1 Introduction
Accurate description of vascular geometry is important for understanding vascular anatomy, physiology, and pathology. Morphologic changes to vascular anatomy are commonly associated with clinical diagnosis and assessment. In the case of aortic aneurysms, for example, the simple measurement of aneurysm diameter can be a predictor of rupture, and aneurysm volume has been shown to correlate with the risk of clinical sequelae [
3,
14]. Moreover, descriptions of dynamic anatomy, in the form of vascular deformations due to cardiac, respiratory, and musculoskeletal influences, can provide further insight into the physiological and pathological processes associated with disease development. For example, radial aortic compliance can help characterize degenerative disease in the aorta and lower extremities [
4,
5]. Axial deformation and elasticity of the superficial femoral artery can be used as an indicator of lower extremity arterial health and a predictor of stent fracture [
6]. Also, degree of in vivo deformation of implanted stents provides insight to predict long-term performance of the stents inside stenotic arteries [
20].
Three-dimensional (3D) medical imaging is capable of providing exquisite geometric information, from which 3D geometric models can be constructed. These geometric models can then be used to quantify vascular deformation for device evaluation and development as well as perform hemodynamic and vessel structure simulations [
2,
6,
8‐
10,
12,
13,
21‐
24]. One of the most established 3D lumen modeling methods is based on centerline construction, orthogonal 2D segmentations, and surface lofting [
28]. While these methods allow for analysis of motion and deformation of lumen centerlines and cross sections, they lack the ability to robustly and fully characterize 3D vascular surface geometry. For example, they cannot fully quantify the variation in surface curvature along a highly curved vessel, such as in the aortic arch.
Developing more nuanced methods to quantify 3D geometric and morphological features of the human vascular system, and their dynamic changes, is needed to better understand how devices interact with the vascular system and the biomechanical characteristics that determine a patient’s prognosis and potential response to treatment. For example, recent efforts in surface modeling and analysis have demonstrated excellent promise for better predicting aortic aneurysm rupture [
12,
16,
18,
19]. However, because the key parameter for evaluating mechanical fatigue of a medical device is based on alternating strain at a particular material points, a Lagrangian-based method to quantify deformation is also warranted. Here we present a method for creating a Lagrangian description of approximate cylindrical structures based on a cylindrical coordinate system. Building from a vessel centerline and lumen cross-sectional contours, this coordinate system can describe complex surface geometry, including longitudinal and circumferential curvature, cross-sectional eccentricity, and the orientation of eccentricity. We validate this method with idealized software phantoms and demonstrate the wide potential of this method by analyzing dynamic changes of blood vessel geometry in patient-specific examples of the thoracic aorta, abdominal aorta, and iliofemoral vein.
4 Discussion
In this paper, we describe a robust method for fully describing the surface geometry of complex anatomic tubular structures by using a Lagrangian cylindrical coordinate system. The method was validated on idealized software phantoms and then applied to three human data sets of different anatomies.
To ensure a robust and widely applicable system, window sizes for curvature calculation need to be standardized based on the anatomy of interest. The window size needs to be small enough to ensure accurate estimation of maximum curvature values, yet not so small as to cause substantial spurious oscillations. Based on experiments on the complex idealized phantom in Fig.
4b, the optimal window size for longitudinal and circumferential curvature calculations was 30 mm and π/4 rad, respectively. These correspond roughly the diameter of the vessel and one eighth of the circumference, respectively. These guidelines were used for all subsequent analyses.
Application to the simple and complex idealized phantoms from Fig.
4 shows excellent agreement with analytic solutions of surface curvature and eccentricity. Figure
7 shows that the straight sections of the simple phantom were correctly calculated to have 0 cm
−1 longitudinal curvature and that the inner and outer curvatures at the bend correctly corresponded with 2 and 4 cm radius of curvature, respectively. In addition, the circumferential curvature was correctly calculated to correspond with 1 cm radius of curvature (2 cm diameter).
Analysis of the complex idealized phantom had similar excellent correspondence with analytic solutions, such as longitudinal curvature values of 0.5 and 1.0 cm
−1 at the outer curve of the first bend and inner curve of the second bend, respectively (Fig.
8). For the contour cross sections, the circumferential curvatures calculated for the circular cross sections and the flat sections of the elliptical cross sections were 1.0 and 0.25 cm
−1, respectively, which perfectly match the analytic solutions. The calculated circumferential curvature of 1.5 cm
−1 at the vertices of the ellipse is 25% below the analytic solution; however, increased point sampling around the contour would greatly improve this calculation. The analytic eccentricity of an ellipse with a 4 cm major diameter and 2 cm minor diameter is
\( \sqrt{3}/2 \) (dotted pink line in Fig.
8d), which is almost exactly the value calculated by our method. In addition, the orientation of eccentricity calculated by the method closely approximated the π/2 and 0 rad analytic solutions (Fig.
8e). Both eccentricity graphs capture the transition point at the correct longitudinal location where the idealized model bent from one plane to an orthogonal plane.
Figure
9 shows applicability of the method to actual medical images of different vascular structures. For the thoracic aortic endograft, the eccentricity was calculated to be higher (less circular) at the ends of the graft, which makes sense due to the decrease in hoop strength at the free ends of the stent graft (Fig.
9d). In the abdominal aorta example, circumferential curvature calculations show higher curvatures at the normal proximal aorta (equivalent to diameter ≈ 2 cm), lower curvatures at the aneurysm (equivalent to diameter ≈ 6 cm), and moderate curvature at the transition between normal and aneurysmal sections (Fig.
9e). For the iliofemoral vein, the longitudinal curvature was predominantly 0.2 cm
−1 or equivalent to approximately 5 cm radius of curvature (Fig.
9f).
The CT images and 3D geometric models of the thoracic aortic endograft in Fig.
10a–d illustrate how difficult it is to qualitatively identify subtle differences in surface geometry and morphology of vessels at different physiological states. However, with the color map of pointwise longitudinal surface curvature (Fig.
10e, f), the coordinate system analysis shows the power of its quantitative sensitivity. For example, the diastolic phase (Fig.
10f) shows a larger region of high longitudinal curvature along the inner curve of the proximal descending aorta as compared to the systolic phase (Fig.
10e). This makes sense because the during the high pressure pulse of systole, we would expect the endograft to straighten slightly.
Several methods have been developed to model human vessels and quantify geometric curvature. Classic methods have employed 2D level set segmentation creating a set of orthogonal contours following vascular lumen, acquiring centerlines from the contours, and computing the curvature along the centerline [
6,
7,
21,
22,
24,
25]. Other available methods include 3D segmentation with growing seeds, acquiring centerline from inscribing spheres in lumenal surface, and computing the centerline curvature [
11,
15]. It is also possible to acquire surface curvature directly across 2D surface patches such as a built-in function provided by the Vascular Modeling Toolkit (VMTK) [
1]. The method presented in this study improved the classic centerline-based method by computing the surface curvature using 2D-segmented contours as an input. In addition to the traditional centerline curvature, we believe that surface curvature provides additional, relevant information for characterizing vascular dynamics and designing novel medical devices. Furthermore, the method in this study provides curvature output along surface lines, not 2D surface patches like the VMTK methods above. The main reason we computed surface curvature was to quantify inner and outer line geometries. These surface line curvatures are crucial to understand how medical device dynamically deform under in vivo condition and if in-stent restenosis occurs along the surface line with challenging geometry [
29]. The methods computing surface curvature across 2D surface patches can also calculate surface line curvature with secondary calculations and averaging, but this requires additional steps and relies on averaging across 2D areas which results in lower resolution calculations.
More complete characterization of vascular geometry may help predict disease severity, such as aneurysm rupture risk [
12,
16,
18,
19,
26], quantify pre- and post-operative geometric alterations to determine the mechanical impact of devices on the native anatomy [
25], establish boundary conditions with which to evaluate and predict device failures due to cyclic fatigue [
7], and more fully describe dynamic anatomy to come up with better device solutions. For example, in vivo arterial motions have often been implicated in cyclic mechanical fatigue and stent fracture [
6,
7,
13,
17,
21]. Additional aortic endograft design challenges for aneurysm or dissection repair include cardiac-induced deformation, hemodynamic forces, and vulnerable and complex anatomies [
27]. The Lagrangian coordinate system described in this paper improves our ability to evaluate deformations of anatomy and implanted devices at material points.
Because this cylindrical coordinate system method uses piecewise linear centerlines with linear interpolation between cross-sectional contours, the quality of the geometric model and analyses is dependent of the quality and quantity of the cross-sectional contours. Specifically, the coordinate system requires that the original 2D contours to be sufficiently orthogonal to the centerline and sufficiently densely-spaced. However, this method can include spline interpolation between centerline points and cross-sectional contours, relieving some of the need for densely packed contours. In addition, the method can be generalized to input a volumetric model, derived from any number of segmentation methods, and then use that model to create arbitrary cross-sectional contour densities based on need. These method extensions, along with applying these techniques to a wide range of anatomic structures including vascular, pulmonary, gastrointestinal, and reproductive, will be topics of future research.