Three-Dimensional Ultrasound-Derived Physical Mitral Valve Modeling

Abstract

Purpose

Advances in mitral valve repair and adoption have been partly attributed to improvements in echocardiographic imaging technology. To further educate and guide repair surgery, we have developed a methodology to quickly produce physical models of the valve using novel 3D echocardiographic imaging software in combination with stereolithographic printing.

Description

Quantitative virtual mitral valve shape models were developed from 3D transesophageal echocardiographic images using software based on semi-automated image segmentation and continuous medial representation (cm-rep) algorithms. These quantitative virtual shape models were then used as input to a commercially available stereolithographic printer to generate a physical model of the each valve at end systole and end diastole.

Evaluation

Physical models of normal and diseased valves (ischemic mitral regurgitation and myxomatous degeneration) were constructed. There was good correspondence between the virtual shape models and physical models.

Conclusions

It was feasible to create a physical model of mitral valve geometry under normal, ischemic and myxomatous valve conditions using 3D printing of 3D echocardiographic data. Printed valves have the potential to guide surgical therapy for mitral valve disease.

Introduction

Despite the established superiority [1] of mitral valve repair, these procedures remain extensively underutilized relative to valve replacement [2]. This high degree of underutilization has been attributed to the steep learning curve associated with mitral valve repair surgery [2].

The development and adoption of repair techniques has always been closely link with advances in valve imaging – specifically echocardiography. Carpentier pioneered the advent of mitral repair procedures using only direct intraoperative observation. More generalized application of his techniques required the widespread availability of 2D echocardiography (2DE) in the late 1970’s and early 1980’s and progressed further with availability of intraoperative 2D transesphogeal echocardiography, beginning in the 1990’s [3–5]. Despite ostensibly overcoming many of the limitations of 2DE, the introduction of 3D echocardiography (3DE) has not had a similar transformative affect.

Commercially available 3DE analysis packages allow only for a limited number of quantitative measures to be made off-line. Custom software algorithms that permit interactive visualization and automated quantification have been developed but these techniques are time-consuming and labor-intensive [6–8].

It has become increasingly apparent that further advances in mitral repair may be dependent on improvements in enabling imaging technology. Herein we report such a technique. We have developed automated image analysis algorithms that quickly and quantitatively describe mitral valve geometry at any point in the cardiac cycle. These algorithms can be used to generate digital data input to commercially available 3D printers to produce physical models of mitral valves. Such tangible models may be useful in facilitating both the teaching and development of repair techniques.

Technology and Technique

Patient Recruitment

The study was approved by the Institutional Review Board of the University of Pennsylvania. Intra-operative 2D and 3D transesophageal echocardiography was performed in 4 patients after induction of anesthesia and prior to sternotomy. Two patients suffered from severe ischemic mitral regurgitation (IMR), one patient from severe myxomatous mitral regurgitation (MMR) and 1 patient had a normal mitral valve.

Image Acquisition

2D, m-mode, spectral and color Doppler and 3D echocardiographic data was collected with an ultrasound platform (iE33 Model, Philips Medical Systems, Andover, MA) equipped with a 2–7 MHz transesophageal matrix-array transducer. 3D data was collected during 4 consecutive heart beats, and the images were reconstructed with the following imaging parameters: frame rate = 17–30 Hz, depth =12–16 cm, image dimensions = 224 × 208 × 208 voxels, isotropic resolution = 0.6–0.8 mm³. Image analysis was performed on mid-systolic and diastolic image frames.

Image Segmentation and Geometric Modeling

Virtual models of the mitral valve were constructed from 3D ultrasound data using a 3D model-based segmentation method detailed in references [9,10]. Briefly, the algorithm involves several steps of user-initialization, including (1) identification of leaflet location along the long-axis dimension of the image volume and (2) outlining of the mitral annulus and anterior leaflet in 2D projection images generated from the 3D image. Points automatically identified on the leaflet surfaces are then detected and dilated to create a tight region of interest (ROI) containing the mitral leaflets. 3D active contour evolution generates binary segmentations of the mitral leaflets in the ROI. To geometrically model the mitral leaflets in these segmented images, a medial template (cm-rep) of the mitral leaflets is deformed through a Bayesian optimization process to capture leaflet geometry in the segmented image data.

3D Printing

The virtual model data was used as input to a Dimension Elite 3D Printer to create physical models from ABS plastic material. The stereolithographic file was imported into CatalystEX and final model dimensions were set. The point thickness of each layer was set for 0.254 mm (0.010 in). The highest resolution settings of the printer were 0.178 mm (0.007 in).

Clinical Experience

We created physical mitral valve models from one patient with normal valve function, who underwent cardiovascular surgery unrelated to the mitral valve. This patient had normal diastolic leaflet motion and systolic coaptation without mitral regurgitation by transesophageal echocardioraphy. A model in end-diastole was created from echocardiographic data in Figure 1A. Diastolic views of the virtual model and the physical model are shown in Figures 1B–E. Similar views of the virtual model and physical model in their mid-systolic configuration are shown in Figures 1F–I. The saddle shape of the mitral valve annulus can be appreciated from views of the valve in profile in diastole (Fig. 1D–E) and systole (Fig. HI). The systolic augmentation of annular saddle shape can be seen by comparing Fig. 1E to 1I.

Figure 1.

3DE, shape and physical models of a patient with normal mitral valve morphology and function. (A) Long-axis view of the left ventricle and atrium during diastole. The relative positions of the anterior (AL) and posterior mitral valve leaflets (PL) are indicated. (B) Diastolic virtual model and (C) physical model viewed obliquely from the atrium. (D) Diastolic virtual model and (E) physical model, from a commissure to commissure view. The saddle-shape annulus can be appreciated as it curves upward at the anterior annulus and downwards at the posterior annulus. (F) Systolic virtual model and (G) physical model viewed from the atrium. (H) Systolic virtual model and (I) physical model in profile, viewed obliquely from the aortic valve position.

Physical models from two patients with severe ischemic mitral regurgitation are shown in Figure 2. Comparison of the two patients demonstrates how variable the geometric valve distortions can be in IMR. In both patients, the annulus was dilated. In patient 1 (Figure 2 A–F), the annulus was markedly flatter than normal and there was both anterior and posterior leaflet tethering. In Patient 2 (Figure 2 G–L), the saddle shape is relatively preserved and the majority of leaflet tethering is on the P2 region of the posterior leaflet and is much more pronounced than in patient 1. Patient 2 also had thicker leaflets than patient 1.

Figure 2.

Virtual and physical models of two patients with severe ischemic mitral regurgitation depicted in systole. (A) Virtual model and (B) physical model viewed from the atrium. (C) Virtual model and (D) physical model as viewed from anterior to posterior commissure, in which the flat annulus can be appreciated (cfa Fig. 1D, E). (E) Virtual model and (F) physical model as viewed obliquely from the atrium. (G) Virtual model and (H) physical model of a second patient with ischemic mitral regurgitation viewed from the atrium. (I) Virtual model and (J) physical model viewed commissure to commissure. (K) Virtual model and (L) physical model as viewed obliquely from the atrium.

Virtual and physical models for a patient with moderate myxomatous valve degeneration are shown in Figure 3. The valve was moderately thickened and there was severe mitral regurgitation directed toward the septum, which corresponded with a partially flail P2 segment. This patient underwent mitral valve repair with resection of the flail segment and annuloplasty, which resolved the mitral regurgitation. The flat shape of the posterior annulus (Fig. 3D–E) should be compared to the saddle shape of the normal valve (Fig. 1).

Figure 3.

3DE, virtual and physical models of a patient with severe mitral regurgitation and a flail P2 leaflet segment. (A) Long-axis view of the left ventricle and atrium during systole. The relative positions of the anterior (AL, red) and posterior mitral valve leaflets (PL, green) are indicated. (B) Virtual model and (C) physical model viewed from the atrium. (D) Virtual model and (E) physical model as viewed in profile from anterior to posterior commisure.

Comment

Successful mitral valve repair offers significant clinical benefits over valve replacement. Current ACC/AHA guidelines recommend repair over replacement when valve morphology and surgical expertise indicate that the likelihood of successful is high [1]. Mitral valve repair has also been shown to be more cost-effective in the long-term than valve replacement [2]. In spite of the established benefits of valve repair, the procedure remains extensively underutilized relative to valve replacement.

Advances in the application of 2DE during the last two decades of the 20^th century improved and expanded the performance of mitral valve repair surgery. The integration of this pre-operative echocardiographic data with the results of intraoperative anatomic valve analysis represents the current state-of-the-art for mitral valve repair. This approach can be quite challenging for less experienced surgeons. It requires the assimilation of dynamic echocardiographic images with direct intra operative valve analysis in the surgically exposed, arrested heart - all needing to be done in an efficient manner under the time constraints of the aortic cross clamp.

Three-dimensional echocardiography (3DE) was introduced nearly a decade ago, and was heralded as a possible solution to the inherent complexities of teaching and performing mitral valve repair procedures. Unfortunately it has yet to have a demonstrable impact. Over the past decade our group has developed novel image analysis software that facilitates quantitative assessment of mitral valve leaflet and annular geometry [9]. The most recent of these advances have produced automated valve imaging algorithms that allow quantitative images of the valve to be made at any point in the cardiac cycle with.

The combination of our automated imaging techniques with routinely available 3D printing devices potentially provides a powerful tool for helping surgeons hone their mitral valve repair techniques. Having a physical model of the valve (at any or multiple points in the cardiac cycle) for the surgeon to observe and manipulate while interpreting preoperative 2D and 3D echocardiographic imaging should be helpful to both the experienced surgeon and surgeons dedicated to learning valve repair techniques.

These automated imaging methods are still in their early stages and require refinement to better describe leaflet coaptation, chordae, papillary muscle and left ventricular geometry. Considering the progress made to date we feel that the routine and rapid imaging of these structures will be achievable in the near future – at which point 3D printing devices will be able to make complete physical models of the mitral valve, at any point in the cardiac cycle, that will greatly facilitate operative planning, procedural improvements and the teaching of mitral valve repair procedures.