3D reconstruction of coronary artery bifurcations from intravascular ultrasound and angiography

Coronary bifurcation lesions represent a challenging anatomical subset, and the understanding of their 3D anatomy and plaque composition appears to play a key role in devising the optimal stenting strategy. This study proposes a new approach for the 3D reconstruction of coronary bifurcations and plaque materials by combining intravascular ultrasound (IVUS) and angiography. Three patient-specific silicone bifurcation models were 3D reconstructed and compared to micro-computed tomography (µCT) as the gold standard to test the accuracy and reproducibility of the proposed methodology. The clinical feasibility of the method was investigated in three diseased patient-specific bifurcations of varying anatomical complexity. The IVUS-based 3D reconstructed bifurcation models showed high agreement with the µCT reference models, with r2 values ranging from 0.88 to 0.99. The methodology successfully 3D reconstructed all the patient bifurcations, including plaque materials, in less than 60 min. Our proposed method is a simple, time-efficient, and user-friendly tool for accurate 3D reconstruction of coronary artery bifurcations. It can provide valuable information about bifurcation anatomy and plaque burden in the clinical setting, assisting in bifurcation stent planning and education.

Interventional cardiology faces challenges in treating coronary bifurcation lesions because of low procedural success rates and increased rates of adverse cardiovascular events 1,2 .Because of these unique anatomical locations, different stenting strategies have been developed, which have been the subject of ongoing debate.To determine the best bifurcation stenting strategy and achieve favorable clinical outcomes, several factors, including bifurcation anatomy and disease extent, must be considered 3,4 .In recent years, interventional cardiologists have expressed the need for the three-dimensional (3D) representation of bifurcation anatomy and disease burden, as this can lead to a better understanding of the anatomical complexity of bifurcation disease and the optimization of stenting strategies.
In this study, we present a new method that combines IVUS imaging and angiography with advanced modeling techniques to perform 3D reconstruction of coronary bifurcations, including plaque materials.The study aimed to: (1) provide a detailed methodology for the 3D reconstruction of coronary bifurcations along with plaque materials, and (2) evaluate the methodology's accuracy, feasibility, and reproducibility in both patientspecific silicone bifurcation models and patient cases with varying degrees of disease.

Experimental studies
Experimental coronary artery models, flow chamber studies, and imaging procedures Figure 1 depicts a flowchart that explains the 3D reconstruction process.Three patient-specific silicone coronary bifurcation models were created using the previously described technique 28 .In brief, the 3D bifurcation geometries were created using two angiographic projections in 3D CAAS Workstation 8.2 (Pie medical imaging, Maastricht, The Netherlands).Negative moulds were created based on the geometries and 3D printed in acrylonitrile butadiene styrene.Polydimethylsiloxane was mixed with its curing agent before being poured into dry-clean moulds for curing.The silicone models were transferred to an acetone beaker to dissolve the acrylonitrile butadiene styrene material.The silicone models were placed in a custom-made flow chamber.A bioreactor circuit was connected to the flow chamber's inlet and outlet, allowing steady blood-mimicking flow at a 100 ml/ min rate at room temperature.
The models were first imaged with angiography in two projections with at least a 30-degree difference in viewing angles to provide the main vessel (MV) and side branch (SB) centerlines.The three silicone models were subjected to IVUS imaging with Opticross 6 HD, 60 MHz (Boston Scientific, Marlborough, MA, USA).The imaging catheter was advanced using a 6F guiding catheter.An automatic pullback was performed at a constant speed of 0.5 mm/s or 1.0 mm/s, with a distance of 0.017/0.033mm between two consecutive frames.All the pullback frames were analyzed offline, and the lumen sections were segmented using EchoPlaque 4.0 (INDEC Medical Systems, Los Altos, CA, USA) (Fig. 2).Finally, after injecting an iodinated contrast media (37%), the bifurcation models were imaged with µCT (Skyscan 1172 version 1.5, Antwerp, Belgium) with the following parameters: image pixel size 26.94 µm, voltage 100 kV, current 100 µA, and slice thickness 27 µm.The bifurcations were 3D reconstructed from µCT images with 3D medical imaging software (Rhino3dmedical, Mirrakoi, Switzerland) and smoothed with modeling software Meshmixer (Autodesk Research, New York, NY, USA).

IVUS processing for bifurcation lumen reconstruction
The detailed steps of the bifurcation lumen reconstruction are listed in Fig. 1.We applied Electrocardiogram (ECG)-gating to obtain the IVUS frames at the end-diastolic phase.The frames were uploaded into Grasshopper (GH) 3D, a Rhinoceros plug-in.These frames were then aligned and packaged in a straight line along the central path of the IVUS catheter.The packaged frames were oriented in 3D space by placing them perpendicular to the centerline and registering each frame centroid.The MV and SB frames were then rotated to align with the same bifurcation carina reference to match the relative position.The lofting function of GH created the original surfaces of MV and SB.This process's detailed steps can be found in our previous work on 3D reconstruction using optical coherence tomography (OCT) images 28 .To secure a smooth 3D reconstruction of the MV and SB surfaces, especially at the carina where the MV/SB surfaces exhibited irregular protrusions, in each IVUS frame, we identified the catheter point.All the catheter points of the entire sequence of IVUS frames created a virtual catheter path.We noticed that the catheter points at the carina were deviating.To correct the surface protrusion at the carina, we manually aligned the deviating catheter points (Fig. 3).To further refine the 3D reconstructed lumen surface, we mapped back the segmented IVUS frames to the 3D reconstructed surface (Fig. 4).Using a Rhinoceros tool, named subdivision surfaces (SubD) 29 , we were able to directly adapt the 3D geometry of the lumen to the segmented IVUS frames.

Validation of the 3D reconstruction method
The 3D IVUS reconstructed bifurcation models were compared to the corresponding 3D µCT reconstructed ones.The carina point was used to co-register the 3D IVUS and µCT reconstructed models.For comparisons, two metrics were used: lumen diameter and shape.To minimize biases, different operators performed the IVUSbased 3D reconstruction, µCT-based 3D reconstruction, and the comparisons of IVUS-and µCT-based models.
In the IVUS and µCT models, serial cross-sections were identified at every 0.1 mm along the lumen of the MV and SB.We observed a consistent difference in lumen diameter between IVUS and µCT (Supplementary Fig. S1), with IVUS measurements overestimating the actual lumen size 30 .For that reason, the lumen diameters were normalized using the z-score 31 to account for the systemic discrepancy in lumen size between the IVUS and µCT.
In each cross-section, the ratio of the maximum distance between the two furthest points of the circumference (distance X) to the maximum length perpendicular to distance X (distance Y) was calculated to compare the lumen shape.The lumen was assumed to be oval-shaped, and the ratio of distance Y/distance X was used as a lumen shape indicator.

Reproducibility
To calculate the reproducibility of the IVUS-based 3D reconstruction method, two independent operators 3D reconstructed all the silicone models, respectively.The reconstructed models were compared in terms of the lumen diameter and shape.

Clinical studies
The clinical feasibility and processing times of our bifurcation reconstruction method were evaluated in n = 3 patient coronary artery bifurcations (Supplementary Table S1).The above-mentioned imaging protocols were used to collect IVUS and angiography data.Using a customized platform (based on Rhinoceros and GH), we manually segmented the vessel lumen and EEM, as well as the plaque materials (calcium, fibrosis, and fibrolipid) (Fig. 3).To correct the twisting of the IVUS catheter due to the pulsatile motion of the artery 28 , we rotated the successive IVUS frames (including lumen, EEM, and plaque materials) around the catheter center until the segmented lumens were aligned (Supplementary Fig. S2).Finally, the correctly oriented IVUS frames (lumen, EEM, and plaque materials) were lofted and smoothened to create the 3D bifurcation reconstruction.The mapping back technique was also used to further refine the quality of the 3D reconstructed lumen, EEM, and plaque materials.To evaluate the time efficiency of our 3D reconstruction method, we calculated the processing time for each step in all three clinical cases.

Statistical analyses
Statistical analyses were performed with the statistical package GraphPad Prism 9.5 (GraphPad Inc., San Diego, CA, USA).Continuous variables were expressed as median (Inter quartile range; IQR).The lumen diameters of IVUS and μCT models were normalized by calculating the Z score as (absolute diameter-µ)/s with µ representing the mean diameter and s the standard deviation of the mean.The comparisons and reproducibility were performed with linear regression and Bland-Altman analysis.p value < 0.01 was considered as the level of significance.

Lumen diameter
All n = 3 silicone models were successfully 3D reconstructed (Fig. 5a).In the diameter/length graphs, the normalized lumen diameters (z-score) of the 3D reconstructed bifurcations from 3D IVUS vs. µCT show high agreement (Fig. 5b).Linear regression analysis revealed r 2 values ranging from 0.88 to 0.99, with slopes close to one and intercepts near zero (Table 1).

Reproducibility
Table 2 displays the reproducibility test results for our method.The lumen diameters of the 3D IVUS reconstructed bifurcation models from two independent operators showed high agreement, with r 2 ranging from 0.96 to 1.00 (p < 0.0001), indicating that our method was quite reproducible.

Clinical feasibility
The validated 3D reconstruction algorithm successfully reconstructed all n = 3 patient bifurcations.The reconstructed bifurcation models, including lumen, EEM, and plaque materials, are shown in Fig. 6.The lumen shape was in good agreement with the corresponding angiograms (Fig. 6b, f, j).The mapping back technique was used in all models to test the accuracy of the plaque materials reconstruction, showing that the reconstructed lumen and plaque materials matched their respective IVUS images in location and size (Fig. 6d, h, l).Table 3 summarizes the processing time for each step, from image processing to final 3D lumen and plaque reconstruction.The average time for image pre-processing was about 100 min, and for 3D reconstruction of a bifurcation about 60 min.

Discussion
In this study, we proposed a new method for 3D reconstruction of coronary bifurcations, including plaque materials, by combining IVUS imaging and coronary angiography.We performed in-vitro validation using silicone models and clinical feasibility tests, demonstrating that our reconstruction technique has high accuracy, reproducibility, and clinical feasibility.Our validated method has the potential to be used in the clinical setting to provide reliable information about the anatomy and plaque burden of coronary bifurcations, thereby facilitating clinical planning and decision-making during cardiac catheterization procedures.
The first studies on IVUS-based coronary vessel 3D reconstruction date back more than 20 years [5][6][7] .Thereafter, several studies have performed 3D reconstruction of non-bifurcated coronary arteries by combining IVUS with angiography [5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22][23] .To date, two studies have performed 3D reconstruction of coronary bifurcation arteries by combining IVUS and angiography, but both studies used CT to correct the IVUS frames' orientation 24,25 .Our study is www.nature.com/scientificreports/ the first to perform 3D reconstruction of coronary artery bifurcation by combining IVUS and angiography only without the need for CT imaging.While some previous studies have performed 3D reconstruction of coronary plaque materials based on IVUS, they have certain limitations.Balocco et al. 21performed segmentation and reconstruction of fibrotic, calcified, and fibro-lipidic plaques considering thickness, but simplified the models by assigning plaques' morphological values to vessel surface mesh.By collecting the IVUS-VH (Virtual Histology) cross-section image, Liang et al. 26 reconstructed fibrous tissue, necrotic core, and calcification for single vessels.Following the 3D reconstruction methods proposed by Tang et al. 32 , a series of works applying IVUS or IVUS-VH were performed by Tang's research group 22,23,27 .They reconstructed 3D models of fibrous tissue, lipid core, and calcification but the reconstructions were limited to non-bifurcated vessels.Compared to these prior works, our study provides a more comprehensive 3D bifurcation reconstruction incorporating all major coronary plaque materials: calcium, fibrosis, and fibro-lipid.Table 4 provides a comprehensive comparison of our methodology to that of other published studies.We validated our work using 3D models reconstructed from µCT imaging, widely accepted as a gold standard 33,34 , after successfully reconstructing patient-specific silicone bifurcation models based on IVUS and angiography.The linear regression analysis on lumen diameter revealed that the r 2 values ranged from 0.88 to 0.99, with a mean difference of -0.003 in the median lumen shape ratio.The linear regression analysis of the reproducibility test, performed by two independent operators, revealed high r 2 values ranging from 0.96 to 1.00, suggestive of the high accuracy and reproducibility of the proposed method.During the clinical feasibility tests, the mapping back of the IVUS images onto the corresponding 3D reconstructed model demonstrated that our 3D reconstruction method could accurately locate and size the plaque materials.
Notably, in our work, we used two innovative techniques: (1) creation of the virtual catheter path to correct the irregular surface protrusion at the carina, and (2) mapping back technique combined with the SubD surfaces, which allowed us to directly modify and adapt the 3D reconstructed bifurcation model to the segmented IVUS frames (Supplementary Fig. S4).This study's findings have important clinical implications and future applications.Our 3D reconstruction method has the potential to improve procedural planning and clinical outcomes by providing interventional cardiologists with precise information about the anatomy and severity of coronary bifurcation disease.Our technique can also perform patient-specific computational simulations for bifurcation stenting.This can advance knowledge about bifurcation stenting mechanics and potentially lead to more personalized stenting techniques.This can be especially beneficial to the pharmaceutical industry when developing next-generation stents.Furthermore, if the imaging data is available to extract the required information, the innovative methodology of our technique has the potential to improve other invasive imaging modalities, such as OCT, or even non-invasive imaging, such as coronary computed tomography angiography.Finally, when combined with improved visualization techniques, the accurate 3D reconstructed bifurcation models can provide medical students with novel education, such as flying through the vessel using virtual reality.

Limitations
There were several limitations in our study.First, our study was performed on a small dataset (three silicone models and three clinical cases).However, the primary goal of our proof-of-concept study was to introduce a novel 3D reconstruction method by IVUS and angiography, validate this with silicone models, and show its clinical applicability.Second, the back-and-forth movement of the IVUS catheter (swinging effect) during pulsation could affect the accuracy of the 3D reconstruction 35 .However, the application of ECG-gating could minimize the swinging effect, as suggested by the previous works.Finally, manual segmentation of IVUS images and plaque material is a time-consuming process.To address this issue, we are currently working on machine learning-based codes that can be integrated into GH to fully automate the IVUS segmentation process, reducing processing time and allowing our algorithm to be applied in near real-time.

Conclusion
This study presents a new methodology for 3D reconstruction of coronary artery bifurcations using IVUS and angiography.The agreement between the IVUS-based 3D reconstructed bifurcation models and the µCT reference models was remarkably high, indicating the accuracy of our approach.Furthermore, our methodology was found to be clinically feasible and time-efficient for three patient-specific bifurcations, with a reconstruction time of about 60 min.Our technique provides a simple, easy-to-use, and accurate 3D reconstruction of coronary bifurcations, which can help with pre-procedural planning and clinical decision-making for bifurcation stenting procedures.

Figure 1 .
Figure 1.Flowchart of the 3D reconstruction of coronary artery bifurcation for patient-specific silicone models and clinical cases.IVUS, intravascular ultrasound; MV, main vessel; SB, side branch.

Figure 3 .
Figure 3. Virtual catheter pathway method to correct surface protrusions near carina during reconstruction.

Figure 6 .
Figure 6.3D reconstruction of coronary artery bifurcation geometries.(a, e, i) Reconstructed lumen of n = 3 patient-specific geometries using angiography and IVUS imaging.(b, f, j) Angiography of the reconstructed coronary bifurcation lumen showing the main vessel (MV) and side branch (SB).(c, g, k) Reconstructed lumen, calcium, fibrosis, and fibrolipid of n = 3 patient-specific geometries using angiography and IVUS imaging.(d,h, l) Mapping back of a cross-section of the reconstructed lumen, calcium, fibrosis, and fibrolipid to its respective IVUS image, showing the high accuracy of the reconstruction.IVUS, intravascular ultrasound; MV, main vessel; SB, side branch.