-resolution MRI, a time-consuming protocol. To reduce scan time, automatic selection of regions of interest is proposed. We developed a left atrial target probability model for paroxysmal AF ablation, based on intraprocedural ablation targeting data of fifteen patients, to support the selection of these regions. A common mesh serves as a reference for registration of the electroanatomical meshes and ablation targets using landmark registration and the Iterative Closest Points algorithm. This is followed by projection of the ablation targets onto the mean mesh model, closure of isolated ablation voids on the surface and Gaussian smoothing of the probability distribution.
The final probability distribution clearly shows PVI contours as suggested in the consensus statement by European associations. The right inferior pulmonary vein (RIPV) shows a lower ablation probability, which may be due to limited maneuverability of the ablation catheter and the proximity of the RIPV ostium and the transseptal puncture, where the catheter enters the left atrium.
Ahmad Al-Agamy, Rashed Karim, Aruna Arujuna, James L. Harrison, Steven E. Williams, Kawal S. Rhode, Hans C. van Assen

Factors Affecting Optical Flow Performance in Tagging Magnetic Resonance Imaging

Changes in cardiac deformation patterns are correlated with cardiac pathologies. Deformation can be extracted from tagging Magnetic Resonance Imaging (tMRI) using Optical Flow (OF) techniques. For applications of OF in a clinical setting it is important to assess to what extent the performance of a particular OF method is stable across different clinical acquisition artifacts. This paper presents a statistical validation framework, based on ANOVA, to assess the motion and appearance factors that have the largest influence on OF accuracy drop. In order to validate this framework, we created a database of simulated tMRI data including the most common artifacts of MRI and test three different OF methods, including HARP.
Patricia Márquez-Valle, Hanne Kause, Andrea Fuster, Aura Hernàndez-Sabaté, Luc Florack, Debora Gil, Hans C. van Assen

Multi-modal Validation Framework of Mitral Valve Geometry and Functional Computational Models

Computational models of the mitral valve (MV) exhibit significant potential for patient-specific surgical planning. Recently, these models have been advanced by incorporating MV tissue structure, non-linear material properties, and more realistic chordae tendineae architecture. Despite advances, only limited ground-truth data exists to validate their ability to accurately simulate MV closure and function. The validation of the underlying models will enhance modeling accuracy and confidence in the simulated results. A necessity towards this aim is to develop an integrated pipeline based on a comprehensive in-vitro flow loop setup including echocardiography techniques (Echo) and micro-computed tomography. Building on [1] we improved the acquisition protocol of the proposed experimental setup for in-vitro Echo imaging, which enables the extraction of more reproducible and accurate geometrical models, using state-of-the art image processing and geometric modeling techniques. Based on the geometrical parameters from the Echo MV models captured during diastole, a bio-mechanical model is derived to estimate MV closure geometry. We illustrate the framework on two data sets and show the improvements obtained from the novel Echo acquisition protocol and improved bio-mechanical model.
Sasa Grbic, Thomas F. Easley, Tommaso Mansi, Charles H. Bloodworth, Eric L. Pierce, Ingmar Voigt, Dominik Neumann, Julian Krebs, David D. Yuh, Morten O. Jensen, Dorin Comaniciu, Ajit P. Yoganathan

Robust Detection of Mitral Papillary Muscle from 4D Transesophageal Echocardiography

Mitral valve (MV) diseases, one of the most common valvular diseases, often require surgical repair to reduce mitral regurgitation and improve cardiac pump function. These procedures however are very complex and require careful planning. In particular, chordae replacement or sub-valvular repair demands a precise assessment of the relative position of the papillary muscles with respect to the leaflets in the beating heart. This can be achieved only before opening the chest through imaging like computerized tomography or trans-esophageal echocardiography (TEE). Yet, quantitative analysis of the MV structure and dynamics, in particular the papillaries, is still tedious and prone to user variability. This manuscript presents a novel approach to automatically detect and track papillary muscle tips in 4D TEE. The proposed data-driven method combines the Marginal Space Learning method with Random Sample Consensus and Belief Propagation cope with varying image quality and signal drop-offs. Experiments on 30 randomly-selected volumes show that the accuracy of our algorithm falls within inter-rater variability (5.58mm out of 6.94mm for the anterior tip and 5.75mm out of 7.06mm for the posterior tip), while being extremely fast (under 3 seconds). The proposed method could therefore provide the surgeon with quantitative MV evaluation for optimal therapy planning.
Mihai Scutaru, Ingmar Voigt, Tommaso Mansi, Anand Tatpati, Razvan Ionasec, Helene Houle, Dorin Comaniciu

Reusability of Statistical Shape Models for the Segmentation of Severely Abnormal Hearts

Statistical shape models have been widely employed in cardiac image segmentation. In practice, however, the construction of the models is faced with several challenges, in particular the need for a sufficiently large training database and a detailed delineation of the training images. Moreover, for pathologies that induce severe shape remodeling such as for pulmonary hypertension (PH), a statistical model is rarely capable of encoding the significant and complex variability of the class. This work presents a new approach for the segmentation of abnormal hearts by reusing statistical shape models built from normal population. To this end, a normalization of the pathological image data is first performed towards the space of the normal shape model, which is then used to guide the segmentation process. Subsequently, the model recovered in the space of normal anatomies is propagated back to the pathological images space. Detailed validation with PH image data shows that the method is both accurate and consistent in its segmentation of highly remodeled hearts.
Xènia Albà, Karim Lekadir, Corné Hoogendoorn, Marco Pereanez, Andrew J. Swift, Jim M. Wild, Alejandro F. Frangi

Registration of Real-Time and Prior Imaging Data with Applications to MR Guided Cardiac Interventions

Recently, there has been increased interest in using magnetic resonance imaging (MRI) to guide interventional procedures due to its excellent soft tissue contrast and lack of ionizing radiation. One of the applications is the use of MRI to guide radio-frequency (RF) ablations for the treatment of cardiac arrhythmia. However, MRI is challenging as there exists significant tradeoffs between the imaging quality and acquisition time. High quality, pre-operative 3D MR images can be acquired with excellent spatial resolution at the expense of long acquisitions. Alternatively, 2D real-time MR imaging during the intervention sacrifices image quality for the ability to visualize dynamic motion of the heart. Therefore, to improve the MRI guidance capabilities for cardiac interventions, we propose a novel registration method to align the real-time and prior imaging data, which corrects for motion errors between the two datasets. The proposed method uses a hybrid metric within a multi-resolution registration framework to achieve the desired clinical accuracy for cardiac interventions. Registration experiments were performed with in vivo human images, and the mean alignment error between real-time and prior images after registration was 3.91 \(\pm \) 1.52 mm.
Robert Xu, Graham A. Wright

Restoration of Phase-Contrast Cardiovascular MRI for the Construction of Cardiac Contractility Atlases

Cardiac Atlases are promising tools for the interpretation of functional and anatomical structures of the heart. Myocardial viability is reflected by both global and regional contractile abnormalities. Atlases incorporating contractility information of a population can assist the diagnosis of myocardial disease and myocardial infarction. For the analysis of myocardial contractility phase-contrast MRI (PC-MRI) is emerging as a valuable clinical tool. The myocardial velocity distribution depicted by PC-MRI provides important insights into the intrinsic mechanics of the heart. As with many imaging techniques, there is an inherent trade-off between imaging resolution and noise. The main purpose of this study is to reduce the noise exhibited in phase-contrast MRI by applying a total variation restoration algorithm. The restoration algorithm has been evaluated on a spiral phase-contrast MRI sequence from a group of normal subjects. The results have shown that the proposed method is able to restore the myocardial velocity distribution whilst preserving the fidelity of the underlying contractile behavior.
Christina Koutsoumpa, Robin Simpson, Jennifer Keegan, David Firmin, Guang-Zhong Yang

Manifold Learning for Cardiac Modeling and Estimation Framework

In this work we apply manifold learning to biophysical modeling of cardiac contraction with the aim of estimating material parameters characterizing myocardial stiffness and contractility. The set of cardiac cycle simulations spanning the parameter space of myocardial stiffness and contractility is used to create a manifold structure based on the motion pattern of the left ventricle endocardial surfaces. First, we assess the proposed method by using synthetic data generated by the model specifically to test our approach with the known ground truth parameter values. Then, we apply the method on cardiac magnetic resonance imaging (MRI) data of two healthy volunteers. The post-processed cine MRI for each volunteer were embedded into the manifold together with the simulated samples and the global parameters of contractility and stiffness for the whole myocardium were estimated. Then, we used these parameters as an initialization into an estimator of regional contractilities based on a reduced order unscented Kalman filter. The global values of stiffness and contractility obtained by manifold learning corrected the model in comparison to a standard model calibration by generic parameters, and a significantly more accurate estimation of regional contractilities was reached when using the initialization given by manifold learning.
Radomir Chabiniok, Kanwal K. Bhatia, Andrew P. King, Daniel Rueckert, Nic Smith


Weitere Informationen

Premium Partner