Fig 12 - uploaded by Maxime Sermesant
Content may be subject to copyright.
Anatomical areas assigned to the myocardium mesh from the myocardium segmentation of the Visible Human Project. (This figure is available in colour, see the on-line version.)
Source publication
This article describes a methodology for creating a generic volumetric biomechanical model from different image modalities and segmenting time series of medical images using this model. The construction of such a generic model consists of three stages: geometric meshing, non-rigid deformation of the mesh in images of various modalities, and image-t...
Citations
... D EFORMABLE image registration (DIR) establishes dense spatial correspondences between different medical image acquisitions [1]. It plays a critical role in various tasks of medical image analysis including anatomical change diagnosis [2], longitudinal studies [3] and statistical atlas building [4]. Given a source image s and a target image t on a spatial domain Ω ∈ R d , the goal of DIR is to find an optimal non-linear dense transformation or field ϕ : Ω × R → Ω that minimizes the energy: ...
Deformable image registration plays a critical role in various tasks of medical image analysis. A successful registration algorithm, either derived from conventional energy optimization or deep networks requires tremendous efforts from computer experts to well design registration energy or to carefully tune network architectures for the specific type of medical data. To tackle the aforementioned problems, this paper proposes an automated learning registration algorithm (AutoReg) that cooperatively optimizes both architectures and their corresponding training objectives, enabling non-computer experts, e.g., medical/clinical users, to conveniently find off-the-shelf registration algorithms for diverse scenarios. Specifically, we establish a triple-level framework to deduce registration network architectures and objectives with an auto-searching mechanism and cooperating optimization. We conduct image registration experiments on multi-site volume datasets and various registration tasks. Extensive results demonstrate that our AutoReg may automatically learn an optimal deep registration network for given volumes and achieve state-of-the-art performance, also significantly improving computation efficiency than the mainstream UNet architectures (from 0.558 to 0.270 seconds for a 3D image pair on the same configuration).
... First, tagged images with 0.5 mm pixel size are generated from the time-resolved LV meshes through rasterization [Sermesant et al., 2003], and using a simplified complementary spatial modulation model [Ryf et al., 2002] for 3D tagging without tag-line fading ( Figure 5.1b-1). We assume a standard image generating function used in CSPAMM to create the tagged image stacks: ...
Cardiovascular diseases (CVDs) remain the leading cause of death and disability worldwide. Moreover, the life quality of patients is severely decreased by the long-term care required, causing a significant increase in healthcare costs. Although several advancements have been made for CVD management in terms of diagnostic, prognostic and therapeutic techniques, the development of novel and more effective approaches is still required to decrease the prevalence of cardiac-related diseases. Early and accurate diagnosis of cardiac dysfunction plays a significant role to reduce or stop disease progression. In this respect, efforts have been mainly focused on non-invasive characterization of functional measures of myocardium that can increase the diagnostic accuracy of traditional techniques.In clinical cardiology, cardiac performance is mostly assessed based on global functional parameters, e.g., ventricular volume, ventricular mass and ejection fraction (EF). Although EF has been the key parameter for disease diagnosis and management, it has been shown to be in normal ranges despite specific dysfunctional cases, e.g. in heart failure with preserved ejection fraction. Strain, strain rate and torsion, however, have allowed for a deeper insight into myocardial function, providing additional diagnostic and prognostic characterization on the regional level in addition to the conventional functional measures. Global longitudinal and circumferential strains are commonly used in clinical setting for assessing myocardial contractility. Although several research studies have reported left-ventricular radial strain and torsion to be early indicators of impaired myocardial contractility, they are rarely utilized in the clinical setting.Computational models have become a powerful tool to characterize cardiac physiology in health and disease and to extract functional parameters, e.g., stress and contractility, that cannot be obtained directly by any in vivo techniques. Recent cardiac models account for many aspects of cardiac behavior, spanning from electrophysiology, fluid mechanics and material behavior. Several computational models have been proposed to study dysfunctional cases, including myocardial infarction and ischemia, valvular disease and cardiomyopathies. Moreover, incorporation of patient-specific cardiac geometries and functional parameters has brought a novel perspective to cardiac research. Given the recent improvements in medical imaging technology, the generation of more detailed individualized cardiac models is a critical step towards bringing translational computational models into clinic.Several imaging modalities are utilized for diagnostic purposes in the clinics. Among them, cardiovascular magnetic resonance (CMR) is accepted as the gold standard to assess myocardial function. In the context of this thesis, the focus lies on two magnetic resonance (MR) sequences: cine and tagged CMR, having their own advantageous features. Cine CMR provides excellent tissue contrast to visualize cardiac anatomy for the characterization of global functional measures, e.g., left-ventricular mass, volumes and EF. Tagged CMR allows for the assessment of local deformation, e.g. strain, strain rate and torsion. Although global circumferential and longitudinal strains have a good reproducibility across image post-processing techniques both on cine and tagged CMR, radial strain is mostly underestimated on tagged CMR and varies significantly among different techniques regardless of the image type utilized.To investigate the accuracy and precision of strain quantification, we performed a detailed analysis using synthetic 3D tagged CMR, generated from a biomechanical model of the left ventricle. Several image characteristics were varied including image resolution, tag line distance and the signal-to-noise ratio (SNR). A finite element-based image registration technique was employed to track tissue motion over time. The resulting displacement and strain fields are compared to ground truth to assess the contribution of each image characteristic to tracking errors. Radial strain is shown to be sensitive to changes in image resolution and SNR while circumferential and longitudinal components are relatively robust with respect to changes in image characteristics. This study stands as a systematic investigation of image requirements for myocardial deformation quantification.To address the shortcoming of individual cine and tagged CMR data analysis, we proposed a combined image analysis technique of tagged and cine CMR to improve radial strain and twist quantification. For this purpose, tracking is first performed on cine images and the resulting displacement field is utilized to mask the tagged images; tracking is then performed on the masked tagged images. The performance of the combined technique is shown both on human and porcine datasets in terms of strain and twist quantification. The analysis results reveal the superiority of combined image registration over tagged-only analysis in terms of radial strain quantification while more physiological twist is achieved compared to cine-only registration.
... Non-invasive imaging, such as ultrasound, CT and MRI, can be used to determine patient-specific cardiac geometry, microstructural orientation [463,323] and deformation [430,126]. Non-invasive imaging modalities which provide temporal geometric information can also be used to obtain deformation in three dimensions by utilizing geometric meshing and image-to-mesh information mapping through rasterization [106,442,134,478,454,437], which can be further improved in accuracy through the addition of a constitutive model [377] and machine learning [517]. Patient-specific parameters of the constitutive model can be estimated by minimizing the difference between modeled versus measured deformation [174,497,19,20,499]. ...
The hierarchical construction of the myocardium plays a pivotal role in the biomechanics of the heart muscle and the resulting flow of blood. In disease, the construction of the heart remodels, altering the structure of the tissue from the subcellular level all the way to the whole organ. Elucidating the impact of these fundamental alterations on the biomechanics of the heart presents challenges to diagnosis, therapy planning and treatment. Computational modeling provides an innovative tool, enabling the simulation of complex biomechanics that capture the complexity of tissue, its growth and remodeling and the resulting blood flow. In this chapter, we review the key ways that computational models can address challenging biomechanical questions in the heart and how these tools can change the way treatment is approached across a range of heart diseases.
... Digital heart atlases (DHA) are the useful tools for computational simulation of cardiovascular function and diseases. [1] The applications of DHA-based simulations included the studies of the biomechanical and electrophysiology cardiovascular function, [2][3][4][5][6] clinical diagnostics, and treatment of heart diseases. [7] In the field of medical imaging, DHAs are used for simulating anatomical modelling and motion modelling. ...
... [38,39] The establishment of 4D cardiac models enabled quantification of bi-ventricular motion malfunction (e.g., Takotsubo cardiomyopathy, hypokinesis, and apical ballooning [40,41] ) from cardiac magnetic resonance imaging (MRI), computed tomographic (CT), ultrasound, and nuclear medicine images. [3,[9][10][11]34,[42][43][44][45][46][47][48] Recently, the interactions between ventricular shape and deformation were also modelled with manifold learning approach, endowing the inherent structural and functional correlations into the heart atlas. [49] Summarizing the development history, DHA have evolved from the preliminary simple geometry ventricular models to whole-heart scale anatomically realistic models. ...
Background and Objectives
Digital heart atlases play important roles in computational cardiac simulation and medical image analysis. During the past decades, various heart anatomy models were developed, but they mostly focused on the ventricular part. Recently, a number of whole-heart atlases were developed but they rarely modelled the motion features. This study constructed a whole-heart atlas incorporating dynamic cardiac motion.
Materials and Methods
The shape and motion features of the atlas were learnt from a training set of 57 dynamic computed tomographic angiography images including 20 cardiac phases. Inter-subject variations of the heart anatomy and motion were incorporated into the atlas using the statistical shape modelling approach. Clinically relevant physiological parameters (e.g., chamber volumes, ejection fraction, and percentage of systolic phase) were correlated with the shape and motion variations using the linear regression approach. The shape and motion pattern of the atlas can be adapted by adjusting the physiological parameters.
Results
Quantitative experiments were conducted to measure the anatomical accuracy of the atlas for whole-heart shape reconstruction of different subjects, a mean Dice score of 0.89-0.93 and a mean surface distance of 1.02-1.91 mm were achieved for the four heart chambers, respectively.
Conclusions
This atlas provides a novel computational tool with adjustable shape and motion parameters for cardiac simulation research.
... The modified LV model is further utilized to generate synthetic 3D-tagged images as follows. First, tagged images with 0.5 mm pixel size are generated from the time-resolved LV meshes through rasterization [39], and using a simplified complementary spatial modulation model [7] for 3D tagging without tag-line fading (Fig 1b-1). We assume a standard image generating function used in CSPAMM to create the tagged image stacks: ...
Cardiac Magnetic Resonance Imaging (MRI) allows quantifying myocardial tissue deformation and strain based on the tagging principle. In this work, we investigate accuracy and precision of strain quantification from synthetic 3D tagged MRI using equilibrated warping. To this end, synthetic biomechanical left-ventricular tagged MRI data with varying tag distance, spatial resolution and signal-to-noise ratio (SNR) were generated and processed to quantify errors in radial, circumferential and longitudinal strains relative to ground truth. Results reveal that radial strain is more sensitive to image resolution and noise than the other strain components. The study also shows robustness of quantifying circumferential and longitudinal strain in the presence of geometrical inconsistencies of 3D tagged data. In conclusion, our study points to the need for higher-resolution 3D tagged MRI than currently available in practice in order to achieve sufficient accuracy of radial strain quantification.
... [10][11][12] A common process consists in using a biomechanical finite element (FE) model that simulates in a physically realistic way the deformations produced in both the surface and internal tissues of the breast during the mammographic acquisition. Finite element models have been widely used in various medical applications, including brain, 13 heart, 14 liver, 15 lungs, 16 or prostate 17 imaging, composing a wide bibliography. [18][19][20] In breast modeling, they have been used in several challenging problems, such as the colocalization of information between different image modalities, 21 temporal studies, 22 identifying lesions or tumors, 23 tracking of these lesions during biopsy, 24 review of the progress of suspicious lesions or evaluation of the effectiveness of treatments and, even, aiding implant selection for breast augmentation procedures. ...
Breast magnetic resonance imaging (MRI) and X-ray mammography are two image modalities widely used for the early detection and diagnosis of breast diseases in women. The combination of these modalities leads to a more accurate diagnosis and treatment of breast diseases. The aim of this paper is to review the registration between breast MRI and X-ray mammographic images using patient-specific finite element-based biomechanical models. Specifically, a biomechanical model is obtained from the patient's MRI volume and is subsequently used to mimic the mammographic acquisition. Due to the different patient positioning and movement restrictions applied in each image modality, the finite element analysis provides a realistic physics-based approach to perform the breast deformation. In contrast with other reviews, we do not only expose the overall process of compression and registration but we also include main ideas, describe challenges and provide an overview of the used software in each step of the process. Extracting an accurate description from the MR images and preserving the stability during the finite element analysis require an accurate knowledge about the algorithms used, as well as the software and underlying physics. The wide perspective offered makes the paper suitable not only for expert researchers but also for graduate students and clinicians. We also include several medical applications in the paper, with the aim to fill the gap between the engineering and clinical performance. This article is protected by copyright. All rights reserved.
... Rasterization is an old problem, well known in computer graphics, that find the inner voxels of each tetrahedron. Sermesant at al., [19] used a multi-stage approach for this problem by first considering the tetrahedron intersection with each the relevant horizontal planes. Then, finding the bounding edges and inner voxels located in horizontal lines of each such plane. ...
... Then, finding the bounding edges and inner voxels located in horizontal lines of each such plane. We propose here a more convenient and less complicated solution to the rasterization problem compared to [19] Given a tetrahedron T ,any point p ∈ T divides it into four sub tetrahedrons. ...
Image registration is a central field in biomedical image analysis. Despite the large body of work on automatic image registration algorithms, in practice, results are often imperfect and can benefit from manual editing. In the current study, we propose a novel interactive registration tool that can be used for editing and refinement of automatic image registration results. The method is based on displacements to control points applied by the user to a finite element mesh representation of an organ of interest. The user-applied-displacements are translated to local forces using the organ’s elastic properties. The local user-derived forces are applied on the mesh and deform the image accordingly. We test the current method on 2D x-ray hand simulated data with non-continuous motion field and on 3D cardiac CT image data.
... The difficulties of this process arise from the presence of noise, lack of edge information for the epicardium, and intensity inhomogeneities. Accordingly, many approaches have been developed for myocardial contour delineation to overcome these issues, including the use of dynamic programming ( [5]), probability atlas ( [6]), active shape model (ASM) ( [7], [8]), active appearance model (AAM) ( [9], [10]), combination of ASM and AAM ( [11]), constrained pattern matching strategy ( [12]), proactive deformable model ( [13]), and motion-guided segmentation framework ( [14], [15]). In addition to the continuing progress in cardiac image segmentation, considerable efforts have also been focused on motion and deformation recovery from image sequences. ...
Although accurate and robust estimations of the deforming cardiac geometry and kinematics from cine tomographic medical image sequences remain a technical challenge, they have significant clinical value. Traditionally, boundary or volumetric segmentation and motion estimation problems are considered as two sequential steps, even though the order of these processes can be different. In this paper, we present an integrated, spatiotemporal strategy for the simultaneous joint recovery of these two ill-posed problems. We use a mesh-free Galerkin formulation as the representation and computation platform, and adopt iterative procedures to solve the governing equations. Specifically, for each nodal point, the external driving forces are individually constructed through the integration of data-driven edginess measures, prior spatial distributions of myocardial tissues, temporal coherence of image-derived salient features, imaging/image-derived Eulerian velocity information, and cyclic motion model of myocardial behavior. The proposed strategy is accurate and very promising application results are shown from synthetic data, magnetic resonance (MR) phase contrast, tagging image sequences, and gradient echo cine MR image sequences.
... Registration of the original cardiac model to the provided imaging modality is a mandatory step so as to integrate various pieces of information in a single model [235]. Our work performed rigid and non-rigid or local registrations of the available 3D mesh models on the given ultrasound data. ...
... A force is generated from the model-data similarity in order to deform the model object and shape it in the form of the contours of the observed object. as the distance-to-contour-based similarity [235,[239][240][241] and the gradient vector flow [242][243][244]. More recently, a formalism named currents has been introduced [245][246][247] whose objective is to designate a surface or a contour in a vector space by studying its action on a space of vector fields. ...
Cardiac hypertrophy is a risk factor of cardiovascular mortality and is rou- tinely examined using ultrasound (US) imaging. This imaging modality does not allow the differentiation of the types of hypertrophy, such as the infiltrative caused by amyloidosis and the sarcomeric type resulting from hypertrophic cardiomyopathy. This leads to errors in this disease’s prognosis and treatment.
The myocardial tissue undergoes modifications specific to every disease ex- pressed in ultrasound by changes in texture difficult to be perceived by the naked eye. Our work is focused on the detection and quantification of speckle texture from the ultrasound image database of different types of cardiac hyper- trophy.
The images undergo texture characterization using Gabor filters of different orientations, sizes, and decomposition levels. The first and second-order statis- tical features are then determined from the resulting images. Next, Principal Component Analysis (PCA) is used for feature reduction followed by Linear Discriminant Analysis (LDA) for supervised clustering. This work gives good classification results differentiating the two main classes of cardiac hypertro- phy in addition to the three sub-classes of cardiac amyloidosis. This serves as a good basis for further studies of amyloidosis and its diagnosis in clinical practice.
We also performed rigid and local registrations of dynamic cardiac models with ultrasound patient-specific data. Such modeling complements our texture characterization approach in order to provide non-invasive clinical healthcare for the different stages of this pathology’s detection and treatment.
... To create a specific-patient heart, it requires magnetic resonance images synchronized with ECG and breathing in order to reduce the noise and motion artefacts caused by to the cardiac cycle and breathing movements. Heart tissue can also be personalized using MRI [5], such as the location and extension of the Myocardium. The CCS and the fiber orientation cannot be personized yet because of limited information. ...
Understanding the human heart anatomy is important in order to extract information and to understand how it works. Magnetic resonance imaging (MRI) can be a robust solution to extract some information about the heart anatomy. Data from multiple image planes can be combined to create a 3D model of the cardiac system. The shape of the human heart is an irregular shape that makes it difficult to model and it takes many hours to create a 3D shape of the human heart with a high details and precision. A new three-dimensional heart model has been developed including structural information from MRI. The method uses Marching Cubes (MC) Algorithm for the extraction of an equipotential surface (subsurface) of a 3D mesh structured and uniform model. In order to visualize the 3D virtual model created in a real environment, we use Augmented Reality techniques (AR).
