Towards patient-specific modeling of mitral valve repair: 3D transesophageal echocardiography-derived parameter estimation
Graphical abstract
Introduction
The mitral valve apparatus is a complex anatomical unit composed of the mitral annulus, anterior and posterior valve leaflets, and chordae tendineae that tether the leaflets to papillary muscles (as shown in Fig. 1). While problems with either opening or closing can cause symptoms, mitral regurgitation is much more prevalent than mitral stenosis, especially in developed countries (Mann et al., 2014). Mitral regurgitation can be classified as either primary (with anatomic valvular pathology) or secondary (with functional abnormality despite normal valve leaflets). The preferred intervention for severe primary mitral regurgitation is valvular repair, when possible, rather than replacement. Surgical mitral valve repair, however, is a technically challenging operation that requires considerable expertise to achieve optimal outcomes (Nishimura et al., 2016). The procedure typically involves leaflet resection and annuloplasty ring placement. Occasionally, there may be placement of edge-to-edge leaflet sutures or artificial chords (Glower, 2012).
Noninvasive imaging of the mitral valve by transesophageal echocardiography (TEE) is critical to treatment planning. TEE, especially with three-dimensional imaging, allows definitive classification of the mechanism of mitral regurgitation and allows surgeons to estimate the likelihood of a successful repair. While patient-specific anatomic and functional information can be visualized with this approach, the prediction of outcomes is still primarily based upon the surgeon’s experience and intuition.
There has been growing interest in developing tools to allow patient-specific modeling, both static and dynamic, of mitral valve anatomy and function (Sun, Martin, Pham, 2014, Kunzelman, Einstein, Cochran, 2007). These models can generally be categorized as either geometrical or biomechanical models. Geometrical models use semi-automated or automated analysis of medical images to reconstruct the mitral valve apparatus, track the motion of its components, and provide quantitative measurements. While these parameters can be used to evaluate mitral valve geometry and dynamics, they have limited ability to predict behavior after pathologic changes or interventions. Biomechanical models attempt to simulate mitral valve dynamics by incorporating initial geometry, assumed constitutive properties, and boundary conditions. Direct predictions can then be made regarding postsurgical function.
While increasingly sophisticated biomechanical models have been developed, truly patient-specific mitral valve modeling has not been fully realized yet. An important limitation involves the inability to accurately estimate in vivo tissue properties. Typical investigations of mitral valve material properties have been performed on in vitro specimens from animals or humans whose characteristics may or may not be similar to those of specific patients. Such studies have, in fact, shown that aging, hypertension, and myxomatous degeneration are all associated with altered mechanical properties (Pham, Sun, 2014, Stephens, de Jonge, McNeill, Durst, Grande-Allen, 2009, Barber, Kaspera, Ratliff, Cosgrove, Griffin, Vesely, 2001). It seems likely that the degree of myxomatous degeneration (the most common cause of primary mitral regurgitation) will also have an effect. A different approach is to obtain invasive in vivo measurements in animals (e.g., from sonomicrometers sutured to the leaflets (Rausch et al., 2013)), but these values will also have limited applicability to specific patients. Accordingly, development of methods to estimate in vivo patient-specific tissue properties through noninvasive imaging would be extremely helpful towards allowing mitral valve modeling to become useful in clinical practice (Sun et al., 2014).
The study presented here uses an approach combining image analysis and biomechanics in order to perform patient-specific mitral valve modeling. Three-dimensional TEE images are used in a semi-automated personalization framework that estimates chordae rest lengths and leaflet material parameters. In other words, the biomechanical model is iteratively calibrated to image-based observations to find an optimized set of geometric and biomechanical parameters describing the mitral valve dynamics for the patient whose echocardiographic images are being analyzed.
The goal of the study is to develop streamlined framework to build patient-specific biomechanical models based on information from medical images. The study presented in this paper extends our previous work (Kanik, Mansi, Voigt, Sharma, Ionasec, Comaniciu, J., 2013a, Kanik, Mansi, Voigt, Sharma, Ionasec, Comaniciu, J., 2013b) from three perspectives: First, a comprehensive framework is described to personalize biomechanical models based on medical images. This framework is not limited to a specific biomechanical model or image modality. Second, the chordae optimization algorithm is improved with better computational efficiency. Third, the framework is tested on a larger data set of 15 clinical examples. The rest of the paper is organized as follows. The framework for patient-specific model is presented in Section 2 including the image analysis module, biomechanical model module and optimization module. The framework is evaluated on 15 examples to simulate mitral valve closure and the simulation results are compared to the ground truth in Section 3. Finally, the paper is concluded with discussions in Sections 4 and 5.
Section snippets
Methods
An overview of the framework is shown in Fig. 2. It starts from 3D real time TEE images and applies an image processing algorithms to estimate the geometry and dynamics of the mitral valve (Section 2.1). The image analysis module outputs finite element meshes of the mitral valve at open state (g0) and closed state (gN). The open valve mesh g0 is fed into the biomechanical model to generate a simulated closed valve geometry (Section 2.2). is then compared to the segmented gN to adjust
Data set and experiments
The performance of the personalization of the mitral valve biomechanical model is evaluated on 15 3D Real-Time TEE images from 14 patients. The images were randomly collected on patients diagnosed of various heart diseases including mitral valve diseases, aortic valve diseases, and myocardium infarction. The 3-D TEE acquisition used an iE33 Philips console with an X7-2t TEE probe with image resolution 0.75–1.58 mm. The average total number of frames is 16 and the mitral valve closure can be
Discussion and future direction
The personalization algorithm estimates the chordae rest length and material parameters by incorporating the image analysis module and the biomechanical module. Here we list several potential advancements in the future that could improve the performance of this parameter estimation.
Image analysis: The image analysis module reconstructs the morphology and dynamics of the mitral valve from in-vivo measurements. The open valve configuration is used in the biomechanical model as the initial
Conclusions
The paper presented an streamlined framework to build a patient-specific biomechanical model of the mitral valve from TEE images with minimal user interaction. The framework combines image analysis and biomechanics to optimize chordae rest length and material parameters to ensure the biomechanical model simulated mitral valve closure matches the image based observations. Through the personalization, the mitral valve model is calibrated for individual patient. During the personalization,
Acknowledgments
The authors thank Siemens Corporation, Corporate Technology, for its funding to support the work, and Dr. Ben A. Lin for his insights during the discussions.
References (46)
- et al.
Mechanical properties of myxomatous mitral valves
J. ThoracicCardiovasc. Surg.
(2001) Surgical approaches to mitral regurgitation
J. Am. College Cardiol.
(2012)- et al.
Computational model of aortic valve surgical repair using grafted pericardium
J. Biomech.
(2012) - et al.
Stress-strain behavior of mitral valve leaflets in the beating ovine heart
J. Biomech.
(2009) - et al.
Mitral valve diseasecurrent management and future challenges
The Lancet
(2016) - et al.
Mitral annulus segmentation from four-dimensional ultrasound using a valve state predictor and constrained optical flow
Med. Image Anal.
(2012) - et al.
Mitral leaflet modeling: importance of in vivo shape and material properties
J. Biomech.
(2011) - et al.
Effect of ventricle motion on the dynamic behaviour of chorded mitral valves
J. Fluids Struct.
(2008) - et al.
Patient-specific modeling and analysis of the mitral valve using 3d-tee
Information Processing in Computer-Assisted Interventions
(2010) - et al.
Efficient nonlinear fem for soft tissue modelling and its gpu implementation within the open source framework sofa
Biomedical Simulation
(2008)
Robust extended kalman filtering
IEEE Trans. Signal Process.
Fluid-structure interactions of the mitral valve and left heart: comprehensive strategies, past, present and future
Int. J. Numer. Methods Biomed. Eng.
Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography
Comm. ACM
Fast image-based model of mitral valve closure for surgical planning
Computational Biomechanics for Medicine (MICCAI Workshop)
The stiffness of normal and abnormal mitral valves
Ann. Thoracic Cardiovasc. Surg.
Patient-specific modeling and quantification of the heart valves from multimodal cardiac images
Automatic personalization of the mitral valve biomechanical model based on 4d transesophageal echocardiography
Statistical atlases and computational models of the heart (MICCAI workshop)
Estimation of patient-specific material properties of the mitral valve using 4d transesophageal echocardiography
ISBI
Material properties of the ovine mitral valve anterior leaflet in vivo from inverse finite element analysis
Am. J. Physiol. Heart Circulatory Physiol.
Mechanical properties of basal and marginal mitral valve chordae tendineae
ASAIO Trans.
Fluid-structure interaction models of the mitral valve: function in normal and pathological states
Philos. Trans. Royal Soc. B
State-space analysis of cardiac motion with biomechanical constrains
IEEE Trans. Image Process.
Maximum a posteriori strategy for the simultaneous motion and material property estimation of the heart
IEEE Trans. Biomed. Eng.
Cited by (21)
A geometry-based finite element tool for evaluating mitral valve biomechanics
2023, Medical Engineering and PhysicsArtificial intelligence applied to support medical decisions for the automatic analysis of echocardiogram images: A systematic review
2021, Artificial Intelligence in MedicineCitation Excerpt :Manual segmentation by specialists did not have significantly lower sensitivity (60% vs 90%, P < 0.001), specificity (91% vs 97%, P = 0.001), and precision (83% vs 95%, P < 0.001) compared to experts. Zhang et al. [30] proposed and developed a semi-automated framework that combined ML models with geometric and biomechanical image analysis models to build a specific representation of the patient's mitral valve that incorporated material properties derived from the image. They used 3D TEE images of the open and closed mitral valve.
3D Printing, Computational Modeling, and Artificial Intelligence for Structural Heart Disease
2021, JACC: Cardiovascular ImagingFinite element analysis of MitraClip procedure on a patient-specific model with functional mitral regurgitation
2020, Journal of BiomechanicsCitation Excerpt :The FMR model includes the MV, LV myocardium and papillary muscles (PM), detailed chordae structure and distribution, accurate leaflet geometry and thickness, dynamic MA and chordae origins, anisotropic hyperelastic material models, and human age- and gender-matched material properties. Additionally, by fine-tuning chordae lengths, chordae tethering forces due to PM displacement were accounted for accurate modeling of MV dynamics (Pham et al., 2017; Zhang et al., 2017). Compared with previous studies, the inclusion of the LV myocardium allowed for more realistic loading and boundary conditions, which in turn was essential to model and quantify the mitral apparatus deformation in response to MitraClip.
Estimation of in vivo constitutive parameters of the aortic wall using a machine learning approach
2019, Computer Methods in Applied Mechanics and EngineeringCitation Excerpt :To reduce computational complexity, some studies suggested the use of simplifications in material models and/or geometries. For example, Liu and Shi [2], Zhang and co-workers [3] and Franquet and co-workers [4] identified linear elastic material parameters. By assuming a perfect cylindrical shape of the arteries, Schulze-Bauer and Holzapfel [5] identified Fung-type material parameters, Stålhand, Olsson and Klarbring, Masson and co-workers [6–9] estimated material parameters using the constitutive model proposed by Holzapfel and co-workers [10], and Smoljkić and co-workers [11] estimated the Gasser–Ogden–Holzapfel (GOH) model [12] parameters.
Estimation of in vivo mechanical properties of the aortic wall: A multi-resolution direct search approach
2018, Journal of the Mechanical Behavior of Biomedical MaterialsCitation Excerpt :To estimate the in vivo mechanical properties, simplifications and assumptions were made in several studies. By assuming linear elastic constitutive relations, (Liu and Shi, 2009), (Zhang et al., 2017) and (Franquet et al., 2013) identified linear elastic material parameters from in vivo images. By assuming a perfect cylindrical shape of the arteries, (Schulze-Bauer and Holzapfel, 2003) identified Fung-type material parameters, (Masson et al., 2011, 2008; Olsson and Klarbring, 2006; Stålhand, 2009) estimated material parameters using the constitutive model proposed in (Holzapfel et al., 2000) together with geometrical parameters, (Smoljkić et al., 2015) estimated the Gasser–Ogden-Holzapfel (GOH) model (Gasser et al., 2006) parameters. (