3D Vessel Centerline Research Articles

Non-rigid registration based on hierarchical deformation of coronary arteries in CCTA images.

In this paper, we propose an accurate and rapid non-rigid registration method between blood vessels in temporal 3D cardiac computed tomography angiography images of the same patient. This method provides auxiliary information that can be utilized in the diagnosis and treatment of coronary artery diseases. The proposed method consists of the following four steps. First, global registration is conducted through rigid registration between the 3D vessel centerlines obtained from temporal 3D cardiac CT angiography images. Second, point matching between the 3D vessel centerlines in the rigid registration results is performed, and the corresponding points are defined. Third, the outliers in the matched corresponding points are removed by using various information such as thickness and gradient of the vessels. Finally, non-rigid registration is conducted for hierarchical local transformation using an energy function. The experiment results show that the average registration error of the proposed method is 0.987mm, and the average execution time is 2.137s, indicating that the registration is accurate and rapid. The proposed method that enables rapid and accurate registration by using the information on blood vessel characteristics in temporal CTA images of the same patient.

Point-Cloud Method for Automated 3D Coronary Tree Reconstruction From Multiple Non-Simultaneous Angiographic Projections.

X-ray angiography is the most commonly used imaging modality for the detection of coronary stenoses due to its high spatial and temporal resolution of lumen contour and its utility to guide coronary interventions in real time. However, the high inter- and intra-observer variability in interpreting the geometry of 3D vascular structure based on multiple 2D image projections is a limitation in the accurate determination of lesion severity. This could be addressed by the 3D reconstruction of the coronary arterial (CA) tree. The automated reconstruction of 3D CA tree from 2D projections is challenging due to the existence of several imaging artifacts, such as vessel overlap, foreshortening, and most importantly respiratory and cardiac motion. Along with these artifacts, the acquisition geometry introduces the possibility of generating false vessel segments in the reconstruction. Our approach aims to reduce the motion artifacts in angiographic projections by developing a new method for rigid and non-rigid motion correction. A novel point-cloud based approach is subsequently introduced for reconstruction of 3D vessel centerlines by iteratively minimizing the reconstruction error. The performance of the proposed 3D reconstruction is evaluated using angiographic projections from 45 patients, producing average reprojection errors of 0.092 ±0.055 mm and 0.910 ±0.352 mm for 3D centerlines reconstruction, when co-registered with the parent vessels on projection planes that were/were not used to derive the 3D reconstruction, respectively. A comparison of the reconstructed 3D lumen surface with optical coherence tomography (OCT) measurements has been performed, showing no statistically significant difference in the luminal cross-sections reconstructed with our method, compared to OCT.

Quantitative Analysis of 3D Artery Volume Reconstructions Using Biplane Angiography and Intravascular OCT Imaging.

Diameter and volume are frequently used parameters for cardiovascular diagnosis, e.g., to identify a stenosis of the coronary arteries. Intra-vascular OCT imaging has a high spatial resolution and promises accurate estimates of the vessel diameter. However, the actual images are reconstructed from A-scans relative to the catheter tip and imaging is subject to rotational artifacts. We study the impact of different volume reconstruction approaches on the accuracy of the vessel shape estimate. Using X-ray angiography we obtain the 3D vessel centerline and the 3D catheter trajectory, and we propose to align the A-scans using both. For comparison we consider reconstruction along a straight line and along the centerline. All methods are evaluated based on an experimental setup using a clinical angiography system and a vessel phantom with known shape. Our results illustrate potential pitfalls in the estimation of the vessel shape, particularly when the vessel is curved. We demonstrate that the conventional reconstruction approaches may result in an overestimate of the cross-section and that the proposed approach results in a good shape agreement in general and for curver segments, with DICE coefficients of approximately 0.96 and 0.98, respectively.

Numerical Simulation of Myocardial Bridging in Patients with Hypertrophic Cardiomyopathy

A myocardial bridge (MB), a congenital anomaly of the coronary artery, occurs when a segment of the epicardial coronary artery goes underneath the myocardium. MBs are often observed in the middle part of the left anterior descending (LAD) artery. MB squeezes the vessel wall periodically and induces hemodynamic abnormalities which are correlated with angina and myocardial ischemia. The level of hemodynamics disturbances induced by MB depends on the myocardial bridge length, the degree of myocardial contractility, thickness, and location [1]. Hypertrophic cardiomyopathy (HCM), characterized by abnormal thickening of the heart wall, is a leading cause of death in patients of all ages [2]. The prevalence and severity of myocardial bridging are higher among patients with HCM, up to 30% [3]. Therefore, understanding the effects of MB in HCM patients is of clinical significance. The purpose of this study was to determine the role of MB on coronary hemodynamics in HCM patients with various level of MB using computational fluid dynamics (CFD) modeling. We analyzed three HCM patients with MB in the middle LAD who underwent coronary angiography at Fuwai hospital (Tab. 1). All patients had a significant MB (diameter stenosis ≥ 50%) as diagnosed by arteriography. <br/> Biplane angiographic images were recorded at 15 frames/s using Artis zee III floor (Siemens Ltd. China) with a pixel spacing of 154μm. 2D vessel centerlines of two different oblique views were extracted from angiographic data and then combined to create a 3D vessel centerline in SolidWorks^®. Lumen diameter of LAD at both end-diastole and end-systole were measured to reconstruct the LAD model (Fig. 1). Linear interpolation was used to predict vessel wall movements during the cardiac cycle using a user-defined function (UDF). The arbitrary Lagrangian-Eulerian (ALE) method was employed for the moving boundaries. Blood was assumed to be homogenous Newtonian fluid with a density of 1050 kgm^-3 and dynamic viscosity of 0.00316 Pa.s. To ensure fully developed flow, extended lengths of 10-diameter and 15-diameter were added to the inlet and outlet sections, respectively [4]. Intracoronary pressure waveforms were applied as the inlet and outlet boundary conditions [5]. The model was meshed with hexahedral elements with a finer mesh near the wall. The shear stress transport (SST) turbulence model was used for simulating flow in ANSYS Fluent (v 17.1). The Navier-Stokes equations were solved using semi-implicit method (Simple) with the maximum residual of less than 10^-6. To achieve a periodic solution, three cardiac cycles were simulated and the last cycle results were used for the interpretation. <br/> To better understand the effect of MB, simulations were repeated for the same patient without MB (vessel compression was artificially removed). CFD results demonstrated that the presence of MB reduced the time-averaged wall shear stress (TAWSS) in the proximal (0.83 ± 0.29 Pa with MB versus 1.25 ± 0.59 without MB) and distal (2.76 ± 0.85 Pa with MB versus 3.82 ± 0.85 without MB) segments while increased TAWSS in the bridge segment (4.04 ± 1.29 Pa with MB versus 2.72 ± 1.23 without MB). The normal range of wall shear stress at arterial wall is from 1 to 7 Pa [6]. MB increased the maximum residence time of blood particles (Tab. 2), which may trigger thrombosis in the proximal segment. Increasing the degree of stenosis showed a significant decrease in the volume-averaged velocity and volumetric flow rate. Our analysis illustrated that the MB had significant effects on the distribution of wall shear stress and blood flow in HCM patients.

In vivo comparison of arterial lumen dimensions assessed by co-registered three-dimensional (3D) quantitative coronary angiography, intravascular ultrasound and optical coherence tomography

This study sought to compare lumen dimensions as assessed by 3D quantitative coronary angiography (QCA) and by intravascular ultrasound (IVUS) or optical coherence tomography (OCT), and to assess the association of the discrepancy with vessel curvature. Coronary lumen dimensions often show discrepancies when assessed by X-ray angiography and by IVUS or OCT. One source of error concerns a possible mismatch in the selection of corresponding regions for the comparison. Therefore, we developed a novel, real-time co-registration approach to guarantee the point-to-point correspondence between the X-ray, IVUS and OCT images. A total of 74 patients with indication for cardiac catheterization were retrospectively included. Lumen morphometry was performed by 3D QCA and IVUS or OCT. For quantitative analysis, a novel, dedicated approach for co-registration and lumen detection was employed allowing for assessment of lumen size at multiple positions along the vessel. Vessel curvature was automatically calculated from the 3D arterial vessel centerline. Comparison of 3D QCA and IVUS was performed in 519 distinct positions in 40 vessels. Correlations were r = 0.761, r = 0.790, and r = 0.799 for short diameter (SD), long diameter (LD), and area, respectively. Lumen sizes were larger by IVUS (P < 0.001): SD, 2.51 ± 0.58 mm versus 2.34 ± 0.56 mm; LD, 3.02 ± 0.62 mm versus 2.63 ± 0.58 mm; Area, 6.29 ± 2.77 mm2 versus 5.08 ± 2.34 mm2. Comparison of 3D QCA and OCT was performed in 541 distinct positions in 40 vessels. Correlations were r = 0.880, r = 0.881, and r = 0.897 for SD, LD, and area, respectively. Lumen sizes were larger by OCT (P < 0.001): SD, 2.70 ± 0.65 mm versus 2.57 ± 0.61 mm; LD, 3.11 ± 0.72 mm versus 2.80 ± 0.62 mm; Area 7.01 ± 3.28 mm2 versus 5.93 ± 2.66 mm2. The vessel-based discrepancy between 3D QCA and IVUS or OCT long diameters increased with increasing vessel curvature. In conclusion, our comparison of co-registered 3D QCA and invasive imaging data suggests a bias towards larger lumen dimensions by IVUS and by OCT, which was more pronounced in larger and tortuous vessels.

Calibration‐free device sizing using an inverse geometry x‐ray system

Quantitative coronary angiography (QCA) can be used to support device size selection for cardiovascular interventions. The accuracy of QCA measurements using conventional x-ray fluoroscopy depends on proper calibration using a reference object and avoiding vessel foreshortening. The authors have developed a novel interventional device sizing method using the inverse geometry scanning-beam digital x-ray (SBDX) fluoroscopy system. The proposed method can measure the diameter and length of vessel segments without imaging a reference object and when vessels appear foreshortened. SBDX creates multiple tomosynthetic x-ray images corresponding to planes through the patient volume. The structures that lie in the plane are in focus and the features above and below the plane are blurred. Three-dimensional localization of the vessel edges was performed by examining the degree of blurring at each image plane. A 3D vessel centerline was created and used to determine vessel magnification and angulation relative to the image planes. Diameter measurements were performed using a model-based method and length measurements were calculated from the 3D centerline. Phantom validation was performed by measuring the diameter and length of vessel segments with nominal diameters ranging from 0.5 to 2.8 mm and nominal lengths of 42 mm. The phantoms were imaged at a range of positions between the source and the detector (+/- 16 cm relative to isocenter) and with a range of foreshortening angles (0 degrees-75 degrees). Changes in vessel phantom position created magnifications ranging from 87% to 118% relative to isocenter magnification. Average diameter errors were less than 0.15 mm. Average length measurements were within 1% (0.3 mm) of the true length. No trends were observed between measurement accuracy and magnification. Changes in vessel phantom orientation resulted in decreased apparent length down to 28% of the original nonforeshortened length. Average diameter errors were less than 0.25 mm across all vessel angulations; errors were less than 0.1 mm for smaller diameter vessels and low to moderate vessel angles. Diameter errors increased with true diameter and vessel angle relative to the image plane. Average length measurement errors were also within 1% (0.3 mm) for each angulation. Tomosynthetic imaging with SBDX can accurately measure dimensions of vessels in various magnifications and angulations without calibration. This method may be more accurate and convenient than conventional QCA techniques.

Clinical evaluation of angiographic multiple-view 3D reconstruction

Endovascular treatment of stroke, a leading cause of death in the United States, involves guidance of devices to the intervention site often through tortuous vessels. Typically, these interventions are performed under two- dimensional (2D) fluoroscopy. To facilitate these procedures, we developed and previously presented a multiple-view self-calibration method involving two steps: (1) calibration of the imaging geometry, and (2) reconstruction of the 3D vessel centerline. Only those 2D angiograms obtained during the procedure are used for reconstruction. In this manuscript, we evaluate this technique on a large set (117 cases) of clinical data obtained over a 12-month period. We evaluated the technique using (1) the RMS difference between the calculated 3D centerlines and the average centerline (before and after application of our method), (2) the difference between the projected 3D centerlines and the 2D indicated centerlines, (3) the translations and rotations calculated by our technique, and (4) intra- and inter-user variations. Our approach (1) reduces the RMS 3D differences by a factor of 10, (2) increases the number of projected 3D centerline points lying within 1 mm of the indicated 2D centerline points by over a factor of 2 (from 28 to 71%), (3) provides an assessment of the variations in the gantry geometry as provided by the imaging system, and (4) is insensitive to user variations in indication (<1 mm differences in 3D are seen). These results indicate that this technique will provide more reliable vessel centerlines in the clinical setting without requiring additional acquisitions or increasing dose to the patient.

Using flow information to support 3D vessel reconstruction from rotational angiography

For the assessment of cerebrovascular diseases, it is beneficial to obtain three-dimensional (3D) morphologic and hemodynamic information about the vessel system. Rotational angiography is routinely used to image the 3D vascular geometry and we have shown previously that rotational subtraction angiography has the potential to also give quantitative information about blood flow. Flow information can be determined when the angiographic sequence shows inflow and possibly outflow of contrast agent. However, a standard volume reconstruction assumes that the vessel tree is uniformly filled with contrast agent during the whole acquisition. If this is not the case, the reconstruction exhibits artifacts. Here, we show how flow information can be used to support the reconstruction of the 3D vessel centerline and radii in this case. Our method uses the fast marching algorithm to determine the order in which voxels are analyzed. For every voxel, the rotational time intensity curve (R-TIC) is determined from the image intensities at the projection points of the current voxel. Next, the bolus arrival time of the contrast agent at the voxel is estimated from the R-TIC. Then, a measure of the intensity and duration of the enhancement is determined, from which a speed value is calculated that steers the propagation of the fast marching algorithm. The results of the fast marching algorithm are used to determine the 3D centerline by backtracking. The 3D radius is reconstructed from 2D radius estimates on the projection images. The proposed method was tested on computer simulated rotational angiography sequences with systematically varied x-ray acquisition, blood flow, and contrast agent injection parameters and on datasets from an experimental setup using an anthropomorphic cerebrovascular phantom. For the computer simulation, the mean absolute error of the 3D centerline and 3D radius estimation was 0.42 and 0.25 mm, respectively. For the experimental datasets, the mean absolute error of the 3D centerline was 0.45 mm. Under pulsatile and nonpulsatile conditions, flow information can be used to enable a 3D vessel reconstruction from rotational angiography with inflow and possibly outflow of contrast agent. We found that the most important parameter for the quality of the reconstruction of centerline and radii is the range through which the x-ray system rotates in the time span of the injection. Good results were obtained if this range was at least 135 degrees. As a standard c-arm can rotate 205 degrees, typically one third of the acquisition can show inflow or outflow of contrast agent, which is required for the quantification of blood flow from rotational angiography.

WE‐C‐L100J‐10: Three Dimensional Vessel Centerline Reconstruction and Imaging System Calibration From a Single Vessel in Angiograms

Purpose: Three dimensional (3D) vessel reconstructions can be useful in assisting clinicians with diagnosing and treating vascular disease, providing information about vessel diameter, length, and tortuosity. Reconstruction of 3D vessels from two views requires knowledge or calibration of the geometry relating the two imaging systems. We introduce a geometry calibration technique that performs this calibration using only a single indicated vessel segment in the angiograms. Methods: Vessel segments of interest were indicated in two angiographic views. The initial imaging geometry is estimated using each view's gantry information, and corresponding points along the vessels in both views are determined using epipolar constraints. The 3D position of the vessel is determined by triangulation of the found corresponding points. The geometry is corrected by varying nine of the imaging system's parameters using the Nelder‐Mead Downhill Simplex Method, with an objective function that minimizes the distance between the reprojection of the reconstructed 3D vessel centerlines and the 2D indicated vessel centerlines in both views. Results were compared with the enhanced‐Metz‐Fencil (EMF) geometry correction technique, which requires identification of additional corresponding points in both views. Results: Variations in the shapes obtained from our single vessel technique (SVT) and the EMF were comparable, median RMS of 0.47 and 0.34 mm, respectively, with magnification variations of 2.2% and 0.7%, respectively. Median errors in 2D reprojections of the 3D data for our technique and the EMF were both 0.12 mm, indicating very good agreement with the 2D indicated centerlines. Conclusion: We have developed an imaging geometry correction technique for two views based on alignment of reprojected 3D data with its respective 2D image information using only a single indicated vessel segment. This technique is reliable and comparable to other geometry correction techniques requiring additional user input.Supported by NIH grant ♯HL52567 and Toshiba Medical Systems Corporation.

Coronary x‐ray angiographic reconstruction and image orientation

We have developed an interactive geometric method for 3D reconstruction of the coronary arteries using multiple single-plane angiographic views with arbitrary orientations. Epipolar planes and epipolar lines are employed to trace corresponding vessel segments on these views. These points are utilized to reconstruct 3D vessel centerlines. The accuracy of the reconstruction is assessed using: (1) near-intersection distances of the rays that connect x-ray sources with projected points, (2) distances between traced and projected centerlines. These same two measures enter into a fitness function for a genetic search algorithm (GA) employed to orient the angiographic image planes automatically in 3D avoiding local minima in the search for optimized parameters. Furthermore, the GA utilizes traced vessel shapes (as opposed to isolated anchor points) to assist the optimization process. Differences between two-view and multiview reconstructions are evaluated. Vessel radii are measured and used to render the coronary tree in 3D as a surface. Reconstruction fidelity is demonstrated via (1) virtual phantom, (2) real phantom, and (3) patient data sets, the latter two of which utilize the GA. These simulated and measured angiograms illustrate that the vessel center-lines are reconstructed in 3D with accuracy below 1 mm. The reconstruction method is thus accurate compared to typical vessel dimensions of 1-3 mm. The methods presented should enable a combined interpretation of the severity of coronary artery stenoses and the hemodynamic impact on myocardial perfusion in patients with coronary artery disease.

SU-FF-I-62: Planning Image-Guided Endovascular Interventions: Models to Determine Point-Specific Vessel Tortuousity

Purpose: Clinicians often have difficulty passing guidewires and stents during endovascular interventions. Therefore, we are developing models to assess the access path based on stent as well as vessel dimensions. Method and Materials: Multi-view images of the carotid are acquired during clinical endovascular interventions. The vessel centerlines in the images are indicated and fit with splines. The vessel sizes are measured in the images. The multi-view imaging geometry is determined using the vessel information, and the three-dimensional (3D) vessel centerline is reconstructed using epipolar constraints. The vessel lumen is generated using the measured sizes and a circular cross-section model. Simulated models of devices are created and passed through the 3D patient-specific vessel. The models for this study included (a) a two-segment flexing device centered on the centerline, (b) a four-segment flexing device centered on the centerline, and (c) an unbound six-segment flexing device. The models were placed at all points along the vessel length and the point-specific tortuousity was calculated at each point. The tortuousity was calculated (a) as the ratio of the device length divided by the maximum length of the device inside the vessel, (b) and (c) as the inverse of the cosine of the average angular deviation from a straight device while keeping all segments inside the vessel. Results: The tortuousities calculated for each model are in good agreement with the apparent tortuousity of the vessel, with tortuousity decreasing from models one through three, reflecting the increased flexibility of the stent model. Conclusion: These models should provide the basis for more accurate assessment of the passage of devices through access vessels and perhaps be useful in decisions regarding stent lengths and flexibilities.

Generating of 3D data during neurovascular interventions by using multi-projection imaging

Three-dimensional (3D) vessel data from CTA or MRA are not always available prior to or during endovascular interventional procedures, whereas multiple 2D projection angiograms often are. Therefore, we are developing methods for combining vessel data from multiple 2D angiographic views obtained during interventional procedures to provide 3D vessel data during these procedures. Multiple projection views of vessel trees are obtained. Vessel regions to be analyzed are selected. One of the 2D images is selected as a common image. Initial pairwise imaging geometry relationships are calculated from the gantry information, and 3D vessel centerlines are calculated using pairwise epipolar constraints. The imaging geometries of each of the other views (relative to that of the common image) are then refined by using multidimensional optimization so as to minimize the differences between all pairwise calculated 3D vessel centerlines, and an average centerline is calculated. This final 3D centerline of the carotid vessel is used for calculation of the tortuousity, a quantity related to curvature. The centerlines calculated using the multiple projection approach are in better agreement (1.3 mm) than those calculated using the biplane technique (9.0 mm), and the tortuousity is more continuous along the vessel. Use of multiple projections improves 3D reconstruction of vessel centerlines and tortuousity reliability, which may facilitate image guided interventions.

Registration of vascular 3D data sets obtained from multiple-view reconstructions

Three-dimensional (3D) vessel data from CTA or MRA are not always available prior to or during endovascular interventional procedures, whereas multiple 2D projection angiograms often are. Therefore, we are developing methods for combining vessel data from multiple 2D angiographic views obtained during interventional procedures to provide 3D vessel data during these procedures. Multiple projection views of vessel trees are obtained. Vessel regions to be analyzed are selected. One of the 2D images is selected as a common image. Initial pairwise imaging geometry relationships are calculated from the gantry information, and 3D vessel centerlines are calculated using pairwise epipolar constraints. The imaging geometries of each of the other views (relative to that of the common image) are then iteratively refined so as to minimize the differences between all pairwise calculated 3D vessel centerlines, and an average centerline is calculated. Without refinement, the average difference between the centerlines is approximately 9 mm, whereas after refinement, the average difference is approximately 1 mm. In addition, the average centerline is smoother. Our results indicate that this technique will provide more reliable vessel centerlines without increasing dose to the patient.

An accurate coronary modeling procedure using 2D calibrated projections based on extracted 3D vessel centerlines

This paper describes a method for the three-dimensional (3D) reconstruction of coronary arterial segments from three projection boundaries. The method is based on the 3D generalized cylinder (GC) model with elliptic cross-sections and mainly consists of a model-based estimation of the vessel boundaries, precise computation of the ellipse parameters of arterial cross-sections from the estimated boundaries, and obtaining the 3D structures by linking consecutive cross-sections. Synthetic and real experiments showed that the method is robust to degradation sources and consequently superior to the two-view method because it uses boundary instead of gray-level information.