## JACC: Cardiovascular Interventions

# Fractional Flow Reserve and Coronary Bifurcation AnatomyA Novel Quantitative Model to Assess and Report the Stenosis Severity of Bifurcation Lesions

## Author + information

- Received September 8, 2014.
- Revision received December 11, 2014.
- Accepted December 31, 2014.
- Published online April 20, 2015.

## Author Information

- Shengxian Tu, PhD
^{â},^{â }^{â}(sxtu{at}sjtu.edu.cn), - Mauro Echavarria-Pinto, MD
^{â¡}, - Clemens von Birgelen, MD, PhD
^{Â§},^{â}, - Niels R. Holm, MD
^{Â¶}, - Stylianos A. Pyxaras, MD
^{#}, - Indulis Kumsars, MD
^{ââ}, - Ming Kai Lam, MD
^{Â§}, - Ilona Valkenburg, BSc
^{Â§}, - Gabor G. Toth, MD
^{#}, - Yingguang Li, MSc
^{â }, - Javier Escaned, MD, PhD
^{â¡}, - William Wijns, MD, PhD
^{#}and - Johan H.C. Reiber, PhD
^{â }

^{â}Biomedical Instrument Institute, School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China^{â }Division of Image Processing, Department of Radiology, Leiden University Medical Center, Leiden, the Netherlands^{â¡}Cardiovascular Institute, Hospital Clinico San Carlos, Madrid, Spain^{Â§}Department of Cardiology, Thoraxcentrum Twente, Medisch Spectrum Twente, Enschede, the Netherlands^{â}Department of Health Technology and Services Research, MIRA Institute, University of Twente, Enschede, the Netherlands^{Â¶}Department of Cardiology, Aarhus University Hospital, Skejby, Aarhus, Denmark^{#}Cardiovascular Centre Aalst, Onze-Lieve-Vrouwziekenhuis Clinic, Aalst, Belgium^{ââ}Department of Cardiology, Paul Stradins Clinical Hospital, Riga, Latvia

- âµâ
**Reprint requests and correspondence:**

Dr. Shengxian Tu, Med-X Research Institute, Shanghai Jiao Tong University, No. 1954, Hua Shan Road, Room 123, Shanghai 200030, China.

## Abstract

**Objectives** The aim of this study was to develop a new model for assessment of stenosis severity in a bifurcation lesion including its core. The diagnostic performance of this model, powered by 3-dimensional quantitative coronary angiography to predict the functional significance of obstructive bifurcation stenoses, was evaluated using fractional flow reserve (FFR) as the reference standard.

**Background** Development of advanced quantitative models might help to establish a relationship between bifurcation anatomy and FFR.

**Methods** Patients who had undergone coronary angiography and interventions in 5 European cardiology centers were randomly selected and analyzed. Different bifurcation fractal laws, including Murray, Finet, and HK laws, were implemented in the bifurcation model, resulting in different degrees of stenosis severity.

**Results** A total of 78 bifurcation lesions in 73 patients were analyzed. In 51 (65%) bifurcations, FFR was measured in the main vessel. A total of 34 (43.6%) interrogated vessels had an FFR ≤0.80. Correlation between FFR and diameter stenosis was poor by conventional straight analysis (ρ = −0.23, p < 0.001) but significantly improved by bifurcation analyses: the highest by the HK law (ρ = −0.50, p < 0.001), followed by the Finet law (ρ = −0.49, p < 0.001), and the Murray law (ρ = −0.41, p < 0.001). The area under the receiver-operating characteristics curve for predicting FFR ≤0.80 was significantly higher by bifurcation analysis compared with straight analysis: 0.72 (95% confidence interval: 0.61 to 0.82) versus 0.60 (95% confidence interval: 0.49 to 0.71; p = 0.001). Applying a threshold of ≥50% diameter stenosis, as assessed by the bifurcation model, to predict FFR ≤0.80 resulted in 23 true positives, 27 true negatives, 17 false positives, and 11 false negatives.

**Conclusions** The new bifurcation model provides a comprehensive assessment of bifurcation anatomy. Compared with straight analysis, identification of lesions with preserved FFR values in obstructive bifurcation stenoses was improved. Nevertheless, accuracy was limited by using solely anatomical parameters.

Assessing the severity of coronary bifurcation lesions is a frequently faced challenge. The pitfalls and limitations of the conventional quantitative coronary angiography (QCA) approach—the straight (single segment) analysis—have been widely recognized (1,2). Three-dimensional (3D) reconstruction on the basis of 2 angiographic projections might improve assessment of bifurcation lesions (3). However, the greatest challenge of bifurcation analysis remains to define the true reference vessel size of the main vessel (MV) and the side branch (SB), and, in particular, the true size of the bifurcation core, or the polygon of confluence (i.e., where the MV and the SB merge upstream into a single branch). Several bifurcation models have been developed to incorporate the step-down phenomenon in calculating the reference vessel size for quantitative assessment of bifurcation lesions (2,4,5). However, these models have not completely addressed the physical interpretation of the reference vessel size at the bifurcation core, which hampers the comprehensive assessment of the entire bifurcation lesion that extends from the proximal MV into the distal MV and/or the SB. Development of dedicated quantitative bifurcation models could assist in: 1) understanding the relation between bifurcation anatomy and fractional flow reserve (FFR), a standard of reference for inducible myocardial ischemia (6); and 2) assessing interventional devices aiming at this complex clinical scenario.

This study aims to present a new quantitative bifurcation model for comprehensive assessment of anatomical severity in the entire bifurcation lesion including its core. The model constructs a number of bent oval planes for measurement in the bifurcation core. This resolves the ambiguity in defining stenosis severity in the core and seamlessly integrates lesion assessment at the 2 sides of the lateral wall opposite to the carina. The accuracy of this model, empowered by 3D QCA to predict the functional significance of coronary bifurcation stenosis, was evaluated using FFR as the reference standard.

## Methods

### Study population

This retrospective, observational, and analytical study randomly selected patients who were admitted in 5 European hospitals (Hospital Clinico San Carlos, Madrid, Spain; Thoraxcentrum Twente at Medisch Spectrum Twente, Enschede, the Netherlands; Aarhus University Hospital, Skejby, Denmark; OLV Clinic, Aalst, Belgium; and Paul Stradins Clinical Hospital, Riga, Latvia) in the context of clinical FFR studies including bifurcation lesions (TWENTE trial [7], Nordic-Baltic Bifurcation Study III [8], Tryton kissing balloon investigation [9], and Echavarria-Pinto et al. [10]). Inclusion criteria were: 1) bifurcation lesions with >40% diameter stenosis (DS%) by visual estimation in the main coronary arteries; 2) absence of coronary artery bypass graft to the interrogated vessel; 3) measurement of FFR in the MV prior to revascularization or in the SB if no stent was implanted in the SB; and 4) 2 angiographic projections ≥25° apart recorded by flat-panel systems.

Exclusion criteria were: 1) excessive overlap or foreshortening (>90%); 2) poor angiographic images quality for delineation of lumen contours; and 3) presence of a third branch with diameter >1 mm at the interrogated bifurcation, forming a trifurcation structure.

Diagnostic angiography was obtained after intracoronary nitrates in all cases, and FFR interrogation was performed as part of the clinical evaluation of patients or according to the corresponding study protocol. Written informed consent was obtained from all patients according to institutional regulations.

### The new quantitative bifurcation model

A bifurcation is associated with 3 segments: the proximal MV, the distal MV, and the SB. They are connected by the bifurcation core, where the distal MV and the SB merge upstream into the proximal MV (2,11). Existing quantitative models (2,4,5) assess the disease along several flat planes at the bifurcation core. Figure 1 presents the new quantitative bifurcation model. Instead of measuring diameter and stenosis along flat planes, the new model creates a number of bent oval planes, with the degree of bending depending on the bifurcation angle of each individual bifurcation. The bending of the oval plane linearly decreases toward proximal, changing into a flat plane at the beginning of the bifurcation core. The sum of the 2 paired diameters at each oval plane was defined as the core diameter. Accordingly, the reference core diameter was defined as the sum of the 2 paired reference diameters at each oval plane. The reference diameters/cords were constructed as followed: a virtual plane going through the carina and the center of the first cross section of the bifurcation core was created (blue line in Figure 1C). The virtual plane separated the core into 2 half cores, 1 extending into the distal MV and the other into the SB. The sizes of the cords in each half core changed linearly from proximal toward distal. The stenosis function is calculated from the arterial core diameter function (solid yellow curve in Figures 1D and 1E) and the reference core diameter function (dashed red curve in Figures 1D and 1E), and represents the DS% at each location along the interrogated vessel. This stenosis function is divided into 2 separated functions at the carina: 1 in the distal MV and the other in the SB. A possible jump depending on the relative stenosis severity between the distal MV and the SB can be observed. In Figure 1, the SB had higher DS% compared with the distal MV, resulting in an increased DS% when crossing the carina into the SB and decreased DS% when crossing the carina into the distal MV.

### 3D QCA and analysis protocol

Angiographic images were recorded by monoplane or biplane x-ray systems (AXIOM-Artis, Siemens, Malvern, Pennsylvania; AlluraXper, Philips Healthcare, Best, the Netherlands; INTEGRIS Allura, Philips Healthcare; Innova 3100, GE Healthcare, Chalfont, Buckinghamshire, United Kingdom). 3D QCA was performed by an experienced analyst (S.T.) using a validated software package (QAngio XA 3D research edition 1.2, Medis Specials BV, Leiden, the Netherlands) (3). Automatic calibration was used when the geometrical parameters required by 3D reconstruction were stored in the DICOM files. Otherwise, catheter calibration was applied. The reconstruction consisted of the following steps: 1) selection of 2 angiographic image sequences ≥25° apart; 2) selection of proper contrast-filled end-diastolic frames; 3) identification of anatomical landmarks for automated correction of system distortions; 4) delineation of the bifurcation lumen in the 2 projections; and 5) 3D reconstruction of the lumen and the reference surface (i.e., the estimated healthy lumen as if there was no stenosis). Three commonly used fractal laws—the Murray, Finet, and HK laws (12)—were implemented in the software to support optimization of reference diameter function (RDF) for the diffusely diseased segment if applicable. Details of the fractal laws are available in the Online Appendix.

Analyses were performed following this protocol: when FFR was measured at both the distal MV and the SB, only the segment with lower FFR was included for comparison. The 3D QCA software automatically calculated an RDF for each of the 3 segments (proximal MV, distal MV, and SB) separately and constructed a reference bifurcation core. After documenting the resulting DS% (annotated as DS%Separate), the analyst chose 2 segments with less diffused disease and optimized the RDF for the third segment using the Murray, Finet, and HK laws, respectively. Accordingly, 3 new reference bifurcation cores were reconstructed, and the corresponding DS% were annotated as DS%Murray, DS%Finet, and DS%HK, respectively. Conventional straight analysis was also performed, and the generated DS% (annotated as DS%Straight) was used for comparisons with the bifurcation analyses.

All imaging data were analyzed at a core laboratory (ClinFact, Leiden, the Netherlands), where 3D QCA performed with QAngio XA 3D on repeated analysis of the same images has intraobserver and interobserver variability of 0.02 ± 0.08 mm and 0.03 ± 0.11 mm for minimum lumen diameter, and 0.02 ± 0.21 mm^{2} and 0.04 ± 0.40 mm^{2} for minimum lumen area (MLA), respectively.

### Statistics

Continuous variables are expressed as mean ± SD or as median (interquartile range [IQR]) as appropriate. Categorical variables are expressed as percentages. Data were analyzed on a per-patient basis for clinical characteristics and on a per-vessel basis for the remaining calculations. Normal distribution was determined using the Shapiro-Wilk test. Association among continuous variables was determined by Spearman’s rank correlation coefficient (ρ value). Pairwise comparisons were made with the Student *t* test or Mann-Whitney *U* tests, as appropriate. No post-hoc p value corrections were performed. Because each bifurcation lesion poses specific anatomy and physiological repercussions, independence was assumed for vessel analyses. The performance of 3D QCA in predicting functionally-significant stenosis was assessed using sensitivity, specificity, positive predictive value, negative predictive value, and diagnostic accuracy, together with their 95% confidence intervals (CIs). The area under the receiver-operating characteristics curve (AUC) by receiver-operating curve (ROC) analysis was used to assess the diagnostic accuracy of 3D QCA using FFR ≤0.80 as the standard of reference. The Youden index (highest sum of sensitivity and specificity) was used as a criterion to identify the optimal diagnostic cut-off value. Comparisons between AUCs were performed with the DeLong method using MedCalc version 13.0 (MedCalc Software, Mariakerke, Belgium). Other statistical analyses were performed with IBM SPSS software version 20.0 (SPSS Inc., Chicago, Illinois). A 2-sided p value of <0.05 was considered significant.

## Results

### Baseline clinical and lesion characteristics

A total of 85 bifurcation lesions from 80 patients were included. Seven bifurcation lesions were excluded, resulting in a final analysis of 78 obstructed bifurcations from 73 patients. In 51 (65%) bifurcations, FFR was measured in the MV. The patients’ clinical characteristics are listed in Table 1. Hyperemia was induced by intracoronary and intravenous administration of vasodilators in 17 (21.8%) and 56 (78.2%) patients, respectively. Table 2 shows lesion characteristics. Overall, the study comprised stenoses of intermediate physiological severity (FFR: 0.81 ± 0.12, median: 0.83 [IQR: 0.74 to 0.89]), and categorically, abnormal FFR ≤0.80 was measured in 34 (43.6%) vessels: 33 in the MV and only 1 in the SB.

### Correlation between FFR and 3D QCA

Figure 2 shows a representative example of analysis applying the new bifurcation model. Comprehensive quantification of characteristics of the bifurcation lesion involving the core can be performed. Due to the similar degrees of stenosis severity in the ostium of the distal MV and the SB, a continuous diameter stenosis function crossing the bifurcation core was generated for the entire MV (Figure 2B2). The maximum DS% was found in the middle of the bifurcation core. Figure 3 shows an example of quantification in a diffusely-diseased proximal MV segment by optimizing its RDF applying the fractal laws. The ostium of the distal MV had a higher degree of stenosis compared with the SB, resulting in a small jump in the diameter stenosis function at the carina (asterisk in Figure 3C2).

The correlations between FFR and DS% from different analyses are shown in Figure 4. Poor correlation (ρ = −0.23, p < 0.01) was observed between FFR and the conventional DS%Straight. This correlation, however, significantly improved when applying the new bifurcation model, especially when integrating the fractal laws. Among the 4 bifurcation analyses, correlation with FFR was the highest for DS%HK (ρ = −0.50, p < 0.001), followed by DS%Finet (ρ = −0.49, p < 0.001), DS%Murray (ρ = −0.41, p < 0.001), and DS%Separate (ρ = −0.39, p < 0.001).

### Agreement among different quantitative analyses

The method for calculating RDF influences the resulting DS%, percent area stenosis, and lesion length. Table 3 shows the comparison of lesion characteristics by the different analyses. DS%Straight was higher compared with all bifurcation analyses, including DS%Separate (difference: 3.0 ± 5.6%; p < 0.001), DS%Murray (difference: 1.4 ± 6.5%; p = 0.057), DS%Finet (difference: 2.3 ± 7.2%; p = 0.006), and DS%HK (difference: 2.5 ± 7.1%; p = 0.002). When the 3 fractal laws were applied for the optimization of the RDF, a good agreement was observed between DS%HK and DS%Finet (difference: 0.3 ± 1.2%; p = 0.070). A small but statistically-significant bias toward higher DS%Murray was observed compared with DS%HK (difference: 1.1 ± 3.6%; p = 0.008) and DS%Finet (0.9 ± 3.3%; p = 0.026). Substantially shorter lesion length was observed in straight analysis compared with all bifurcation analyses, including the separate analysis (difference: 5.4 ± 11.2 mm; p < 0.001), the Murray law (difference: 8.0 ± 13.3 mm; p < 0.001), the Finet law (difference: 8.3 ± 13.6 mm; p < 0.001), and the HK law (difference: 8.3 ± 13.5 mm; p < 0.001). No significant difference was observed among the lesion lengths assessed by the 3 bifurcation analyses applying fractal laws.

### Accuracy of 3D QCA for predicting functionally-significant stenoses

Importantly, all bifurcation models were superior to the conventional DS%Straight for diagnosing FFR ≤0.80 stenoses (Table 4). DS%Straight had a lower AUC (0.56 [95% CI: 0.44 to 0.67]) compared with the bifurcation analyses including DS%Separate (0.64 [95% CI: 0.52 to 0.75]; difference, 0.08 [95% CI: 0.00 to 0.16]; p = 0.044), DS%Murray (0.69 [95% CI: 0.58 to 0.79]; difference, 0.13 [95% CI: 0.04 to 0.22]; p = 0.004), DS%Finet (0.72 [95% CI: 0.60 to 0.81]; difference, 0.16 [95% CI: 0.06 to 0.25]; p = 0.002), and DS%HK (0.72 [95% CI: 0.61 to 0.82]; difference, 0.16 [95% CI: 0.06 to 0.25]; p = 0.001). Among the bifurcation analyses optimized by the 3 fractal laws, DS%HK had the highest AUC. However, the AUC differences respecting DS%Murray and DS%Finet were small and statistically nonsignificant (0.03 [95% CI: −0.03 to 0.08]; p = 0.328, and 0.00 [95% CI: −0.01 to 0.02]; p = 0.622, respectively).

Categorization of the bifurcation stenosis (n = 78) on the basis of the common limit of ≥50% diameter stenosis resulted in 23 true positives, 27 true negatives, 17 false positives, and 11 false negatives. Bifurcation analysis resulted in reclassification of the straight analysis in 19 lesions: 16 correctly reclassified and 3 incorrectly reclassified. The best cut-off values for determining FFR ≤0.80 were DS%HK 56%, minimum lumen diameter 1.35 mm, and MLA 1.70 mm^{2}. The associated diagnostic performances are listed in Table 5. Comparable performances were observed for the 4 discrimination criteria on the basis of bifurcation analysis applying the fractal laws, among which DS%HK ≥56% had the highest overall accuracy of 68%, with sensitivity 53%, specificity 80%, positive predictive value 67%, and negative predictive value 69%.

## Discussion

This study presents a new quantitative model for comprehensive anatomical assessment of coronary bifurcation lesions including the bifurcation core. When applying this model in the quantification of obstructive bifurcation lesions, accuracy in diameter stenosis for predicting FFR ≤0.80 was higher by the bifurcation model compared with conventional straight analysis: 0.72 (95% CI: 0.61 to 0.82) versus 0.60 (95% CI: 0.49 to 0.71; p = 0.001). Applying a threshold of ≥50% diameter stenosis, assessed by the bifurcation model, to predict FFR ≤0.80 resulted in 23 true positives, 27 true negatives, 17 false positives, and 11 false negatives.

### QCA and FFR

The diagnostic accuracy of QCA in discriminating stenosis physiological severity as defined by FFR has been reported by various studies, with differing results (13–16). Characteristics of included lesions, angiographic quality, and techniques in image analysis may explain these differences. In a study using online QCA analysis, Toth et al. (13) found that roughly one-third of a large patient population showed discordance between DS% ≥50% and FFR ≤0.8. In another study that included mainly nonobstructive coronary stenoses (DS% = 34 ± 13%), Pyxaras et al. (15) found a good correlation between FFR and QCA as analyzed by a dedicated 3D QCA system. The AUC for MLA in determining FFR ≤0.8 was 0.89. However, another separate study (16) that applied 3D QCA in coronary lesions that were on average intermediate (DS% = 51 ± 14%) showed a lower accuracy (AUC 0.79). Recently, Tu et al. (14) studied a population of relatively homogeneous intermediate lesions (DS% = 46.6 ± 7.3%) and observed a modest accuracy in all anatomical parameters assessed by dedicated 3D QCA. MLA had the highest AUC at 0.73. These data suggest that QCA might not be sufficient to predict the physiological behavior of individual intermediate stenosis, where physiological assessment is most commonly required in clinical practice.

Yet, analysis techniques continue to advance, which is of particular interest in the setting of bifurcation lesions. It is well recognized that conventional straight analysis is not reliable for analyzing bifurcation lesions due to the step-down phenomenon (2). We developed a novel quantitative bifurcation model and applied it in the present study. Our findings suggest that this bifurcation model increased the diagnostic accuracy of angiography using FFR as the reference standard, compared with straight analysis. However, the accuracy was limited. This finding agrees with the recent statement by Johnson et al. (17) that coronary anatomy alone will probably not be sufficient to predict physiological behavior at the patient level. It also supports the hypothesis by Toth et al. (13) that the discordance between angiography and FFR is related to physiologic factors, including the amount of myocardium downstream to the stenosis and the functional status of the coronary microcirculation. Notably, a higher optimal cut-off value of 56% for DS% in predicting FFR ≤0.80 was observed in the present study, compared with the commonly-used threshold of 50%. This might be partly explained by a unique characteristic of bifurcation lesions. When 1 daughter branch is severely obstructed while the other is relatively spared, flow distribution at the bifurcation could be altered due to the increased resistance of the obstructed branch. Hence, hyperemic flow can be redistributed and mainly directed to the unobstructed branch, resulting in a higher FFR value in the obstructed branch compared with the same stenosis geometry in a theoretical straight vessel.

### Comparison to the existing solutions for bifurcation analysis

Two solutions have been reported for dedicated bifurcation analysis: the CAAS bifurcation application (Pie Medical, Maastricht, the Netherlands) (5) and the QAngio XA bifurcation application (Medis Medical Systems, Leiden, the Netherlands) (1,4). The latter contains 2 bifurcation models: 1 for T-shaped bifurcations and 1 for Y-shaped bifurcations. All of these solutions quantified diameters along a number of flat planes, resulting in ambiguity of the stenosis severity in the bifurcation core. Consequently, assessment of bifurcation anatomy and its underlying physiology is hampered. We developed a new bifurcation model that could overcome the aforementioned limitations by quantifying stenosis severity along a number of bent oval planes, with the bending depending on individual bifurcation angle. By applying this solution, the quantification at every location in the bifurcation core is unique, and lesions at the 2 sides of the lateral wall opposite to the carina are seamlessly integrated. In addition, the pattern of the RDF in the bifurcation core is clearer with this solution (increasing downstream from the mother diameter to the sum of the 2 daughter diameters), facilitating future standardization and comparison of bifurcation analyses. Last but not least, fractal laws were integrated in the model for optimization of the RDF. This could be particularly helpful when 1 of the 3 segments is diffusely diseased, during which the calculation of the RDF on the basis of the diseased segments alone is unreliable. This is because fractal laws express the diameter relation of the 3 segments prior to development of disease. Therefore, the integration of these laws could potentially improve the accuracy of the RDF calculation for the diffusely-diseased segments. Importantly, although the 3 fractal laws applied in this study have been tested in phantoms and healthy coronary bifurcations (12), they have not been systematically compared in clinical populations including bifurcation lesions. Indeed, our data showed that the correlations between DS% and FFR improved after applying the fractal laws. Among the 3 commonly-used fractal laws, we found that DS% optimized by the HK law had the highest accuracy in predicting FFR ≤0.80. However, the difference respecting the Murray and Finet laws was small and statistically nonsignificant, which might partially be explained by our relatively small sample size.

### Clinical implications

Determining whether a bifurcation lesion is in real need of revascularization remains challenging due to the complexity of its anatomy. Moreover, the percutaneous treatment of these lesions is associated with relatively high restenosis and complication rates (18). Integration of anatomical and physiological assessments, tailored selection of interventional devices and treatment strategies, and optimal sizing and deployment of the devices are important factors that may contribute to better patient outcomes. The current study presented in detail a new model for the quantitative assessment of bifurcation lesions. In this proof-of-concept study, we demonstrated that the proposed model could be applied to the analysis of x-ray images acquired by different angiographic systems with various protocols, requiring only 2 different angiographic projections at least 25° apart; therefore, it is readily available during diagnostic angiography. This is an important basis for future studies, especially for assessing the efficacy of stents or vascular scaffolds implanted in coronary bifurcations. We also demonstrated that applying this bifurcation model reduced the scatter in the relation between angiography-determined DS% and FFR, especially when bifurcation fractal laws were applied. However, the capability of predicting functionally-significant lesions by solely anatomical parameters remained limited. Our results further support the hypothesis that the discordance between angiography and FFR cannot be completely removed by improved anatomical analysis techniques. Various physiological factors, including the area of downstream supplied territory and microvascular resistance, may have contributed to the discordance. Integration of anatomical modeling and physiological factors in the computation of FFR including QCA-based FFR (14) and computed tomography–based FFR (19) might evolve as new tools for assessing functional coronary stenosis severity. These new techniques estimate patient-specific hyperemic flow and utilize computational fluid dynamics to solve complex interactions between flow and luminal boundary, shedding light on the understanding of coronary anatomy and physiology.

### Study limitations

The present study examined a limited sample size. We were unable to perform a meaningful ROC analysis for the cohort with FFR measurement in the side branch; therefore, the effect of lesion location (MV or SB) on the diagnostic accuracy of this novel bifurcation model requires further studies. We analyzed images from 4 studies that predominantly enrolled patients undergoing PCI. Therefore, selection bias cannot be ruled out. In the cohort with FFR measurements in the SB, only a small proportion of the study population had a functionally-significant lesion. Nevertheless, a previous study that examined consecutive patients with jailed side branches also reported that most side branch obstructions did not show functional significance (20), which might be partly explained by the fact that side branches generally supply perfusion to smaller amounts of myocardium, resulting in less pressure drop compared with the same stenosis in the main vessel. Our primary analysis showed that applying the bifurcation model in image analysis reduced the scatter in the relation between diameter stenosis and FFR. This warrants further prospective evaluation of the diagnostic value of bifurcation QCA in subsets of bifurcation lesions and in guiding optimal treatment strategies for bifurcation interventions on the basis of the integration of anatomy and physiology.

## Conclusions

A new model for the analysis of coronary bifurcation lesions is proposed. This model provides comprehensive assessment of bifurcation anatomy including its core. When compared with straight analysis, identification of lesions with preserved FFR values in obstructive bifurcation stenoses was improved. However, accuracy was limited by using solely anatomical parameters.

**WHAT'S KNOWN?** Current bifurcation QCA models have limitations in assessing stenosis severity in the bifurcation core.

**WHAT'S NEW?** A new quantitative bifurcation model for comprehensive assessment of bifurcation anatomy was developed, resulting in improved accuracy in identifying lesions with preserved FFR values. Nevertheless, accuracy was limited, even by including bifurcation fractal laws in the analysis. Discrepancy between QCA and FFR cannot completely be removed by solely anatomical analysis.

**WHAT'S NEXT?** Integration of QCA and coronary flow in computational FFR might improve assessment of bifurcation stenoses.

## Appendix

## Appendix

For an expanded Methods section, please see the online version of this article.

## Footnotes

This work was supported in part by the Natural Science Foundation of China under grant 61271155. Dr. Tu had an employment contract with Medis Medical Imaging Systems BV until June 2014; currently, Shanghai Jiao Tong University receives institutional grant support on his behalf from Medis. Drs. Echavarria-Pinto and Escaned have served as speakers in educational events organized by St. Jude Medical and Volcano Corporation. Dr. von Birgelen has served as a consultant to Abbott Vascular, Boston Scientific, and Medtronic; has received travel expenses from Biotronik and lecture fees from Merck Sharp & Dohme; and his institution has received research grants from Abbott Vascular, Biotronik, Boston Scientific, and Medtronic. Dr. Holm has received speaker fees, consultant fees, and research grants from St. Jude Medical; and has received a research grant from Medis. Dr. Kumsars has received speakers honoraria from Tryton Medical and AstraZeneca. Mr. Li is employed by Medis; and has a research appointment at the Leiden University Medical Center (LUMC). Dr. Wijns is a cofounder, shareholder, and board member of Cardio3BioSciences, Genae, and Argonauts Partners; and Cardiovascular Research Center Aalst receives institutional grant support and consultancy fees on his behalf of from St. Jude Medical, Tryton, and other device and pharmaceutical companies. Dr. Reiber is the Chief Executive Officer of Medis; and has a part-time appointment at LUMC as Professor of Medical Imaging. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.

- Abbreviations and Acronyms
- 3D
- 3-dimensional
- AUC
- area under the receiver-operating characteristics curve
- CI
- confidence interval
- DS%
- percent diameter stenosis
- FFR
- fractional flow reserve
- MLA
- minimum lumen area
- MV
- main vessel
- QCA
- quantitative coronary angiography
- RDF
- reference diameter function
- SB
- side branch

- Received September 8, 2014.
- Revision received December 11, 2014.
- Accepted December 31, 2014.

- American College of Cardiology Foundation

## References

- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- von Birgelen C.,
- Basalus M.W.,
- Tandjung K.,
- et al.

- ↵
- ↵
- Pyxaras S.A.,
- Tu S.,
- Barbato E.,
- Reiber J.H.C.,
- Wijns W.

- ↵
- Echavarria-Pinto M.,
- Escaned J.,
- Macías E.,
- et al.

- ↵
- ↵
- Toth G.,
- Hamilos M.,
- Pyxaras S.,
- et al.

- ↵
- Tu S.,
- Barbato E.,
- Kőszegi Z.,
- et al.

- ↵
- ↵
- Yong A.S.,
- Ng A.C.,
- Brieger D.,
- Lowe H.C.,
- Ng M.K.,
- Kritharides L.

- ↵
- Johnson N.P.,
- Kirkeeide R.L.,
- Gould K.L.

- ↵
- Maeng M.,
- Holm N.R.,
- Erglis A.,
- et al.

- ↵
- Koo B.K.,
- Erglis A.,
- Doh J.H.,
- et al.

- ↵
- Koo B.K.,
- Kang H.J.,
- Youn T.J.,
- et al.