Fusion of 3D QCA and IVUS/OCT

Authors: Shengxian Tu Niels R Holm Gerhard Koning Zheng Huang Johan H C Reiber

Publish Date: 2011/01/25

Volume: 27, Issue: 2, Pages: 197-207

PDF Link

Abstract

The combination/fusion of quantitative coronary angiography QCA and intravascular ultrasound IVUS/optical coherence tomography OCT depends to a great extend on the coregistration of Xray angiography XA and IVUS/OCT In this work a new and robust threedimensional 3D segmentation and registration approach is presented and validated The approach starts with standard QCA of the vessel of interest in the two angiographic views either biplane or two monoplane views Next the vessel of interest is reconstructed in 3D and registered with the corresponding IVUS/OCT pullback series by a distance mapping algorithm The accuracy of the registration was retrospectively evaluated on 12 silicone phantoms with coronary stents implanted and on 24 patients who underwent both coronary angiography and IVUS examinations of the left anterior descending artery Stent borders or sidebranches were used as markers for the validation While the most proximal marker was set as the baseline position for the distance mapping algorithm the subsequent markers were used to evaluate the registration error The correlation between the registration error and the distance from the evaluated marker to the baseline position was analyzed The XAIVUS registration error for the 12 phantoms was 003 ± 032 mm P = 075 One OCT pullback series was excluded from the phantom study since it did not cover the distal stent border The XAOCT registration error for the remaining 11 phantoms was 005 ± 025 mm P = 049 For the in vivo validation two patients were excluded due to insufficient image quality for the analysis In total 78 sidebranches were identified from the remaining 22 patients and the registration error was evaluated on 56 markers The registration error was 003 ± 045 mm P = 067 The error was not correlated to the distance between the evaluated marker and the baseline position P = 073 In conclusion the new XAIVUS/OCT coregistration approach is a straightforward and reliable solution to combine Xray angiography and IVUS/OCT imaging for the assessment of the extent of coronary artery disease It provides the interventional cardiologist with detailed information about vessel size and plaque size at every position along the vessel of interest making this a suitable tool during the actual interventionOver the past decades the continuous developments in coronary visualization and quantitative systems have been motivated by the increasing need to better understand and assess coronary atherosclerosis and by the online need for support of coronary interventions in cardiac catheterization laboratories Recently developed threedimensional quantitative coronary angiography 3D QCA systems aimed to resolve some of the limitations in conventional twodimensional 2D analysis 1 2 3 and hence to extend its capacity and reliability in assessing the true dimensions of coronary vascular structures It has been demonstrated that 3D QCA can accurately assess vessel segment length and diameter 4 5 6 7 as well as the optimal viewing angles 8 9 10 for the subsequent interventional stentprocedure By using 3D QCA and based on such more accurate 3D data clinical decision making can be affected thus possibly leading to a more efficient and economic usage of stents in percutaneous coronary intervention PCI 11 This may have significant impact in today’s costconstrained health care systemsDespite the fact that the 3D angiographic reconstruction has important potential values the foremost limitation of Xray angiographybased systems remains the inability to image beyond the vessel lumen as only the contrast lumen is visualized In other words the 3D reconstructed vessel remains a lumenogram though with better 3D capabilities Thus early stages of plaque formation may not be evident with Xray angiography due to the occurrence of coronary artery remodeling 12 and vulnerable plaques can not be recognized for possible implementations of measures to prevent these from rupturing These limitations have been well addressed by intravascular tomographybased imaging techniques among which grey scale intravascular ultrasound IVUS is a wellestablished and validated modality IVUS provides a wealth of information including vessel wall composition which is crucial to the assessment of coronary atherosclerosis Later on the role of intravascular tomographybased imaging techniques was greatly enhanced by the radiofrequency data analysis for plaque characterization and optical coherence tomography OCT for the assessment of the thin fibrous cap atheromas and malapposition of stent struts These new imaging techniques have extended the capabilities in the assessment of coronary artery disease However the fact that intravascular tomographybased imaging does not preserve the global topology information could lead to erroneous interpretations Although a longitudinal view LView is available in most IVUS/OCT consoles to provide an overview of the pullback series the presentation of the LView by stacking crosssectional images along a straightened version of the transducer pullback trajectory is a very unnatural way of conceptualization As a result the interpretation can be quite challengingGiven the different but complementary perspectives provided by Xray angiography XA and IVUS/OCT the fusion/integration of the two imaging modalities by using XA as a roadmap while exploiting detailed vessel wall information from IVUS/OCT will benefit the interpretation of coronary artery disease and the guidance of coronary interventions Currently if IVUS/OCT is performed in the preintervention stage the treatment planning is determined to a great extent by the IVUS/OCT interpretation However since XA fluoroscopy is still the only imaging tool available during stent deployment and positioning the interventionalist must mentally establish the correspondence between XA and IVUS/OCT images This spatial corresponding process is not always easy especially for diseases of early stages or long diffused lesions where lumen narrowing is not clearly evident and no sidebranch is present in the neighborhood of the lesion borders Thus XAIVUS/OCT integrated systems are currently requested in the market to better support coronary interventions The clinical applicability of such fused/integrated systems depends to a great extend on the reliability and robustness of the coregistration approach Once a reliable correspondence between angiographic and IVUS/OCT images is established the issue of fusing/integrating information from the two image modalities becomes relevantAccurate and robust 3D angiographic reconstruction is the foremost important step in the XAIVUS/OCT coregistration Early research on 3D reconstruction can be traced back to decades ago 13 14 However clinicready systems were announced only in recent years and there have not been widespread acceptance of such systems in routine clinical practice One of the reasons is due to the fact that mechanical distortions in Xray systems and noisy angiographic images in routine clinical acquisitions could significantly affect the reliability and robustness of the 3D reconstruction and analysis For monoplane Xray angiographic acquisitions the shift of the whole coronary tree due to the patient’s respiration or the nonisocentric condition could greatly deteriorate the system’s reliability Such system distortions should be corrected before or during the 3D angiographic reconstructionA number of approaches 13 14 15 have been proposed to correct for angiographic system distortions Ideally a couple of reliable features eg catheter tip and sidebranches should be identified on the two angiographic views as reference points for the correction of system distortions However the practical applicability of such approaches in online usage has been hampered by the efforts in identifying many reliable features which turned out to be too time consuming or even impossible to find such reliable features on the two angiographic views especially when there were many overlaps from different vessel segments To guarantee the reliability in the identification of reference points has already been a nontrivial taskTo come up with a more practical and attractive workflow we have developed a new approach by using only one to three pairs of reference points for the correction of system distortions In case of the presence of small perspective projection angles for noisy angiographic images the reliability and robustness of the angiographic reconstruction are further improved by constructing a distance transformation matrix and by searching for the optimal corresponding path in the matrix to refine the correspondence between the two angiographic views 8 The approach has been validated with high accuracy in both phantom and in vivo data 5 8 In short the 3D angiographic reconstruction consists of only a few major steps 1 load two image sequences acquired at two arbitrary angiographic views at least 25 degrees apart in viewing angles 2 select the enddiastolic image frames with the vessel lumen well filled with contrast from the two image sequences for the subsequent 3D reconstruction 3 identify one to three reference points in both angiographic views for the automated correction of system distortions 4 manually define the vessel segment of interest and extract its contours and centerlines using our extensively validated QCA algorithms 1 16 17 in the two angiographic views 5 reconstruct the 3D centerline and crosssections after refining the correspondence between the two extracted centerlines

Keywords: