Abstract
Purpose
To allow for comprehensive noninvasive diagnostics of coronary artery disease (CAD) by using three-dimensional (3D) image fusion of CT coronary angiography, CT-derived fractional flow reserve (CT FFR), whole-heart dynamic 3D cardiac MRI perfusion, and 3D cardiac MRI late gadolinium enhancement (LGE).
Materials and Methods
Seventeen patients (54 years ± 10 [standard deviation], one female) who underwent cardiac CT and cardiac MRI were included (combined subcohort of three prospective trials). Software facilitating multimodal 3D image fusion was developed. Postprocessing of CT data included segmentation of the coronary tree and heart contours, calculation of CT FFR values, and color coding of the coronary tree according to CT FFR. Postprocessing of cardiac MRI data included segmentation of the left ventricle (LV) in cardiac MRI perfusion and cardiac MRI LGE, co-registration of cardiac MRI to CT data, and projection of cardiac MRI perfusion and LGE values onto the high spatial resolution LV from CT.
Results
Image quality was rated as good to excellent (scores: 2.5–2.6; 3 = excellent). CT coronary angiography revealed significant stenoses in seven of 17 cases (41%). CT FFR was possible in 16 of 17 cases (94%) and showed pathologic flow in seven of 17 cases (41%), six of which coincided with cases revealing significant stenoses at CT coronary angiography. Cardiac MRI perfusion identified eight of 17 patients (47%) with hypoperfusion (ischemic burden of 17% ± 5). Cardiac MRI LGE showed myocardial scar in three of 17 cases (18%, scar burden of 7% ± 4). Conventional two-dimensional readout of CT coronary angiography and cardiac MRI resulted in eight of 17 cases (47%) with uncertain findings. Most of these divergent findings could be solved when adding information from CT FFR and 3D image fusion (six of eight, 75%).
Conclusion
Multimodal 3D cardiac image fusion is feasible and may help with comprehensive noninvasive CAD diagnostics.
Supplemental material is available for this article.
© RSNA, 2020
Summary
While reading different imaging modalities for the diagnostic assessment of coronary artery disease in a sequential manner can lead to divergent findings, multimodal three-dimensional image fusion may help to overcome these difficulties.
Key Points
■ The study introduces a method for comprehensive noninvasive diagnostics of coronary artery disease; multiple pathologic aspects of the disease are visualized by using multimodal multiparametric three-dimensional (3D) image fusion and advanced 3D rendering techniques.
■ Information on cardiac and coronary morphology (by CT coronary angiography), hemodynamic relevance of coronary lesions (CT-derived fractional flow reserve [CT FFR]), myocardial perfusion deficits (whole-heart dynamic 3D cardiac MRI perfusion), and extent and transmurality of myocardial scar (3D cardiac MRI late gadolinium enhancement) are included.
■ Conventional two-dimensional readout of CT and MR images resulted in a substantial number of uncertain findings in our study (eight of 17 cases, 47%); most of these divergent findings could be solved when including additional information from CT FFR and 3D image fusion (six of eight, 75%).
Introduction
Accurate diagnostic assessment of coronary artery disease (CAD) is essential for identifying patients and coronary lesions that are likely to benefit from revascularization (1). For all pathologic aspects of the disease, different diagnostic instruments have been established. Because of high spatial resolution, CT lends itself to accurate depiction of coronary arteries (2). To not only assess coronary morphology but also estimate hemodynamic relevance of coronary stenoses, computing the fractional flow reserve from CT images (CT FFR) has been proposed (3,4) and was found to be superior to CT coronary angiography alone (5,6). For myocardial ischemia testing, cardiac MRI stress perfusion proved to be of high diagnostic value (7) without the disadvantage of ionizing radiation. Using cardiac MRI late gadolinium enhancement (LGE), myocardial scar can be distinguished from viable tissue (8). Advanced three-dimensional (3D) acquisition sequences apply 3D excitation and readout instead of sequential acquisition of separate slices (9). Three-dimensional cardiac MRI perfusion acquisition proved accurate at increased spatial resolution (10), whereas 3D cardiac MRI scar imaging provided significantly shorter acquisition times (11).
To assess all above-mentioned aspects of CAD, multimodal imaging can be used (12). To date, results from different imaging modalities are often viewed side by side or in a sequential manner by different readers. To overcome separate readout, different concepts of cardiac image fusion have been proposed (13), demonstrating improved specificity (14).
The most common approach of cardiac image fusion combines CT coronary angiography with myocardial perfusion imaging by SPECT (15). Although it is an excellent long-term predictor of cardiac events (16), the technique uses ionizing radiation. Hybrid PET with cardiac MRI combines morphologic, metabolic, and perfusion information (17) but lacks anatomic information of coronary arteries. Fusion of CT coronary angiography and CT perfusion imaging was introduced (18). Although the combination of these modalities was found to be a good predictor of cardiac events (19), both examinations apply radiation. Seeking to apply the lowest level of radiation possible and to offer the most accurate assessment of CAD, fusion of CT coronary angiography and cardiac MRI perfusion has been introduced for two-dimensional (2D) (20) and 3D (21) acquisition sequences. However, these approaches do not cover myocardial viability.
In summary, to our knowledge, all current approaches toward multimodal cardiac imaging investigate only a subset of pathologic aspects of CAD. In this article, we present a framework for 3D image fusion of CT coronary angiography, CT FFR, whole-heart dynamic 3D cardiac MRI perfusion, and 3D cardiac MRI LGE.
Materials and Methods
Patients
Seventeen patients (53.9 years ± 9.5, range: 32–71 years, one female) who prospectively underwent cardiac MRI between May 2009 and January 2016 because they were suspected of having or known to have CAD were included (combined subcohort of three trials; for details, see Appendix E1 [supplement]). All included patients also underwent cardiac CT and coronary angiography because of clinical indications. None of the patients underwent interventional or surgical revascularization between cardiac MRI and CT imaging. The decision on subsequent treatment was not based on the results of this work. The study had local ethics committee approval, and all patients provided written informed consent. For details on image acquisition, please see the Appendix E1 (supplement).
Conventional 2D Readout
For all conventional 2D readouts, readers were blinded to the patients’ histories and symptoms as well as the results from other imaging tests.
CT coronary angiography.—CT coronary angiography images were analyzed by one reader (H.A., with 14 years of experience in cardiac CT). Image quality was rated on an ordinal Likert scale: score of 3 = excellent, score of 2 = good, score of 1 = moderate but suitable for diagnostic investigation, and score of 0 = insufficient image quality. Coronary artery stenoses narrowing the vessel diameter by more than 50% (hereafter referred to as significant stenoses) were noted. Nomenclature followed applicable guidelines (22). The total Agatston score was documented (23).
CT-derived FFR.—CT FFR analysis was performed by two readers in consensus (M.M. and H.M., both in training). CT FFR values were derived from CT coronary angiography data sets by using a deep learning model trained on a large database of synthetic vessels (3). A cutoff point of 0.75 was used to discriminate pathologic stenosis (hereafter called hemodynamically relevant stenosis). Although a threshold of 0.8 is typically used for invasive FFR, some aspects specific to CT FFR may lead to lowered CT FFR values and may need to be considered during CT FFR interpretation (24). Thus, the lower CT FFR cutoff value of 0.75 was chosen in favor of higher specificity over sensitivity (25).
Cardiac MRI perfusion.—Cardiac MRI perfusion readout was performed by one reader (R.M., with 16 years of experience in cardiac MRI). Image quality was rated using the same Likert scale used for CT coronary angiography. Perfusion deficits were noted according to standard supply territories of the coronary arteries (26).
Cardiac MRI LGE.—Readout was performed by one reader (R.M). Image quality was rated on the same Likert scale used for CT coronary angiography and cardiac MRI perfusion. Regions of myocardial scar were noted according to standard supply territories (26).
Coronary angiography.—In all patients, coronary angiography was performed because of clinical indication. Results from clinical reports were documented on a per-vessel basis without reassessment of original images. If available, invasively measured FFR values were noted. Stenoses with 50% or greater luminal narrowing were considered significant (27).
Postprocessing for 3D Image Fusion
The data flow diagram in Figure 1 illustrates all postprocessing steps. To overcome the mismatch of CT and cardiac MRI data acquired at different time points and in different geometries, preceding studies introduced (21) and refined (18,28) a technique to derive high-resolution, anatomically correct left ventricle (LV) templates from CT coronary angiography data sets onto which cardiac MRI perfusion and scar values could be projected. The approach also solves problems that occur due to low resolution of cardiac MRI data, coarse manual segmentation, and interpolation during co-registration (21). All postprocessing steps are described in brief in the following. For in-depth information on the applied technique, reference is made to the aforementioned publications. Further information on 3D image fusion and advanced rendering techniques are provided in the Appendix E1 (supplement) and Figure 2.
Figure 1:
Data flow diagram illustrates CT and cardiac MRI postprocessing. From CT, segmented heart contour and color-coded coronary tree were derived. From cardiac MRI, cardiac MRI perfusion and cardiac MRI late gadolinium enhancement (LGE) images were derived, segmented, projected onto high-resolution templates of left ventricle (LV), co-registered, and color-coded. Finally, all four resulting volume data sets were rendered in common three-dimensional scene. CT FFR = CT-derived fractional flow reserve.
Figure 2a:
Images demonstrate image-based lighting and Disney’s “principled” reflectance model. (a) Multiple photographs of Andreas Grüntzig catheter laboratory were taken (exemplarily, one panoramic shot is shown). Pictures were assembled to a cube map, which projected entire 720° environment onto six faces of a cube and served as the basis for highly detailed real-world lighting in context of cardiac interventional suite. (b) To demonstrate effect, data from one nonpathologic CT coronary angiography were rendered three times with different surface qualities defined by the “principled” reflectance model. Fully reflective surface mirrors surroundings of catheter laboratory (left); glassy appearance is both reflective and translucent (right); polished red surface texture demonstrates interplay of all rendering aspects (center).
Figure 2b:
Images demonstrate image-based lighting and Disney’s “principled” reflectance model. (a) Multiple photographs of Andreas Grüntzig catheter laboratory were taken (exemplarily, one panoramic shot is shown). Pictures were assembled to a cube map, which projected entire 720° environment onto six faces of a cube and served as the basis for highly detailed real-world lighting in context of cardiac interventional suite. (b) To demonstrate effect, data from one nonpathologic CT coronary angiography were rendered three times with different surface qualities defined by the “principled” reflectance model. Fully reflective surface mirrors surroundings of catheter laboratory (left); glassy appearance is both reflective and translucent (right); polished red surface texture demonstrates interplay of all rendering aspects (center).
For assessing interreader reliability, readily prepared 3D fusion images were interpreted by two readers (J.v.S., R.M.) independently. To assess intrareader reliability, one reader (R.M.) repeated the readout after 1 day.
CT coronary angiography.—The heart contour and coronary tree were segmented automatically (Syngo.via; Siemens Healthineers, Forchheim, Germany). Because the two resulting data sets had the same origin, orientation, and dimensions, no additional image registration was necessary. CT FFR results were projected onto the presegmented CT coronary tree.
From each CT coronary angiography data set, two LV templates for later projection of cardiac MRI perfusion and LGE values were generated (28): First, an elongated rotation ellipsoid (so-called prolate spheroid) with dimensions resembling the individual LV size was defined (approximately 11 × 8 × 8 cm3). The spheroid was masked to the heart contour segmentation described above. Second, the resulting template was eroded by two voxels. While the residual smaller template was used for cardiac MRI perfusion projection, the enveloping two-voxel border was used for projection of cardiac MRI LGE values (28).
Cardiac MRI perfusion.—The image frame with maximum perfusion contrast was chosen from the dynamic perfusion data set. The LV myocardium was segmented manually, excluding basal slices with the LV outflow tract and low-signal apical slices. Areas of pathologic hypoperfusion were identified automatically based on a region of interest manually drawn in healthy myocardium (so-called remote tissue): analyzing the histogram of perfusion values inside the region, pathologic hypoperfusion was assumed for all myocardial voxels with a grayscale value more than 2 standard deviations σ below the average healthy signal intensity µ (10,28). The amount of hypointense myocardium relative to the total LV volume was calculated (so-called ischemic burden) (10,28). Because some perfusion images revealed small areas with hypointense signal, only hypointensities with an extent greater than or equal to 5% of the LV myocardium and apparently following coronary territories were considered pathologic.
Cardiac MRI perfusion data sets were coregistered to CT coronary angiography data by matching three distinct landmarks manually defined in both data sets (ie, basal LV center, apical LV center, and LV outflow tract). Co-registration results were redefined by applying automatic rigid registration (21). The obtained perfusion values were projected onto the inner high-resolution LV template derived from CT coronary angiography data, as explained above.
Cardiac MRI LGE.—For cardiac MRI LGE postprocessing, the LV myocardium was manually segmented on all applicable slices. If scar tissue was present, then LGE was assumed in all voxels with gray values greater than or equal to 0.4 times the maximum myocardial signal intensity (adapted full-width-half-maximum criterion) (28). The amount of LGE relative to the total LV volume was calculated (so-called scar burden) (28,29). On lines escaping radially from the LV’s center, the percentage of scar tissue relative to the myocardial wall width was calculated (so-called transmurality). Cardiac MRI LGE data sets were coregistered to CT coronary angiography data using the transformation matrix previously defined for cardiac MRI perfusion and CT coronary angiography registration. The resulting transmurality values were projected onto the outer high-resolution LV template derived from CT coronary angiography data, as explained above.
Estimation of postprocessing time.—While 3D postprocessing was complex because of the current prototypic architecture, the theoretical time overhead using a future, specifically engineered software tool can be estimated based on data from the literature. No or little user interaction is needed for CT coronary angiography segmentation. Time for CT FFR calculation was reported to be approximately 45 minutes (4,25). Manual segmentation of cardiac MRI perfusion and definition of healthy remote tissue takes 5 minutes ± 1 and 1 minute ± 0, respectively (28). Time for manual segmentation of cardiac MRI LGE is 5 minutes ± 1 (28). Co-registration of CT coronary angiography and cardiac MRI takes 15 seconds ± 2 for manual marker definition (28). In total, postprocessing time for the preparation of 3D fusion images can be estimated to be approximately 56 minutes ± 2 per case.
Results
Results from conventional 2D readout are listed in Table 1. Inconsistent findings are summarized in Table 2 and Figure 3.
Table 1:
Results from Conventional 2D Readout
Table 2:
Consistency of Results and Added Value
Figure 3:
Graph shows consistency of results from conventional two-dimensional (2D) readout, 2D readout including CT-derived fractional flow reserve (CT FFR), and three-dimensional (3D) image fusion. As is usually done in daily clinical routine, multiple imaging modalities were separately analyzed in conventional 2D readout. Areas of hypoperfusion and/or myocardial scar were attributed to one of the main coronary arteries based on their standard supply territories. Using this approach, we found controversial, imprecise, incomplete, inconsistent, and/or incorrect results in 47% of cases. Most of these divergent findings could be solved when adding information from CT FFR and 3D image fusion.
Conventional 2D Readout
CT coronary angiography.—For CT coronary angiography, image quality was rated as good to excellent (mean score, 2.6 ± 0.5). Significant coronary stenoses were found in seven of 17 cases (41%). Average Agatston score was 308 ± 525.
CT-derived FFR.—CT FFR calculations were possible in all but one patient (16 of 17, 94%). In this case (patient 9), exclusion criteria for correct CT FFR analysis applied (stenosis of the left main artery could not be correctly calculated according to software specifications) (30). In one other case (patient 6), CT FFR of the first side branch of the first diagonal branch might not be fully accurate because of prior stent placement. However, the main stenosis and the major drop of CT FFR are already seen in front of the stent and not over its course. CT FFR measurements revealed pathologic values in seven of 17 cases (41%). Six of these patients corresponded to six pathologic cases also identified in conventional CT coronary angiography readout. The remaining case (patient 7) did not show significant findings in CT coronary angiography, but a pathologic CT FFR value (0.68) was found in the distal right coronary artery (RCA).
Cardiac MRI perfusion.—For cardiac MRI perfusion, image quality was rated as good to excellent (mean score, 2.6 ± 0.6). Eight of 17 patients (47%) revealed a pathologic extent of hypoperfused myocardium (ie, ≥5% of the LV volume). Average ischemic burden of pathologic cases was 17% ± 5.
Cardiac MRI LGE.—For cardiac MRI LGE, image quality was rated as good to excellent (mean score, 2.5 ± 0.7). Three of 17 patients (18%) showed myocardial scar of varying extent. All these patients coincided with cases showing pathologic hypoperfusion in cardiac MRI perfusion. Average scar burden of pathologic cases was 7% ± 4.
Coronary angiography.—Coronary angiography reported coronary stenoses in eight of 17 patients (47%). In four of these cases (four of 17, 24%), invasive FFR measurements were performed for lesions of intermediate severity. In two cases, findings from CT FFR and 3D image fusion were retrospectively confirmed by reviewing coronary angiography images (patient 6: stenosis of the first diagonal branch and its first side branch; patient 7: diffuse concentric narrowing of RCA).
Added Value from 3D Image Fusion
Multimodal multiparametric 3D image fusion was possible in all cases (17 of 17, 100%). Analysis from the two readers (J.v.S., R.M.) matched in all but one case (16 of 17, 94%). The first and second readout of one reader (R.M.) yielded identical results (17 of 17, 100%).
Nine patients (nine of 17, 53%; patients 1, 3–5, 8, and 12–15) showed no pathologic findings in any imaging test. Some of these patients showed small hypointensities of nonpathologic extent (patients 1, 3, 5, 8, and 12–14), two of whom revealed a speckled pattern of hypoenhancement not clearly related to any specific vessel (patients 3 and 12). Four of these cases showed hypoenhancement correlated to distal coronary arteries with low but not yet pathologic CT FFR (0.76, 0.8, 0.8, and 0.75 for patients 5, 8, 13, and 14, respectively). In the remaining case (patient 1), one reader (J.v.S.) described a speckled pattern of hypoenhancement, whereas the second reader (R.M.) correlated minor hypoperfusion to the left circumflex artery (LCx) supply territory. Both readers reported identical results in all patients described hereafter.
Patient 2 had severe three-vessel disease with pathologic findings in all imaging modalities (Figs E1, E2 [supplement]). At CT coronary angiography, significant stenoses of the left anterior descending artery (LAD), LCx, and RCA were described. In addition, CT FFR showed low values in the ramus intermedius and obtuse marginal branch. In 3D image fusion, all coronary lesions could be intuitively correlated to perfusion deficits and scars of varying extent.
In patient 6 (Fig 4), CT coronary angiography, CT FFR, and coronary angiography reported an RCA stenosis, which was correlated to an inferior perfusion deficit in cardiac MRI perfusion. Cardiac MRI perfusion revealed a second perfusion deficit of the anterior/anterolateral wall attributed to the LAD/LCx supply territory. Three-dimensional image fusion helped in refining the diagnosis: The anterior/anterolateral perfusion deficit was mainly caused by the first diagonal branch and its first, stented side branch, both also showing pathologic CT FFR values.
Figure 4a:
Images of a 65-year-old man (patient 6). (a) Cardiac MRI perfusion shows perfusion deficit of anterior/anterolateral wall attributed to left anterior descending artery/left circumflex artery (*). (b) CT coronary angiography. (c) Coronary angiography, left anterior oblique projection with caudal angulation. (d) Three-dimensional image fusion helped refine diagnosis: perfusion deficits (*) were most likely caused by narrow first diagonal branch and its first, stented side branch (arrowhead). Retrospectively, denoted lesion could also be found at CT coronary angiography and coronary angiography (arrowheads in b and c, respectively). CT FFR = CT-derived fractional flow reserve, LGE = late gadolinium enhancement.
Figure 4b:
Images of a 65-year-old man (patient 6). (a) Cardiac MRI perfusion shows perfusion deficit of anterior/anterolateral wall attributed to left anterior descending artery/left circumflex artery (*). (b) CT coronary angiography. (c) Coronary angiography, left anterior oblique projection with caudal angulation. (d) Three-dimensional image fusion helped refine diagnosis: perfusion deficits (*) were most likely caused by narrow first diagonal branch and its first, stented side branch (arrowhead). Retrospectively, denoted lesion could also be found at CT coronary angiography and coronary angiography (arrowheads in b and c, respectively). CT FFR = CT-derived fractional flow reserve, LGE = late gadolinium enhancement.
Figure 4c:
Images of a 65-year-old man (patient 6). (a) Cardiac MRI perfusion shows perfusion deficit of anterior/anterolateral wall attributed to left anterior descending artery/left circumflex artery (*). (b) CT coronary angiography. (c) Coronary angiography, left anterior oblique projection with caudal angulation. (d) Three-dimensional image fusion helped refine diagnosis: perfusion deficits (*) were most likely caused by narrow first diagonal branch and its first, stented side branch (arrowhead). Retrospectively, denoted lesion could also be found at CT coronary angiography and coronary angiography (arrowheads in b and c, respectively). CT FFR = CT-derived fractional flow reserve, LGE = late gadolinium enhancement.
Figure 4d:
Images of a 65-year-old man (patient 6). (a) Cardiac MRI perfusion shows perfusion deficit of anterior/anterolateral wall attributed to left anterior descending artery/left circumflex artery (*). (b) CT coronary angiography. (c) Coronary angiography, left anterior oblique projection with caudal angulation. (d) Three-dimensional image fusion helped refine diagnosis: perfusion deficits (*) were most likely caused by narrow first diagonal branch and its first, stented side branch (arrowhead). Retrospectively, denoted lesion could also be found at CT coronary angiography and coronary angiography (arrowheads in b and c, respectively). CT FFR = CT-derived fractional flow reserve, LGE = late gadolinium enhancement.
For patient 7 (Fig 5), CT coronary angiography and coronary angiography reported stenosis-free coronary arteries, whereas cardiac MRI perfusion revealed a large area of hypoperfusion of the inferior/inferoseptal wall. Adding information from CT FFR and 3D fusion showed that the perfusion deficit was most likely caused by an RCA with diffuse concentric narrowing, resulting in pathologic CT FFR of 0.68. Image data of this patient were acquired 6 years after heart transplantation. The diffuse lumen restriction could possibly be explained by cardiac allograft vasculopathy, a disease that many heart transplant recipients develop in the long term and that is characterized by fibrotic proliferation and intimal thickening of the coronary arteries (31). It is known that such diffuse lumen restriction can cause flow alterations similar to those of a focal stenosis of higher degree (32).
Figure 5a:
Images of a 58-year-old man (patient 7). Three-dimensional (3D) image fusion (a) alongside CT coronary angiography (b: curved multiplanar reformation of right coronary artery [RCA]). Conventional two-dimensional readout without CT-derived fractional flow reserve (CT FFR) was inconsistent: In cardiac MRI perfusion, inferior/inferoseptal hypoperfusion was found without significant lesion of RCA in CT coronary angiography or coronary angiography. In this case, CT FFR in combination with 3D image fusion helped: inferior/inferoseptal perfusion deficit (* in a) was most likely caused by diffuse concentric narrowing of RCA (a and b, arrowheads) with pathologic CT FFR. This patient underwent a heart transplant, and diffuse vessel narrowing might have been caused by cardiac allograft vasculopathy. As illustrated in (c), diffuse lumen narrowing can lead to flow restrictions similar to those of focal stenosis of higher degree. LGE = late gadolinium enhancement.
Figure 5b:
Images of a 58-year-old man (patient 7). Three-dimensional (3D) image fusion (a) alongside CT coronary angiography (b: curved multiplanar reformation of right coronary artery [RCA]). Conventional two-dimensional readout without CT-derived fractional flow reserve (CT FFR) was inconsistent: In cardiac MRI perfusion, inferior/inferoseptal hypoperfusion was found without significant lesion of RCA in CT coronary angiography or coronary angiography. In this case, CT FFR in combination with 3D image fusion helped: inferior/inferoseptal perfusion deficit (* in a) was most likely caused by diffuse concentric narrowing of RCA (a and b, arrowheads) with pathologic CT FFR. This patient underwent a heart transplant, and diffuse vessel narrowing might have been caused by cardiac allograft vasculopathy. As illustrated in (c), diffuse lumen narrowing can lead to flow restrictions similar to those of focal stenosis of higher degree. LGE = late gadolinium enhancement.
Figure 5c:
Images of a 58-year-old man (patient 7). Three-dimensional (3D) image fusion (a) alongside CT coronary angiography (b: curved multiplanar reformation of right coronary artery [RCA]). Conventional two-dimensional readout without CT-derived fractional flow reserve (CT FFR) was inconsistent: In cardiac MRI perfusion, inferior/inferoseptal hypoperfusion was found without significant lesion of RCA in CT coronary angiography or coronary angiography. In this case, CT FFR in combination with 3D image fusion helped: inferior/inferoseptal perfusion deficit (* in a) was most likely caused by diffuse concentric narrowing of RCA (a and b, arrowheads) with pathologic CT FFR. This patient underwent a heart transplant, and diffuse vessel narrowing might have been caused by cardiac allograft vasculopathy. As illustrated in (c), diffuse lumen narrowing can lead to flow restrictions similar to those of focal stenosis of higher degree. LGE = late gadolinium enhancement.
For patient 9, a left main artery stenosis was described in CT coronary angiography and coronary angiography but could not be assessed by CT FFR (see above). In cardiac MRI perfusion, areas of myocardial hypoperfusion were reported for the supply territories of all three coronary arteries. While for RCA hypoperfusion no culprit lesion was found at CT coronary angiography or coronary angiography, 3D image fusion revealed that the perfusion deficit was limited to the outermost RCA territory and correlated to borderline CT FFR (0.76).
For patient 10 (Fig 6), CT coronary angiography showed a subtotal occlusion of the proximal LAD and total occlusions of the proximal LCx and distal LAD. Three-dimensional fusion showed hypoperfusion and preserved viability (scar transmurality ≤50%) around the proximal LAD and LCx, but a large area of transmural scar perfectly correlated to the occluded distal LAD.
Figure 6a:
Images of a 56-year-old man (patient 10). Three-dimensional image fusion (a) and conventional two-dimensional images (b: curved multiplanar CT coronary angiography reformation of left anterior descending artery [LAD], c: cardiac MRI late gadolinium enhancement [LGE]). Because of subtotal occlusion of proximal LAD (arrowheads in a and b corresponding to each other), surrounding hypoperfusion can be seen, while myocardial viability is preserved in proximal supply territory (scar transmurality ≤50%). Because of total occlusion of distal LAD, large area of transmural scar has evolved in distal supply territory (* in a and c). Because of proximal stenosis of left circumflex artery (LCx), segmentation algorithm failed to follow course of this vessel (a, dotted line used to visualize presumed course). However, it can be seen that restricted LCx flow is causing myocardial hypoperfusion (a, lateral heart wall) but has not yet caused myocardial scarring of larger extent. CT FFR = CT-derived fractional flow reserve.
Figure 6b:
Images of a 56-year-old man (patient 10). Three-dimensional image fusion (a) and conventional two-dimensional images (b: curved multiplanar CT coronary angiography reformation of left anterior descending artery [LAD], c: cardiac MRI late gadolinium enhancement [LGE]). Because of subtotal occlusion of proximal LAD (arrowheads in a and b corresponding to each other), surrounding hypoperfusion can be seen, while myocardial viability is preserved in proximal supply territory (scar transmurality ≤50%). Because of total occlusion of distal LAD, large area of transmural scar has evolved in distal supply territory (* in a and c). Because of proximal stenosis of left circumflex artery (LCx), segmentation algorithm failed to follow course of this vessel (a, dotted line used to visualize presumed course). However, it can be seen that restricted LCx flow is causing myocardial hypoperfusion (a, lateral heart wall) but has not yet caused myocardial scarring of larger extent. CT FFR = CT-derived fractional flow reserve.
Figure 6c:
Images of a 56-year-old man (patient 10). Three-dimensional image fusion (a) and conventional two-dimensional images (b: curved multiplanar CT coronary angiography reformation of left anterior descending artery [LAD], c: cardiac MRI late gadolinium enhancement [LGE]). Because of subtotal occlusion of proximal LAD (arrowheads in a and b corresponding to each other), surrounding hypoperfusion can be seen, while myocardial viability is preserved in proximal supply territory (scar transmurality ≤50%). Because of total occlusion of distal LAD, large area of transmural scar has evolved in distal supply territory (* in a and c). Because of proximal stenosis of left circumflex artery (LCx), segmentation algorithm failed to follow course of this vessel (a, dotted line used to visualize presumed course). However, it can be seen that restricted LCx flow is causing myocardial hypoperfusion (a, lateral heart wall) but has not yet caused myocardial scarring of larger extent. CT FFR = CT-derived fractional flow reserve.
For the same patient, CT FFR also revealed relevant lesions of smaller vessels (ie, the first diagonal branch and right posterior-lateral branch). Three-dimensional fusion showed hypoperfused myocardium adjacent to the right posterior-lateral branch, which was not seen in 2D readout of cardiac MRI perfusion.
For patient 11, CT coronary angiography depicted significant stenoses in RCA and LCx. CT FFR found pathologic flow in RCA and the obtuse marginal branch. While coronary angiography confirmed the relevance of the latter two lesions (both ≥99%), the LCx stenosis seen at CT coronary angiography was of intermediate severity (invasive FFR = 0.86). Cardiac MRI perfusion depicted perfusion deficits in RCA and LCx supply territories. All findings could be intuitively summarized by using 3D image fusion: Perfusion deficits as well as myocardial scar were most likely correlated to hemodynamically relevant RCA and obtuse marginal branch lesions.
In patient 16, CT coronary angiography showed significant lesions of LAD and ramus intermedius. CT FFR found pathologic values in LAD, ramus intermedius, and additionally in RCA (relevance of all three lesions confirmed by invasive FFR). Cardiac MRI perfusion depicted hypoperfusion in the supply territories of all three main coronary arteries—with anterolateral hypoperfusion in the LCx territory assumed to be false positive. Three-dimensional fusion showed that the anterolateral perfusion deficit was in close proximity to the ramus intermedius, whereas the LCx territory was almost unaffected.
Patient 17 revealed a significant LAD stenosis at CT coronary angiography and coronary angiography with pathologic CT FFR values and hypoperfusion in cardiac MRI perfusion. In cardiac MRI perfusion readout, a second area of hypoperfusion was found in the inferior wall. Three-dimensional fusion showed that this inferior hypoperfusion was located around the right posterior-lateral branch, which also showed pathologic CT FFR of 0.73.
Discussion
In this proof-of-principle study, we explained a method for multimodal multiparametric 3D image fusion combining all diagnostic facets of CAD.
Comparing results from conventional 2D readout and 3D image fusion, we found that consistency of results could be substantially increased. In clinical routine, multiple imaging modalities assessing CAD are usually analyzed separately from each other. Areas of myocardial hypoperfusion and scar are typically attributed to one coronary artery based on standard supply territories (26). Applying this approach, we found uncertain (ie, controversial, imprecise, incomplete, inconsistent, and/or incorrect) results in eight of 17 (47%) cases. Most of these cases could be clarified by including information from CT FFR (six of 17, 35% uncertain) and 3D image fusion (two of 17, 12% uncertain).
As pointed out in the introduction, to our knowledge, all current approaches toward multimodal CAD imaging depict a subset of pathologic aspects only (13,15–21,28). Most of these approaches combine anatomic information from CT coronary angiography with myocardial perfusion data from nuclear imaging (15,16), cardiac MRI perfusion (20,21), or CT perfusion (18,19). In our study, we found that additional inclusion of hemodynamic information from CT FFR to fused 3D images can be superior to anatomic information from CT coronary angiography alone—a finding also backed by the literature (4,5). Further including information from cardiac MRI LGE can help to intuitively delineate myocardial hypoperfusion and scar (28), which allows for identification of hypoperfused but viable myocardium, this being the tissue of interest for interventional or surgical revascularization.
CT FFR calculates FFR from CT coronary angiography in a noninvasive fashion and has attracted growing interest (3–6,24,25,33). In one patient of our study cohort (patient 9 with left main artery stenosis), CT FFR was impossible because of inherent software restrictions. In one other case (patient 6), CT FFR might have been imprecise because of prior intervention. Further exclusion criteria of the CT FFR tool include coronary artery bypass grafts, stenosis at a coronary ostium, and aneurysms (30). All these conditions are not uncommon in cardiologic patients. These cases underline the importance of further improving CT FFR algorithms if they should become part of the daily clinical routine.
Inter- and intrareader reliability of our method was found to be high for both the preparation and the interpretation of 3D fusion images. Regarding preparation, user-dependent variation is limited to manual processing of cardiac MRI data, calculation of CT FFR, and semiautomatic co-registration. While reproducibility has been proven for cardiac MRI processing (27,28) and CT FFR (25,33), the applied optimization algorithm will automatically correct for minor variations of user input when setting three distinct landmarks as a starting point for co-registration. Regarding interpretation of the readily prepared 3D images, inter- and intrareader reliability was assessed by repeated analysis by two readers. While analysis of both readers matched in all but one patient (different interpretation of subpathologic hypoperfusion in patient 1), the repeated readout of one reader (R.M.) yielded identical results.
Our multimodal imaging approach is based on a two-stage diagnostic procedure including both CT imaging and cardiac MRI. While such an approach is feasible for diagnostics of stable CAD (12), the number of patients undergoing both imaging tests in clinical routine is low (34). In a separately analyzed subcohort of the European Cardiovascular Magnetic Resonance Registry, 6% of stress cardiac MRI results were inconclusive and necessitated additional testing (leading to CT coronary angiography in 13% of these cases) (35). However, studies investigating the benefit of multimodal imaging for CAD reported added diagnostic value in a higher number of cases (36). Hybrid imaging of CAD was found to significantly improve specificity (14)—a result that might be important when considering the finding that 59% of patients with one imaging test with positive results do not show obstructive CAD in invasive coronary angiography and do not need revascularization (37).
Admittedly, postprocessing of all data sets and preparation of 3D fusion images described in this study was complex and time-consuming. However, it is important to differentiate between the overhead necessary at the current stage of prototypic development and the estimated extra effort that will be necessary for obtaining the same results with a future software tool specifically engineered for this purpose (56 minutes ± 2). Future advances in (semi-)automatic FFR computation and cardiac MRI segmentation might further decrease the manual overhead. Eventually, the additional expenses (in terms of time and complexity) need to be carefully weighed against possible advantages of a more accurate diagnosis.
In summary, a two-stage diagnostic procedure including CT and cardiac MRI yields diagnostic advantages but is also associated with higher costs and complexity. Therefore, such diagnostic work-up should be reserved for complex pathologic cases with uncertain findings in the first test (12). Complicated cases may particularly benefit from the insights gained by 3D image fusion, as physiologically relevant lesions can be correlated to resulting perfusion deficits and myocardial scar with a high degree of certainty. In this way, a two-step diagnostic approach combined with CT FFR and 3D image fusion might serve as a gatekeeper for coronary angiography, coronary interventions, and surgery.
This study had limitations. First, the number of enrolled patients was relatively low. Moreover, only a limited number of patients had areas of hypoperfusion and myocardial scars. However, the purpose of this pilot study was to develop a technique for multimodal image fusion and show its applicability. It can be assumed that the presented results hold true for a larger patient cohort. Second, a variety of software tools was applied for image processing. This prototypic architecture is too complex for transferring the technique into daily clinical routine. A dedicated software tool specifically implemented for the task at hand should be developed in the future.
In conclusion, the present study proposed a method for diagnostic assessment of CAD including multimodal multiparametric imaging data from both cardiac CT and cardiac MRI. The strength of this approach is the ability to incorporate all the information from both imaging modalities to obtain a comprehensive understanding of the coronary anatomy, myocardial perfusion, and scar burden. This finding holds true especially when results from coronary angiography and perfusion imaging are inconsistent or even contradictory.
APPENDIX
SUPPLEMENTAL FIGURES
Acknowledgments
We acknowledge the help of Max Schoebinger, PhD and Christian Hopfgartner, MSc (both of Siemens Healthineers, Forchheim, Germany) and thank them for valuable discussions on photorealistic rendering of medical data.
Supported in part by the Swiss National Science Foundation (CR3213_132671/1) and Bayer Healthcare.
Disclosures of Conflicts of Interest: J.v.S. disclosed no relevant relationships. M.M. disclosed no relevant relationships. H.M. disclosed no relevant relationships. C.S. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: is employed by and holds stock/stock options in Siemens Healthcare. Other relationships: has patents (pending and issued) with Siemens Healthcare. S.K. disclosed no relevant relationships. F.R. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: is a board member of European Society of Cardiology; is a consultant for Amgen, Fresenius, CITI Research, and Vifor; has grants/grants pending with Abbott/St. Jude Medical, Amgen, Bayer, Novartis, Servier, and Swiss National Foundation; received payment for lectures including service on speakers bureaus from Abbott, AstraZeneca, Boehringer Ingelheim, Hôpitaux Universitaires des Genève/GECORE, Luzerner Kantonsspital, CCE Services (Boston Scientific), Medtronic, Medscape, Novartis, Roche, Ruwag, Sanofi-Aventis, Servier, and Swiss Heart Failure Academy (USZ). Other relationships: disclosed no relevant relationships. H.A. disclosed no relevant relationships. R.M. disclosed no relevant relationships.
Abbreviations:
- CAD
- coronary artery disease
- CT FFR
- CT-derived FFR
- FFR
- fractional flow reserve
- LAD
- left anterior descending artery
- LCx
- left circumflex artery
- LGE
- late gadolinium enhancement
- LV
- left ventricle
- RCA
- right coronary artery
- 3D
- three-dimensional
- 2D
- two-dimensional
References
- 1.Patel MR, Calhoon JH, Dehmer GJ, et al. ACC/AATS/AHA/ASE/ASNC/SCAI/SCCT/STS 2017 appropriate use criteria for coronary revascularization in patients with stable ischemic heart disease: a report of the American College of Cardiology appropriate use criteria task force, American Association for Thoracic Surgery, American Heart Association, American Society of Echocardiography, American Society of Nuclear Cardiology, Society for Cardiovascular Angiography and Interventions, Society of Cardiovascular Computed Tomography, and Society of Thoracic Surgeons. [Published correction appears in J Am Coll Cardiol 2018;71(19):2279–2280]. J Am Coll Cardiol 2017;69(17):2212–2241. [DOI] [PubMed] [Google Scholar]
- 2.Hausleiter J, Meyer T, Hadamitzky M, et al. Non-invasive coronary computed tomographic angiography for patients with suspected coronary artery disease: the Coronary Angiography by Computed Tomography with the Use of a Submillimeter resolution (CACTUS) trial. Eur Heart J 2007;28(24):3034–3041. [DOI] [PubMed] [Google Scholar]
- 3.Itu L, Rapaka S, Passerini T, et al. A machine-learning approach for computation of fractional flow reserve from coronary computed tomography. J Appl Physiol (1985) 2016;121(1):42–52. [DOI] [PubMed] [Google Scholar]
- 4.Tesche C, De Cecco CN, Albrecht MH, et al. Coronary CT angiography-derived fractional flow reserve. Radiology 2017;285(1):17–33. [DOI] [PubMed] [Google Scholar]
- 5.Ihdayhid AR, Norgaard BL, Gaur S, et al. Prognostic value and risk continuum of noninvasive fractional flow reserve derived from coronary CT angiography. Radiology 2019;292(2):343–351. [DOI] [PubMed] [Google Scholar]
- 6.Kishi S, Giannopoulos AA, Tang A, et al. Fractional flow reserve estimated at coronary CT angiography in intermediate lesions: comparison of diagnostic accuracy of different methods to determine coronary flow distribution. Radiology 2018;287(1):76–84. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Takx RA, Blomberg BA, El Aidi H, et al. Diagnostic accuracy of stress myocardial perfusion imaging compared to invasive coronary angiography with fractional flow reserve meta-analysis. Circ Cardiovasc Imaging 2015;8(1):e002666. [DOI] [PubMed] [Google Scholar]
- 8.Demirkiran A, Everaars H, Amier RP, et al. Cardiovascular magnetic resonance techniques for tissue characterization after acute myocardial injury. Eur Heart J Cardiovasc Imaging 2019;20(7):723–734. [DOI] [PubMed] [Google Scholar]
- 9.Vitanis V, Manka R, Giese D, et al. High resolution three-dimensional cardiac perfusion imaging using compartment-based k-t principal component analysis. Magn Reson Med 2011;65(2):575–587. [DOI] [PubMed] [Google Scholar]
- 10.Manka R, Wissmann L, Gebker R, et al. Multicenter evaluation of dynamic three-dimensional magnetic resonance myocardial perfusion imaging for the detection of coronary artery disease defined by fractional flow reserve. Circ Cardiovasc Imaging 2015;8(5):e003061. [DOI] [PubMed] [Google Scholar]
- 11.Goetti R, Kozerke S, Donati OF, et al. Acute, subacute, and chronic myocardial infarction: quantitative comparison of 2D and 3D late gadolinium enhancement MR imaging. Radiology 2011;259(3):704–711. [DOI] [PubMed] [Google Scholar]
- 12.Task Force Members , Montalescot G, Sechtem U, et al. 2013 ESC guidelines on the management of stable coronary artery disease: the task force on the management of stable coronary artery disease of the European Society of Cardiology. Eur Heart J 2013;34(38):2949–3003. [DOI] [PubMed] [Google Scholar]
- 13.Veulemans V, Hellhammer K, Polzin A, Bönner F, Zeus T, Kelm M. Current and future aspects of multimodal and fusion imaging in structural and coronary heart disease. Clin Res Cardiol 2018;107(Suppl 2):49–54. [DOI] [PubMed] [Google Scholar]
- 14.Rizvi A, Han D, Danad I, et al. Diagnostic performance of hybrid cardiac imaging methods for assessment of obstructive coronary artery disease compared with stand-alone coronary computed tomography angiography: a meta-analysis. JACC Cardiovasc Imaging 2018;11(4):589–599. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Gaemperli O, Schepis T, Kalff V, et al. Validation of a new cardiac image fusion software for three-dimensional integration of myocardial perfusion SPECT and stand-alone 64-slice CT angiography. Eur J Nucl Med Mol Imaging 2007;34(7):1097–1106. [DOI] [PubMed] [Google Scholar]
- 16.Pazhenkottil AP, Benz DC, Gräni C, et al. Hybrid SPECT perfusion imaging and coronary CT angiography: long-term prognostic value for cardiovascular outcomes. Radiology 2018;288(3):694–702. [DOI] [PubMed] [Google Scholar]
- 17.Nazir MS, Ismail TF, Reyes E, Chiribiri A, Kaufmann PA, Plein S. Hybrid positron emission tomography-magnetic resonance of the heart: current state of the art and future applications. Eur Heart J Cardiovasc Imaging 2018;19(9):962–974. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.von Spiczak J, Manka R, Mannil M, et al. 3D fusion of coronary CT angiography and CT myocardial perfusion imaging: intuitive assessment of morphology and function. J Cardiovasc Comput Tomogr 2017;11(6):437–443. [DOI] [PubMed] [Google Scholar]
- 19.Chen MY, Rochitte CE, Arbab-Zadeh A, et al. Prognostic value of combined CT angiography and myocardial perfusion imaging versus invasive coronary angiography and nuclear stress perfusion imaging in the prediction of major adverse cardiovascular events: The CORE320 multicenter study. Radiology 2017;284(1):55–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Donati OF, Alkadhi H, Scheffel H, et al. 3D fusion of functional cardiac magnetic resonance imaging and computed tomography coronary angiography: accuracy and added clinical value. Invest Radiol 2011;46(5):331–340. [DOI] [PubMed] [Google Scholar]
- 21.von Spiczak J, Manka R, Gotschy A, et al. Fusion of CT coronary angiography and whole-heart dynamic 3D cardiac MR perfusion: building a framework for comprehensive cardiac imaging. Int J Cardiovasc Imaging 2018;34(4):649–660. [DOI] [PubMed] [Google Scholar]
- 22.Leipsic J, Abbara S, Achenbach S, et al. SCCT guidelines for the interpretation and reporting of coronary CT angiography: a report of the Society of Cardiovascular Computed Tomography Guidelines Committee. J Cardiovasc Comput Tomogr 2014;8(5):342–358. [DOI] [PubMed] [Google Scholar]
- 23.Agatston AS, Janowitz WR, Hildner FJ, Zusmer NR, Viamonte M Jr, Detrano R. Quantification of coronary artery calcium using ultrafast computed tomography. J Am Coll Cardiol 1990;15(4):827–832. [DOI] [PubMed] [Google Scholar]
- 24.Rabbat MG, Berman DS, Kern M, et al. Interpreting results of coronary computed tomography angiography-derived fractional flow reserve in clinical practice. J Cardiovasc Comput Tomogr 2017;11(5):383–388. [DOI] [PubMed] [Google Scholar]
- 25.De Geer J, Sandstedt M, Björkholm A, et al. Software-based on-site estimation of fractional flow reserve using standard coronary CT angiography data. Acta Radiol 2016;57(10):1186–1192. [DOI] [PubMed] [Google Scholar]
- 26.Cerqueira MD, Weissman NJ, Dilsizian V, et al. Standardized myocardial segmentation and nomenclature for tomographic imaging of the heart. A statement for healthcare professionals from the Cardiac Imaging Committee of the Council on Clinical Cardiology of the American Heart Association. Circulation 2002;105(4):539–542. [DOI] [PubMed] [Google Scholar]
- 27.Manka R, Jahnke C, Kozerke S, et al. Dynamic 3-dimensional stress cardiac magnetic resonance perfusion imaging: detection of coronary artery disease and volumetry of myocardial hypoenhancement before and after coronary stenting. J Am Coll Cardiol 2011;57(4):437–444. [DOI] [PubMed] [Google Scholar]
- 28.von Spiczak J, Mannil M, Kozerke S, Alkadhi H, Manka R. 3D image fusion of whole-heart dynamic cardiac MR perfusion and late gadolinium enhancement: intuitive delineation of myocardial hypoperfusion and scar. J Magn Reson Imaging 2018;48(4):1129–1138. [DOI] [PubMed] [Google Scholar]
- 29.Kancharla K, Weissman G, Elagha AA, et al. Scar quantification by cardiovascular magnetic resonance as an independent predictor of long-term survival in patients with ischemic heart failure treated by coronary artery bypass graft surgery. J Cardiovasc Magn Reson 2016;18(1):45. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Schwemmer C. Syngo.via frontier cFFR prototype manual v3.1. Erlangen, Germany: Siemens Healthineers, 2018. [Google Scholar]
- 31.Chih S, Chong AY, Mielniczuk LM, Bhatt DL, Beanlands RS. Allograft vasculopathy: the Achilles’ heel of heart transplantation. J Am Coll Cardiol 2016;68(1):80–91. [DOI] [PubMed] [Google Scholar]
- 32.Gould KL. Does coronary flow trump coronary anatomy? JACC Cardiovasc Imaging 2009;2(8):1009–1023. [DOI] [PubMed] [Google Scholar]
- 33.Kumamaru KK, Angel E, Sommer KN, et al. Inter- and intraoperator variability in measurement of on-site CT-derived fractional flow reserve based on structural and fluid analysis: a comprehensive analysis. Radiol Cardiothorac Imaging 2019;1(3):e180012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Bruder O, Schneider S, Nothnagel D, et al. EuroCMR (European Cardiovascular Magnetic Resonance) registry: results of the German pilot phase. J Am Coll Cardiol 2009;54(15):1457–1466. [DOI] [PubMed] [Google Scholar]
- 35.Moschetti K, Muzzarelli S, Pinget C, et al. Cost evaluation of cardiovascular magnetic resonance versus coronary angiography for the diagnostic work-up of coronary artery disease: application of the European Cardiovascular Magnetic Resonance registry data to the German, United Kingdom, Swiss, and United States health care systems. J Cardiovasc Magn Reson 2012;14(1):35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.van Werkhoven JM, Heijenbrok MW, Schuijf JD, et al. Combined non-invasive anatomical and functional assessment with MSCT and MRI for the detection of significant coronary artery disease in patients with an intermediate pre-test likelihood. Heart 2010;96(6):425–431. [DOI] [PubMed] [Google Scholar]
- 37.Patel MR, Peterson ED, Dai D, et al. Low diagnostic yield of elective coronary angiography. N Engl J Med 2010;362(10):886–895. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.