Functional Imaging and Modeling of the Heart

Arrhythmias are often associated with healing infarcts and could arise from the border zone of the scars. The main purpose of this work was to characterize the infarct scars using in vivo electro-anatomic CARTO maps (recorded in sinus rhythm) and high-resolution ex-vivo MR images in a porcine model of chronic infarct. The MR images were segmented into scar, peri-infarct and healthy ventricular tissue, and, in select slices, the results of segmentation were validated against histology. Further, the segmented volumes and associated fiber directions (derived from diffusion-weighted (DW) MRI as well as from synthetic models), were used as input to a simple two-variable mathematical model that calculates the propagation of depolarization waves and isochronal maps; and these isochronal maps were compared to the measured ones. We further correlated the size of the scar measured during the electrophysiology (EP) study with scar dimensions obtained from MRI using ex-vivo DW-MRI methods. Finally, we present preliminary results from a qualitative comparison between the scar delineation from ex vivo and in vivo MR images.

Detailed knowledge of spatial-temporal behaviours of intracellular calcium dynamics is very important in understanding excitation-contraction coupling of cardiac myocytes under both normal and pathological conditions, such as initiation and propagation of spontaneous calcium waves. A full cell simulation integrating about 10,000 Ca

2 + 

release units (CRUs) in a typical cardiac myocyte is considered a multi-scale, computationally demanding problem. In this paper, we imported an experimentally obtained spatial distribution of Ryanodine Receptor (RyR) clusters into a spatially extended, stochastic model of intracellular Ca

2 + 

dynamics, to investigate the role of the structural bifurcation of Z-disks on the initiation and propagation of intracellular Ca

2 + 

waves from spatially symmetric Ca

2 + 

sparks. Besides, we also proposed a simple parallelization strategy to increase computational speed for this chanllenging problem in cardiac modelling.

In this paper, we present a methodology to estimate cardiac motion directly from the high resolution temporal data provided by an intra-cavity electrical mapping system (CARTO). These data consist in intracardiac electrical measurements, obtained invasively through the contact of a catheter with the endocardial wall at different locations, with simultaneous recording of the position of the measuring catheter over time. The 3D displacement fields between the different timepoints obtained from the position measurements are projected onto the vector pointing from the CARTO points to the centroid of the CARTO cloud, giving a very intuitive vectorial coarse representation of the diastolic and systolic motion. Furthermore, scalar projection 1D curves can be used to identify specific motion patterns. We have applied the proposed methodology to the CARTO acquisitions of nine candidates to cardiac resynchronization therapy, identifying the specific sequence of motion and deformation (septal flash) found in LBBB, which was confirmed by visual inspection of the corresponding MR and 3D-US images.

Patient-specific model adaptation and validation requires a comparison of simulations with measured patient data. For patients suffering from atrial fibrillation, such data is mainly available as intracardiac catheter signals. In this work, we demonstrate the simulation of clinically relevant catheter data as measured using circular mapping catheters (such as Lasso

®

or Orbiter

®

) and coronary sinus catheters using atrial simulations on a realistic geometry. Four circular catheters are modeled using a projection technique for two distinct types of application. We show that in sinus rhythm, the choice of a distinct electrophysiological model does not impair the signal quality. Finally, we compare simulated potentials to a real clinical measurement. In the future, with patient-specific models available, such comparisons can constitute an important interface for personalizing cardiac models.

Cardiac fibre architecture plays a key role in heart function. Recently, the estimation of fibre structure has been simplified with diffusion tensor MRI (DT-MRI). In order to assess the heart architecture and its underlying function, with the goal of dealing with pathological tissues and easing inter-patient comparisons, we propose a methodology for finding cardiac myofibrille trace correspondences across a fibre population obtained from DT-MRI data. It relies on the comparison of geometrical and topological clustering operating on different fibre representation modes (fixed length sequences of 3-D coordinates with or without ordering strategy, and 9-D vectors for trace shape approximation). In geometrical clustering (or k-means) each fibre path is assigned to the cluster with nearest barycenter. In topological (or spectral) clustering the data is represented by a similarity graph and the graph vertices are divided into groups so that intra-cluster connectivity is maximized and inter-cluster connectivity is minimized.

Using these different clustering methods and fibre representation modes, we predict different fibre trace classifications for the same cardiac dataset. These classification results are compared to the human heart architecture models proposed in the literature.

Early models of rabbit cardiac fibre structure where from fitting histological fibre orientation onto finite element models. More recently models have been produced using DT-MRI. In a quantitative comparison of these models, in a selected equatorial slice the fibre helix angle has a transmural change of -111.8±30.8º (mean linear fit ± S.D.) in the histological data [H-1]), -92.4±54.5º in DT-MRI dataset [DTI-1] and -86.3±30.7º in DT-MRI dataset [DTI-2]. Variation is large due to outlier data near the RV posterior insertion, and is less when selected anatomical transmural locations are quantified; the lateral LV has a monotonic transmural change of H-1:-87.5±3.7º (mean linear fit ± S.E of slope); DTI-1: -84.8±2.2º; DTI-2:-77.7±2.5º. There is greater variation in the transmural change of the transverse angle than the helix angle (DTI-1, 4.6±83.0º (mean linear fit ± S.D.), pooled data). Limitations in the datasets from both methodologies are discussed in the light of this analysis.

Mechanical load plays an important role in cardiac growth and remodeling. In a previous mathematical model study the hypothesis was tested that fiber cross-fiber shear acts as a stimulus for remodeling of the pattern of myofiber orientations [7]. It was found that reorientation in response to fiber cross-fiber shear yielded realistic longitudinal and transmural components of myofiber orientation. In this study, we investigated whether circumferential-radial shear, as determined experimentally with magnetic resonance tagging, can be reproduced in the model as well. Our results show that after myofiber reorientation differences in circumferential-radial shear between model and experiment were reduced significantly. However, a complete match could not be obtained through shear-induced myofiber reorientation alone.

Whole heart computer models are being widely used to assess treatment planning like Cardiac Resynchronization Therapy (CRT), a recognized treatment for heart failure. Certain aspects like cardiac geometry and the Purkinje system (PS) are still sometimes neglected on the studies. The present study includes a model of the human ventricles with incorporated anatomical structures such as myofibers orientation and Purkinje system. Two meshes representative of hypertrophic and dilated cardiomyopathies were generated from an original healthy subject mesh. In the context of (III) atrio-ventricular (AV) block, a sequential pacing protocol was tested for different device configurations. The pacing protocols were performed in models including the PS and in models lacking the PS. The results show that the Purkinje System leads to the synchronization of the septum and the left ventricle (LV) lateral wall, thus minimizing the interventricular delay. Also the modifications in the geometry resulted in changes in the activation patterns of the LV lateral wall, differing notably from physiological scenario.

Denoising and interpolation of primary vector fields in DT-MRI are essential for tracking myocardial fibers of the human heart. In this paper, a noise-reduced interpolation method for 3-D primary vector fields in human cardiac DT-MRI is proposed. The method consists of first localizing the noise-corrupted vectors using local statistical properties of the vector fields, then restoring the noise-corrupted vectors by means of Thin Plate Spline (TPS) interpolation method, and finally applying a global TPS interpolation to gain higher resolution in the spatial domain. Experiments and results show that the proposed method allows us to obtain higher resolution and reduce noise, while improving direction-coherence (

DC

) of vector fields, preserving details, and improving fiber tracking.

The anisotropic electrical conduction within myocardial tissue due to preferential cardiac myocyte orientation (‘fibre orientation) is known to impact strongly in electrical wavefront dynamics, particularly during arrhythmogenesis. Faithful representation of cardiac fibre architecture within computational cardiac models which seek to investigate such phenomena is thus imperative. Drawbacks in derivation of fibre structure from imaging modalities often render rule-based representations based on a priori knowledge preferential. However, the validity of using such rule-based approaches within whole ventricular models remains unclear. Here, we present the development of a generic computational framework to directly compare the fibre architecture predicted by rule-based methods used within whole ventricular models against fibre structure derived from DTMRI data, and assess how relative differences influence propagation dynamics throughout the ventricles. Results demonstrate the close overall match between the methods within the rat ventricles, and highlight regions for potential rule-adaption.

To treat defects within the heart, surgeons currently use stopped-heart techniques. These procedures are highly invasive and incur a significant risk of neurological impairment. We are developing methods for performing surgery within the heart while it is beating. New real-time 3-D ultrasound imaging allows visualization through the opaque blood pool, but this imaging modality poses difficult image processing challenges, including poor resolution, acoustic artifacts, and data rates of 30 to 40 million voxels per second. To track instruments within the heart we have developed a Radon transform-based algorithm, which is readily implemented in real-time on graphics processor units. For manipulation of rapidly moving cardiac tissue we have created a fast robotic device that can track the tissue based on ultrasound image features. This allows the surgeon to interact with the heart as if it was stationary. To integrate ultrasound imaging with the robotic device we have developed a predictive controller that compensates for the 50-100 ms imaging and image processing delays to ensure good tracking performance. Our in vitro studies show that this approach enhances dexterity and lowers applied forces. In vivo applications of this technology in atrial septal defect closure and mitral valve annuloplasty procedures demonstrate the potential for improved patient outcomes.

In this paper, intravascular ultrasound (IVUS) grayscale images, acquired with a single-element mechanically rotating transducer, are processed with wavelet denoising and region-based segmentation to extract various layers of lumen contours and plaques. First, IVUS volumetric data is expanded on complex exponential multi-resolution basis functions, also known as Brushlets, which are well localized in the time and frequency domains. Brushlet denoising has previously demonstrated a great aptitude for denoising ultrasound data and removal of blood speckle. A region-based segmentation framework is then applied for detection of lumen border layers, which remains a challenging problem in IVUS image analysis for images acquired with a single element, mechanically rotating 45 MHz transducer. We evaluated a hard thresholding operator for Brushlet denoising, and compared segmentation results to manually traced lumen borders. We observed good agreement and suggest that the proposed algorithm has a potential to be used as a reliable pre-processing step for accurate lumen border detection.

The presence of an infarct scar in the heart generates abnormal electrical pathways that may trigger the occurrence of arrhythmic episodes. While precise models of the electric propagation in the heart have been proposed, we are just starting to observe and analyze infarct scars using high-resolution imaging techniques. Recent observations have shown that the scar is a highly heterogeneous tissue, characterized by a complex interface with surrounding myocardium. For instance, the infarct scar is perforated by tunnels of live tissue, which could generate abnormal activation pathways and therefore facilitate arrhythmia episodes. In order to characterize the role of such structures, we need to first delineate them. In this paper, we propose an automatic method for the detection of these tunnels of normal tissue through scars in high resolution MR images.

During interventional procedures, cardiologists revascularize the ischemic heart by opening occluded arteries. Patency of the artery is assessed by selective injection of dye. Reperfusion of the myocardium is more difficult to estimate in the dim signal produced by the contrast agent in the microcirculation. Some researchers [1-2] have proposed and studied qualitative evaluations based on the observation of contrast appearance and wash-out in the heart muscle. In this paper, we describe a set of tools designed to assist the cardiologist in the assessment of myocardial perfusion. First, we apply synchronized subtraction in order to enhance the tissue opacification by removing the background. Secondly, to quantify the microcirculation temporal behavior, we compute time-density curves normalized with respect to the injection characteristics. Thirdly, we create parametric images that summarize in a single picture the dynamic information directly usable in the interventional setting.

Real-time three-dimensional echocardiography (RT3DE) permits the acquisition and visualization of the beating heart in 3D. However, its actual utility is limited due to missing anatomical structures and limited field-of-view (FOV). We present an automatic two-stage registration and fusion method to integrate multiple single-view RT3DE images. The registration scheme finds a rigid transformation by using a multiresolution algorithm. The fusion is based on the 3D wavelet transform, utilizing the separation of the image into low- and high-frequency wavelet subbands. The qualitative and quantitative results, from 12 subjects, demonstrate that the proposed fusion framework helps in: (i) filling-in missing anatomical information, (ii) extending the FOV, and (iii) increasing the structural information and image contrast.

Atrial fibrillation (AF) is the most common cardiac arrhythmia in the western world. Genetic variants in the cardiac I

Kr

channel have been identified to influence ventricular repolarization. The aim of this work is to investigate the effect of the mutation N588K on atrial repolarization and the predisposition to AF. Experimental data of N588K mutated hERG channel were incorporated in an atrial ionic model using parameter fitting. The effects of the mutation were analyzed in cell and tissue. N588K showed a gain of function effect, causing a rapid repolarization and a shortening of the action potential duration. Computer simulations of a schematic right atrial geometry were used to investigate the excitation conduction properties. The effective refractory period of mutant cells were reduced from 317 to 233

ms

at 1

Hz

. The conduction velocity is not significantly influenced by the mutation. Nevertheless, the wavelength of mutant cells is for all frequencies smaller, indicating that the mutation N588K predisposes the initiation and perpetuation of AF.

Experimentally observed differences in the action potential (AP) properties – primarily, refractoriness – between the left (LA) and right (RA) atria are believed to be important in maintaining atrial fibrillation. We incorporate AP models for single LA and RA cells into 2D atrial tissue models and study the role of tissue heterogeneity in global interaction between reentrant spiral waves in the LA and RA. Our simulations show that shorter refractoriness in the LA translates into a shorter period of spiral rotation, and as a result, reentry in the LA dominates the overall excitation patterns in the atria.

The foremost premise for the success of noninvasive volumetric myocardial transmembrane (TMP) imaging from body surface potential (BSP) recordings is a realistic yet efficient electrocardiographic model which relates volumetric myocardial TMP distribution to BSP distribution. With a view towards this inverse problem, the heart-torso representation and the associated TMP-to-BSP model should balance the accuracy of the model with the feasibility of TMP reconstruction. We present a novel coupled meshfree-BEM platform to this TMP-to-BSP modeling and perform electrocardiographic simulations of various cardiac conditions on personalized heart-torso structures. Simulated QRS integral maps, BSPMs and ECG leads are consistent with existent experimental studies, verifying the plausibility of the presented platform.

The mutations to hERG (the human Ether-a-go-go Related Gene) that cause long QT syndromes produce effects on the rapid delayed rectifier K

 + 

current

I

Kr

and, therefore, action potential duration (APD). These mutations can affect various properties that determine

I

Kr

kinetics. We used computational models of human ventricular myocytes to identify which of these properties, when altered, cause profound changes to APD and transmural dispersion of repolarisation (TDR). Such increases in both APD and TDR is caused by a positive shift of activation

V

0.5

, a negative shift of inactivation

V

0.5

, or by reducing maximal conductance. The largest reduction in APD is achieved by a positive shift of inactivation

V

0.5

. Altering the time constant of activation had relatively little effect. When two or more parameters were altered simultaneously, shifting inactivation

V

0.5

had the dominant effect on APD, except for some extreme shifts of activation

V

0.5

or moderate reductions of maximal conductance. HERG mutations observed clinically lie in the parameter range where maximal conductance has the dominant effect. Bifurcation analysis showed stable steady states (corresponding to physiological resting membrane potential) at all parameter values, and no APD alternans. We conclude that increased APD due to hERG mutations seen clinically are a combined effect of alterations to

I

Kr

kinetic parameters that, in isolation, cause either shortening or prolongation of the AP. Therapeutics that alter

I

Kr

conductance are potentially most beneficial.

In this paper, an adaptive meshless method is described for solving the modified FitzHugh Nagumo equations on a set of nodes directly imported from the voxels of the medical images. The non-trivial task of constructing suitable meshes for complex geometries to solve the reaction-diffusion equations is circumvented by a meshfree implementation. The spatial derivatives arising in the reaction diffusion system are estimated using the Lagrangian form of scattered node radial basis function interpolant. Normal cardiac activation phenomena is fast, with a very steep upstroke and localised as compared to the size of the computational domain. To accurately capture this phenomena, a space adaptive method is presented where extra nodes are placed near the region of the activation front. The performance of the adaptive method is investigated first for synthetic geometry and then applied to a real-life geometry obtained from magnetic resonance imaging. Numerical results suggest that the presented method is capable of predicting realistic electrophysiology simulation effectively.

The heart is a dynamic organ and important characteristics about its function can be derived from 4D (3D + time) data. In this paper we present an approach for the automatic local tracking of the cardiac surfaces with sub-voxel accuracy in time series of computed tomography image volumes. First, a multi-compartment mesh is adapted to a single reconstructed image. This adapted mesh is then propagated through the whole time series. Chamber volumes and local surface motion can be estimated at arbitrary time points. Experiments on 15 time series indicate that the system is very precise and robust with respect to the selection of the start phase of the propagation. The results can be used, e.g., for dyssynchronity assessment or comparison with electromechanical models of the full heart.

Analysing heart motion provides crucial insights on the condition of the cardiac function. Tagged-MRI and 2D-strain ultrasound enable quantitative assessment of the myocardium strain. But estimating 3D myocardium strain from cine-MRI remains attractive: cine-MRI is widely available and it yields detailed 3D+t anatomical images. This paper presents an image-based method to estimate myocardium strain from clinical short-axis cine-MRI. To recover non-apparent cardiac motions, we improve the diffeomorphic demons, a non-linear registration algorithm, by adding two physical constraints. First, myocardium near-incompressibility is ensured by constraining the deformations to be divergence free. Second, myocardium elasticity is modelled using smooth vector filters. The proposed physically-constrained demons are compared with the diffeomorphic demons and evaluated in a healthy subject against tagged MRI. The method is also tested on a patient with congenital pulmonary valve regurgitations and compared with 2D-strain measurements. In both cases, obtained results correlate well with ground truth. This method may become a useful tool for cardiac function evaluation.

Real-time three-dimensional echocardiography (RT3DE) offers an efficient way to obtain complete 3D images of the heart over an entire cardiac cycle in just a few seconds. The complex 3D wall motion and temporal information contained in these 4D data sequences has the potential to enhance and supplement other imaging modalities for clinical diagnoses based on cardiac motion analysis. In our previous work, a 4D optical flow based method was developed to estimate dynamic cardiac metrics, including strains anddisplacements, from 4D ultrasound. In this study, in order to evaluate the ability of our method in detecting ischemic regions, coronary artery occlusion experiments at various locations were performed on five dogs. 4D ultrasound data acquired during these experiments were analyzed with our proposed method. Ischemic regions predicted by the outcome of strain measurements were compared to predictions from cardiac physiology with strong agreement. This is the first direct validation study of our image analysis method for biomechanical prediction and

in vivo

experimental outcome.

Starting from the presentation of a unified perspective of segmentation with deformable models and data assimilation with images, we propose a data assimilation procedure designed to dynamically estimate 3D positions and velocities of the myocardium along the heart cycle, using data consisting of contours extracted from image sequences. We assess this procedure with a test problem based on a realistic computational heart model, and with synthetic data produced from reference simulations by creating binary images of the myocardium. The automatic meshing library CGAL is then employed to create contour meshes for each snapshot, and these meshes are directly used in the model-measurement comparisons. This approach gives very satisfactory qualitative and quantitative results.

Meaningful physiological models are important for studying cardiac physiology. For cardiac image analysis, the models used should be detailed enough to describe the macroscopic physiological behaviors, but should not be too complicated for the inverse problems. To achieve this goal, we propose to use an orthotropic hyperelastic biomechanical model, which has only seven parameters but was reported as the best among the five tested well-known models in a comparative study. Combining with the active contraction forces provided by electromechanical models, the cyclic cardiac dynamics can be available to provide the physiological foundation for cardiac image analysis. To facilitate the complicated inverse problems, the system is implemented under the Cartesian coordinate system, and we propose the corresponding cardiac specific boundary conditions for anatomically realistic deformations. Experiments with analytical solutions were performed to verify the correctness of the implementation, and cardiac cycles were simulated to verify the physiological plausibility of the proposed model.

Cardiac resynchronisation therapy (CRT) has been shown to be an effective adjunctive treatment for patients with dyssynchronous ventricular contraction and symptoms of the heart failure. However, clinical trials have also demonstrated that up to 30% of patients may be classified as non-responders. In this article, we present how the personalisation of an electromechanical model of the myocardium could help the therapy planning for CRT. We describe the four main components of our myocardial model, namely the anatomy, the electrophysiology, the kinematics and the mechanics. For each of these components we combine prior knowledge and observable parameters in order to personalise these models to patient data. Then the acute effects of a pacemaker on the cardiac function are predicted with the

in silico

model on a clinical case. This is a proof of concept of the potential of virtual physiological models to better select and plan the therapy.

In pulmonary arterial hypertension (PH), increase of myofiber contraction duration in the right ventricular (RV) free wall as compared to that in the left ventricular (LV) free wall leads to mechanical asynchrony. We used the lumped TriSeg model describing ventricular mechanics to determine effects of increasing pulmonary resistance and of changing time of RV free wall activation on ventricular pump mechanics and on myofiber mechanics in the ventricular walls. Simulated circumferential strain in the ventricular walls under normal and PH conditions agreed with measurements in patients. When the RV free wall was early activated with respect to the LV free wall and septum, simulations showed decreased RV volume and decreased myofiber work in the RV free wall whereas LV free wall and septal myofiber work increased. This suggests that RV free wall pacing in PH improves RV pump function and increases homogeneity of myofiber load in the ventricular walls.

This work contributes to the systemic interpretation of clinical data for the analysis of the regional cardiac function in the context of heart failure. A two-step patient-specific approach, combining a realistic geometry and a hybrid, tissue-level electromechanical model of the left ventricle is proposed. For the first step, a fast framework to extract a realistic geometry of the left ventricle from MSCT data is proposed. This geometry is then applied to a tissue-level model of the left ventricle, coupling a discrete electrical model, a mechanical model integrating a visco-elastic law, solved by a finite element method and a hydraulic model. A set of simulations carried out with the model are shown and preliminary results of the parameter identification approach, based on real patient data, are presented and discussed.

Recent experimental findings have suggested the important role played by blood vessels within the heart in stabilising arrhythmias. However, this link has yet to be explored computationally. In this paper, we develop a computational framework to model fibre orientation around structural inhomogeneities in myocardial tissue based on information obtained from high-resolution histological and MRI images. This framework allows the simulation of cardiac wavefront propagation for a generalised vessel orientation and position within the ventricular wall and transmural fibre architecture around it. We simulate propagation following different stimulation protocols around a transmural and a sub-epicardial vessel, using both bidomain and monodomain representations. We demonstrate the importance of accurately modelling the fibre structure around blood vessels relative to a simplistic transmurally varying fibre orientation model and suggest how this may impact pro-arrhythmic electrical dynamics.

This paper presents a simulation study on the identifiability of multiple reentrant circuits on the basis of the vectorcardiogram. The methods involved include an advanced tracking of the basic frequencies of the dominant rotors and a supporting identification based on the observed loops of their vectorcardiogram. The vector cardiogram was derived from body surface potentials spatially sampled by different lead systems. The results indicate that up to three independent circuits can be identified reliably.

Vulnerability to atrial fibrillation (AF) is increased in acutely dilated atria and is related to stretch-activated channels (SACs). To investigate the role of atrial anatomy in AF, we apply a computer model of human atrial electromechanics that includes SACs and contraction of the sarcomeres. Trabecular bundle structures are modeled by varying atrial wall thickness in a triangular mesh representing the human atria. Vulnerability to AF is investigated by application of overall stretch, while stimulating near the pulmonary veins. Due to contraction of some areas, stretch increases in other areas, leading to a variation in effective refractory period (ERP). Onset and perpetuation of AF in our model is explained by an increased dispersion in ERP, conduction slowing, and local conduction block. Atrial contraction attributes to the termination of AF through mechanoelectric feedback. We conclude that onset and termination of AF episodes under stretch are related to atrial structure and mechanoelectric feedback.

In this paper we propose a novel segmentation technique for quantification of sparsely sampled single-beat 3D contrast enhanced echocardiographic data acquired with a Fast Rotating Ultrasound transducer (FRU). The method uses a 3D Active Shape Model of the Left Ventricle (LV) in combination with local appearance models as prior knowledge to steer the segmentation. From a set of semi-manually delineated contours, 3D meshes of the LV endocardium are constructed for different cardiac phases. Mesh surfaces are partitioned into a fixed number of regions, each of which is modeled by a local image appearance. During segmentation, model update points are generated based on similarity matches with these local appearance models in multiple curved 2D cross-sections, which are then propagated over a dense 3D mesh. The Active Shape Model effectively constrains the shape of the 3D mesh to a statistically plausible cardiac shape. Leave-one-out cross validation was carried out on single-beat contrast enhanced FRU data from 18 patients suffering from various cardiac pathologies. Experiments show that the proposed method generates segmentation results that agree with the ground truth contours with average Point to Point (P2P) error of 4.1±2.0 mm and average Point to Surface (P2S) error of 2.4±2.1mm. Convergence tests show that the proposed method is capable of producing acceptable segmentation results (with less than 1.5X error compared to favorable initialization) within the range of 18~22 mm of in-plane displacement and 12~14 degrees of long-axial orientation error.

Whole heart segmentation from cardiac MRI is useful in clinics but challenging due to the large shape variation of the heart and many indistinct boundaries commonly presented in the MR images, especially in pathological cases. Image registration for whole heart MR images has been developed and applied to atlas propagation based segmentation . In this paper, we base on this segmentation framework and propose a new non rigid registration, a free-form deformations (FFDs) registration using adaptive control point status, for the segmentation refinement. This method activates and optimises control points according to the combined information of the deformation field and the heart surface from the atlas to improve the registration performance. The segmentation framework using this registration demonstrates a RMS accuracy of 1.8±0.2 mm in 10 pathological data.

It has been demonstrated that 3D anatomical models can be used effectively as roadmaps in image guided interventions. However, besides the anatomical information also the integrated display of functional information is desirable. In particular, a number of procedures such as the treatment of coro nary artery disease by revascularization and myocardial repair by targeted cell delivery require information about myocardial viability. In this paper we show how we can determine myocardial viability and integrate the information into a patient-specific cardiac 3D model. In contrast to other work we associate the viability information directly with the 3D patient anatomy. Thus we ensure that the functional information can be visualized in a way suitable for interventional guidance. Furthermore we propose a workflow that allows the nearly automatic generation of the patient-specific model. Our work is based on a previously published cardiac model that can be automatically adapted to images from different modalities like CT and MR. To enable integration of myocardial viability we first define a new myocardium surface model that encloses the left ventricular cavity in a way that suits robust viability measurements. We modify the model-based segmentation method to allow accurate adaptation of this new model. Second, we extend the model and the segmentation method to incorporate volumetric tissue properties. We validate the accuracy of the segmentation of the left ventricular cavity systematically using clinical data and illustrate the complete method for integrating myocardial viability by an example.

Live 3D trans-esophageal echocardiography (TEE) and X-ray fluoroscopy provide complementary imaging information for guiding minimally invasive cardiac interventions. X-ray fluoroscopy is most commonly used for these procedures due to its excellent device visualization. However, its challenges include the 2D projection nature of the images and poor soft tissue contrast, both of which are addressed by the use of live 3D TEE imaging. We propose to integrate 3D TEE imaging with X-ray fluoroscopy, providing the capability to co-visualize both the interventional devices and cardiac anatomy, by accurately registering the images using an electro-magnetic tracking system. Phantom trials validating the proposed registration scheme indicate an average accuracy of 2.04 mm with a standard deviation of 0.59 mm. In the future, this system may benefit the guidance and navigation of interventional cardiac procedures such as mitral valve repair or patent foramen ovale closure.

In this work, we present a registration framework for cardiac cine MRI (cMRI), tagged (tMRI) and delay-enhancement MRI (deMRI), where the two main issues to find an accurate alignment between these images have been taking into account: the presence of tags in tMRI and respiration artifacts in all sequences. A steerable pyramid image decomposition has been used for detagging purposes since it is suitable to extract high-order oriented structures by directional adaptive filtering. Shift correction of cMRI is achieved by firstly maximizing the similarity between the Long Axis and Short Axis cMRI. Subsequently, these shift-corrected images are used as target images in a rigid registration procedure with their corresponding tMRI/deMRI in order to correct their shift. The proposed registration framework has been evaluated by 840 registration tests, considerably improving the alignment of the MR images (mean RMS error of 2.04mm vs. 5.44mm).

This study investigates a fully automatic left ventricle segmentation method from cine short axis MR images. Advantages of this method include that it: 1) does not require manually drawn initial contours, trained statistical shape or gray-level appearance model; 2) provides not only endocardial and epicardial contours, but also papillary muscles and trabeculations’ contours; 3) introduces a roundness measure that is fast and automatically locates the left ventricle; 4) simplifies the epicardial contour segmentation by mapping the pixels from Cartesian to approximately polar coordinates; and 5) applies a fast Fourier transform to smooth the endocardial and epicardial contours. Quantitative evaluation was performed on 41 subjects. The average perpendicular distance between manually drawn and automatically selected contours over all slices, all studies, and two phases (end-diastole and end-systole) was 1.40±1.18 mm for endocardial and 1.75±1.15 mm for epicardial contours. These results indicate a promising method for automatic segmentation of left ventricle for clinical use.

We have developed a model of active contraction in the whole heart, using an anatomical geometry of a failing human heart, fitted to MRI data. Deformation in this model was driven by a model of active tension generation in human ventricular myocytes. By perturbing model parameters and boundary conditions we have predicted which metrics of cardiac function are sensitive to different parameters or boundary conditions. This allows us to begin to identify parameters that can be determined from different diagnostic modalities.

We present numerical results obtained with a three-dimensional electromechanical model of the heart with a complete realistic anatomy. The electrical activity of the heart-torso domain is described by the bidomain equations in the heart and a Laplace equation in the torso. The mechanical model is based on a chemically-controlled contraction law of the myofibres integrated in a 3D continuum mechanics description accounting for large displacements and strains, and the main cardiovascular blood compartments are represented by simplified lumped models. We considered a normal case and a pathological condition and the medical indicators resulting from the simulations show physiological values, both for mechanical and electrical quantities of interest, in particular pressures, volumes and ECGs.

Biophysically detailed cardiac cell models are based upon stiff ordinary differential equations describing ionic channels and intracellular dynamics aiming at reproducing experimental action potentials (APs) and intracellular calcium ([

$Ca^{2+}]_{i}$

) transients. Channel blocking and bifurcation analyses are local sensitivity analyses in model parameter space. However, all parameters influence model behaviour and require a global sensitivity index quantifying the influence of parameters on model responses. Identification of the influence of individual parameters increases our understanding of models. A global parameter sensitivity index for assessing the sensitivity of model responses to parameters in cardiac cell models is proposed. The robust index was applied to four widely used models. The analysis revealed that whilst models have common sets of parameters influencing AP and

$[Ca^{2+}]_{i}$

transients, there are subtle differences. This sensitivity analysis offers a systematic method for quantifying the influence of individual parameters on model behaviour to assist in model reduction, refinement or development.

We present a method for cardiac motion recovery using the adjustment of an electromechanical model of the heart to cine MRI. This approach is based on a proactive model which consists in a constrained minimisation of an energy coupling the model and the data. The presented method relies on specific image features in order to constrain the motion of the endocardia and epicardium and impose boundary conditions at the base. Thus, image intensity and gradient information are used to constrain the motion of the myocardium surfaces while a 3D block matching technique leads to the motion estimation of base vertices. Finally, we show that the implicit time integration of those forces and personalised boundary conditions lead to a better cardiac motion recovery from cine-MR images.

We propose a new framework for simulating blood flow inside the heart, usable with geometric models of the heart from patient-specific data. The method is geared toward realistic simulation of blood flow, taking into account not only heart wall motion but also valve motion. The simulator uses finite differences to discretize a domain that includes a functional model of the left ventricle and the left atrium of the heart. Navier-Stokes equations are robustly solved throughout the simulation domain, such that one-way momentum transfer between the heart model and the blood is enforced. The valve motion was timed to correspond to valve motions obtained from MRI. The final simulation results are qualitatively consistent to those from MR phase contrast velocity mapping, including high-velocity flow to the aorta during systole and toroidal vortex formation past the mitral valve during diastole.

Current capabilities of computerized radiology equipment have made three-dimensional (3D) and 4D (3D + time) medical images available for cardiac diagnosis, as well as for planning and guidance of therapy. Real-time display and interaction with these images plays an important role. To address this need, we have developed a graphics processing unit (GPU) accelerated 4D medical image visualization and manipulation platform, in which different levels of image representation and operation are employed, and anatomical volume of interest (VOI) can be efficiently enhanced. In addition, a new “depth texture indexing” algorithm is described to allow virtual surgical tools to be included in this environment, and a preliminary user study is performed to test the clinical usefulness and user performance enhancement when using the new techniques implemented in this software platform. The aim of this work is to supply cardiologists and surgeons with a comprehensive visualization software framework for cardiac disease preoperative diagnosis, treatment planning, and surgery guidance.

Achieving patient specific treatments is of fundamental importance in the future provision of cost and patient effective health care services for cardiovascular diseases. Implementing this view requires that software and data are reused and shared over the investigation process and within each single process phase. Within the heart modeling and simulation phase, geometrical models can be reused and shared with interoperable markup languages and a central repository. A markup language has been recently proposed to describe such models whereas a repository is still needed. In this paper we introduce

euHeartDB

, a publicly accessible database that aims to satisfy this role by providing the functionalities to upload, to search and to integrate geometrical models. The database is web-enabled and allows the visualization of the models which functionality is fundamental to human users. In addition,

euHeartDB

uses the Foundational Model of Anatomy ontology to classify the models and to support semantic-based queries.

GIMIAS is a workflow-oriented environment for addressing advanced biomedical image computing and build personalized computational models, which is extensible through the development of application-specific plug-ins. In addition, GIMIAS provides an open source framework for efficient development of research and clinical software prototypes integrating contributions from the Virtual Physiological Human community while allowing business-friendly technology transfer and commercial product development. This framework has been fully developed in ANSI-C++ on top of well known open source libraries like VTK, ITK and wxWidgets among others. Based on GIMIAS, in this paper is presented a workflow for medical image analysis and simulation of the heart.

Automated motion tracking of the myocardium from 3D echocardiography provides insight into heart’s architecture and function. We present a method for 3D cardiac motion tracking using non-rigid image registration. Our contribution is two-fold. We introduce a new similarity measure derived from a maximum likelihood perspective taking into account physical properties of ultrasound image acquisition and formation. Second, we use envelope-detected 3D echo images in the raw spherical coordinates format, which preserves speckle statistics and represents a compromise between signal detail and data complexity. We derive mechanical measures such as strain and twist, and validate using sonomicrometry in open-chest piglets. The results demonstrate the accuracy and feasibility of our method for studying cardiac motion.

This paper proposes a new registration method for the

in vivo

quantification of cardiac deformation from a sequence of possibly noisy images. Our algorithm has been applied to 3D ultrasound (3D-US) images, which currently give a reasonable spatial and time resolution, but suffer from significant acquisition noise. Therefore, this modality requires the design of a robust strategy to quantify motion and deformation with the clinical aim of better quantifying cardiac function e.g. in heart failure. In the proposed method, referred to as Large Diffeomorphic Free Form Deformation (LDFFD), the displacement field at each time step is computed from a smooth non-stationary velocity field, thus imposing a coupling between the transformations at successive time steps. Our contribution is to extend this framework to the estimation of motion and deformation in an image sequence. Similarity is captured for the entire image sequence using an extension of the pairwise mutual information metric. The LDFFD algorithm is applied here to recover longitudinal strain curves from healthy and Left-Bundle Branch Block (LBBB) subjects. Strain curves for the healthy subjects are in accordance with the literature. For the LBBB patient, strain quantified before and after Cardiac Resynchronization Therapy show a clear improvement of cardiac function in this subject, in accordance with clinical observations.

Automatic delineation of the myocardium in real-time 3D echocardiography may be used to aid the diagnosis of heart problems such as ischaemia, by enabling quantification of wall thickening and wall motion abnormalities. Distinguishing between myocardial and non-myocardial tissue is, however, difficult due to low signal-to-noise ratio as well as the efficiency constraints imposed on any algorithmic solution by the large size of the data under consideration. In this paper, we take a machine learning approach treating this problem as a two-class 3D patch classification task. We demonstrate that solving such task using

random forests

, which are the discriminative classifiers developed recently in the machine learning community, allows to obtain accurate delineations in a matter of seconds (on a CPU) or even in real-time (on a GPU) for the entire 3D volume.

Cardiac magnetic resonance (MR) imaging has advanced to become a powerful diagnostic tool in clinical practice. Automatic detection of anatomic landmarks from MR images is important for structural and functional analysis of the heart. Learning-based object detection methods have demonstrated their capabilities to handle large variations of the object by exploring a local region, context, around the target. Conventional context is associated with each individual landmark to encode local shape and appearance evidence. We extend this concept to a landmark

set

, where multiple landmarks have connections at the semantic level, e.g., landmarks belonging to the same anatomy. We propose a joint context approach to construct contextual regions between landmarks. A discriminative model is learned to utilize inter-landmark features for landmark set detection as an entirety. This helps resolve ambiguities of individual landmark detection results. A probabilistic boosting tree is used to learn a discriminative model based on contextual features. We adopt a marginal space learning strategy to efficiently learn and search a high dimensional parameter space. A fully automatic system is developed to detect the set of three landmarks of the left ventricle, the apex and the two basal annulus points, from a single cardiac MR long axis image. We test the proposed approach on a database of 795 long axis images from 116 patients. A 4-fold cross validation results show that about 15% reduction of the errors is obtained by integrating joint context into a conventional landmark detection system.

Minimally invasive beating heart intracardiac surgery is an area of research with many unique challenges. Surgical targets are in constant motion in a blood-filled environment that prevents direct line-of-sight guidance. The restrictive workspace requires compact, yet robust tools for proper therapy delivery. Our novel method for approaching multiple targets inside the beating heart allows their identification and access under augmented reality-assisted image guidance. The surgical platform integrates real-time ultrasound imaging with virtual models of the surgical instruments, along with virtual cardiac anatomy acquired from pre-operative images. Extensive

in vitro

studies were performed to assess the operator’s ability to “deliver therapy” to dynamic intracardiac targets via both transmural and transluminal access, and demonstrated significantly more accurate targeting under augmented reality guidance compared to ultrasound image guidance alone, accompanied by a reduction of procedure time by half. Moreover, preliminary

in vivo

acute studies on porcine models showed successful prosthesis positioning for beating-heart septal defect repair and mitral valve implantation via direct surgical access. While still in its infancy, this work emphasizes the promise of ultrasound-enhanced model-guided environments for minimally-invasive cardiac therapy, whether delivered via a catheter introduced into the vascular system or a cannula inserted through the heart wall.

During open heart coronary artery bypass grafting (CABG), optimal placement of the bypass graft on the diseased target vessel is of utmost importance. To assist the heart surgeon in this matter, a surgical navigation system has been developed. The concept includes preprocedural planning of the optimal position for the distal anastomosis (surgical target) on the target vessel and intraoperative image-guided navigation to that position. Navigation is enabled by registration of pre- and intraoperative data. The registration mechanism was validated retrospectively. Promising results justified subsequent intraoperative live application of the system during 11 open heart CABG surgeries enabling navigation on 21 target vessels on the anterior, lateral and inferior wall of the heart. In each case, the employed registration mechanism ensured good alignment of preoperatively planned and intraoperatively identified surgical targets. The system is very well applicable for in-vivo navigation and allows for bypass grafting precisely at a preoperatively planned position.

Computational models of myocardial perfusion in normal as well as pathological conditions have contributed to recent advances in the personalized diagnosis and treatment of cardiovascular disease.. For the coronary circulation we attempt to develop a vascular tree model based on realistic replicas from the intramural vessels in hearts from animals and man. These virtual replicas are obtained by means of a novel imaging cryomicrotome producing a stack of 4000 images of 2000x2000 pixels. This paper describes the methods developed for segmentation of the images resulting in a 3D virtual vascular model. Also the methods for fluorescent microsphere detection are described used to validate the models in prediction of flow distribution over the walls of the heart chambers. Special attention is provided to the detection of collaterals and collateral flow.

Q-ball imaging (QBI) is an established high-angular resolution diffusion MRI technique enabling to resolve intravoxel fibre structure. Here, we present a detailed study of myocardial fibre orientation using QBI. We compare standard diffusion tensor MRI (DTI) versus QBI in the canine left ventricular free wall (LVFW) and posterior left-right ventricular insertion site. Most voxels within the LVFW show high fractional anisotropy (FA), Gaussian diffusion profiles, and a single population of aligned fibres. In these, the difference between fibre helix angles estimated by DTI and QBI is below 5°. However, we show that reduced FA near the anterior papillary muscle in the LVFW and in most of the left-right ventricular fusion site correlates with non-Gaussian diffusion. The QBI orientation distribution functions (ODF) in these regions reveal complex intravoxel fibrous structure, which cannot be inferred using DTI. Extensive ODF maps of myocardial fibre orientation are presented and discussed for the first time to our knowledge.

The purpose of this computational study was to test the pertinence of the magnitude of the minimum time derivative of the extracellular potential, |dV

es

/dt

min

|, measured in a thin, conducting solution layer adjacent to the tissue, as an index of cardiac excitability. For this purpose, we performed computational studies characterizing the relationship between |dV

es

/dt

min

| and the maximum upstroke velocity of transmembrane voltage, dV

m

/dt

max

, which has been used in previous studies as an index of excitability. A three-dimensional bidomain model of electrical conduction in cardiac tissue was used based on the Noble-Varghese-Kohl-Noble model of ventricular myocytes. The spatial domain included a slab of cardiac tissue with intra- and extracellular anisotropic conductivities surrounded by a layer of solution. The simulations showed linear relationships between |dV

es

/dt

min

| and dV

m

/dt

max

for reduction of maximum sodium current conductance (G

Na

) from 100% to 20%. The relationship was dependent on location and propagation direction. However, when both parameters were normalized, those dependencies disappeared. In summary, our study demonstrated that normalized |dV

es

/dt

min

| is linearly related to normalized dV

m

/dt

max

. The results support our hypothesis that normalized |dV

es

/dt

min

| can be used as an index of cardiac excitability.

The major determinants of the T wave polarity in electrocardiograms (ECGs) are still a debated issue. The aim of this work is to investigate the effects of tissue anisotropy, cellular action potential duration (APD) heterogeneities and excitation wavefront shape on the T wave polarity in unipolar and bipolar ECGs, simulated in a conducting medium surrounding the cardiac tissue at some distance fom endo to epicardium. The study is based on three-dimensional anisotropic Monodomain simulations of the entire depolarization and repolarization phases of propagating action potentials in a parallelepipedal slab. The results show that the T wave of unipolar ECGs is positive at all sites explored and its shape and polarity are mainly determined by the anisotropy of the cardiac tissue, irrespective of cellular APD heterogeneities and shape of the excitation wavefront. On the other hand, bipolar ECGs are mainly affected by their isotropic component and their T wave turns out to be positive for single site stimulations in the presence of transmural APD heterogeneity, while it becomes always negative in case of multiple sites stimulation generating large activation wavefronts, regardless of the considered cellular APD heterogeneities.

We consider the problem of building a standard electrocardiogram (ECG) from the electrical potential provided by a pacemaker in a few points of the heart (electrogram). We use a 3D mathematical model of the heart and the torso electrical activity, able to generate “computational ECG”, and a “metamodel” based on a kernel ridge regression. The input of the metamodel is the electrogram, its output is the ECG. The 3D model is used to train and test the metamodel. We illustrate the performance of the proposed strategy on simulated bundle branch blocks of various severities and a few clinical data.