Fractional flow reserve calculation optimized by myocardial computed tomography perfusion information

Introduction

The coronary fractional flow reserve (FFR), measured using pressure wire technology, plays a key role in clinically assessing myocardial ischemia and allows for precise diagnosis of ischemic heart conditions (1,2). However, its invasiveness and high cost have restricted its widespread clinical use. With the advancement of computed tomography (CT) imaging equipment, featuring enhanced detector resolution, broader detector width, and accelerated scanning speeds, more distinct image data can be captured for dynamic areas such as the heart. The emerging non-invasive FFR calculation, based on computational fluid dynamics (CFD) and medical image segmentation, has shown great potential. This novel approach is called computed tomography angiography-derived fractional flow reserve (CT-FFR).

In recent years, there have been remarkable achievements in simulating coronary blood flow. Given CFD’s high reliance on the accuracy of boundary conditions and the limitations of CT images in providing direct information about flow-related boundary conditions, researchers have been striving to achieve more accurate simulations of real blood flow states through enhanced modeling and computational techniques. Hiromasa Otake et al. articulated three fundamental principles for CFD-based FFR simulation: total coronary flow correlates with myocardial mass, intravascular blood flow is related to vessel size, and microvascular resistance is associated with the level of maximum hyperemia (3). These studies form the basis for CT-FFR technology. They allow for estimating total coronary flow from myocardial mass, determining blood flow distribution across different branches, and simulating FFR under hyperemic conditions. Research by Coenen et al. demonstrated clinical concordance between CFD-simulated CT-FFR and FFR measured by pressure wire (4). However, Rajiah et al. has highlighted the limitations of FFR, such as its inability to address myocardial ischemia resulting from microcirculatory dysfunction, which can be mitigated by myocardial perfusion (5).

Myocardial computed tomography perfusion (CTP) image analysis can provide insights into myocardial blood flow (MBF) and distribution of perfusion. Experimental results from van Assen et al. indicate that the correlation between myocardial perfusion and major adverse cardiac events is less pronounced compared to coronary angiography and FFR (6). Nous et al. substantiated the correlation between MBF and FFR outcomes, suggesting that their combined use holds clinical significance for diagnostic purposes (7). However, the studies mentioned did not delve into a more profound integration of these two methodologies. Schrauwen et al. introduced a technique for estimating outflow boundary conditions using CTP images and proceeded to calculate the FFR and wall shear stress at coronary artery bifurcations (8). Yet, this method has not been tested for its validity across an entire coronary arterial tree.

CTP images can reveal details about myocardial ischemia. However, due to technical challenges, the evaluation of myocardial ischemia using CTP images is not precise enough, particularly for mild ischemia, where sensitivity tends to be lower. CT-FFR, based on the coronary vascular pressure, determines whether the myocardial region supplied by a particular coronary artery is ischemic, but it does not take into account the condition of the myocardium. Integrating CTP image information into the calculation of CT-FFR can provide CT-FFR with information from both the coronary and myocardial dimensions, thereby enhancing the accuracy of CT-FFR and enabling a more accurate evaluation of potential and existing myocardial ischemia in clinical settings.

This study introduces a method that integrates blood flow data from CTP with CFD boundary conditions to enhance the precision of branch outlet and aortic inlet boundaries, ultimately improving FFR calculations. In this article, we first elaborated on the method of using CFD for coronary vessel model simulation, image preprocessing, mesh preprocessing, inlet flow estimation method based on the myocardial mass, coronary branch flow allocation based on Murray’s law (9), CFD solving process, and corresponding result analysis. Subsequently, we introduced MBF calculations based on CTP images, registrations of CTP and computed tomography angiography (CTA) images, and optimizations of boundary conditions based on MBF distributions. Moreover, we proposed two new methods for calculating flow distributions in coronary arteries based on CTP images as boundary conditions for CFD simulations. Finally, the results of the three methods were compared, analyzed, and discussed. We present this article in accordance with the STARD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-24-2172/rc).

Methods

Patient population

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by Ethics Committee of Peking Union Medical College Hospital (No. I-22PJ012) and individual consent for this retrospective analysis was waived. Sixty consecutive symptomatic patients suspected of having coronary artery disease (CAD) were enrolled between March 2016 and June 2018. These patients were initially scheduled for coronary computed tomography angiography (CCTA), CTP, and subsequently underwent invasive coronary angiography (ICA) from Peking Union Medical College Hospital (Figure 1).

Figure 1 Flow chart of patient inclusion. CAD, coronary artery disease; CCTA, coronary computed tomography angiography; CT, computed tomography; CTP, computed tomography perfusion; FFR, fractional flow reserve; ICA, invasive coronary angiography; MBF, myocardial blood flow.

Inclusion criteria: (I) patients diagnosed with obstructive CAD or myocardial ischemia (the moderate to high-risk category); (II) high-quality CCTA and CTP images.

Exclusion criteria: (I) history of stent implantation; (II) history of coronary artery bypass grafting; (III) unclear CCTA and CTP images; (IV) missing FFR records.

Cardiac CTP, CCTA and ICA image acquisition

The patients were all advised to abstain from consuming caffeine for 24 hours before the examination. Using a dual-source CT (second-generation SOMATOM Definition Flash or third-generation SOMATOM Force; Siemens Healthineers, Forchheim, Germany) with attenuation-based tube current modulation and automated tube voltage selection, all enrolled patients underwent a combined adenosine triphosphate (ATP)-stress dynamic myocardial CTP and CCTA imaging examination protocol. The procedures for obtaining and preprocessing CTA and CTP images are thoroughly outlined in the Appendix 1.

ICA was performed using an Allura Xper UNIQ FD10 system (Philips Medical Systems, The Netherlands), with images acquired through multiple projections, and at least two orthogonal projections were obtained to evaluate the target vessels. FFR was measured in the lesions using a 0.014-in pressure wire with sensor tips (Pressure Wire; Radi Medical Systems) during rest and during maximal myocardial congestion induced by intravenous infusion of ATP (140 µg/kg/min). An FFR value of ≤0.8 was considered to be hemodynamically significant.

Steps for calculating FFR with CFD

In this manuscript, we start by explaining the traditional CFD simulation process based on CT imaging, and then discuss how to integrate CTP data, as shown in Figure 2 (detailed calculation steps can be found in Figures S3,S4). The traditional CFD simulation computational process based on CT images encompasses the following steps: acquisition and segmentation of CTA images; development of the mesh model; formulation of boundary conditions; and performance of CFD simulation. With the incorporation of CTP information, the CFD simulation computational process is refined to include: the acquisition and preprocessing of both CTP and CTA images; the registration of CTA and CTP images; the design of boundary conditions; and the execution of CFD simulation. The core of the article emphasizes the registration of CTP and CTA images and the strategic design of diverse boundary conditions.

Figure 2 Schematic diagram of the overall plan. CCTA, coronary computed tomography angiography; CFD, computed fluid dynamics; CTP, computed tomography perfusion; MBF, myocardial blood flow.

Setting of boundary conditions in traditional CFD simulations

The boundary conditions are used to determine the motion state of the fluid at the boundary. In this paper, the no-slip condition of the wall is used, and the vascular boundary is considered a rigid wall for this steady-state CFD simulation. The physical parameters of blood were set as the density 1.05 g/cm³ and viscosity 3.45 mPa⋅s. As detailed in the following text:

Inlet boundary conditions

The boundary conditions of the inlet mainly include the inlet pressure and inlet flow rate. The inlet pressure is estimated based on the patient’s systolic and diastolic blood pressure records. Approximate calculations can significantly reduce the computational workload. The pressure at the inlet can be estimated using the following formula (10):

Pinlet=2∗Pdiastole+Psystole3[1]

P_inlet refers to the inlet pressure, P_diastole refers to the patient’s diastolic pressure, and P_systole refers to the systolic and diastolic pressures.

The estimation of the coronary inlet flow rate is one of the main difficulties in the CFD simulation of coronary arteries. Coronary inlet flows account for approximately 4% of the cardiac output (CO) (11). By calculating the CO, the coronary inlet flow can be estimated. Referencing the research of Choy and Sharma et al. (10,12), the CO in this article can be calculated by the following formula:

QCO=kρVLVM[2]

In this equation, Q_CO refers to the CO, V_LVM refers to the volume of the left myocardium, and ρ refers to the myocardial density, which is 1.2 g/cm³ in this article. Moreover, k is 1.5, representing the conversion coefficient between the left ventricular myocardium and the overall heart. Then the inlet flow of the total coronary arteries is set to 4% of the CO (11).

Outlet boundary conditions

The outlet blood flow distribution is based on the principle of Murray’s Law (9), which states that the flow rate of each branch is correlated with its pipe diameter, and the flow rate of each outlet is allocated. Although some studies have suggested that multi-scale coronary artery models based on lumped parameter resistance models offer higher accuracy (13), these papers typically employ a smaller number of datasets and require manual, fine segmentation of the coronary artery models, and these operation need much time (14). In clinical practice, it is impossible to obtain precise modeling and very accurate boundary condition parameters, and the models need to serve a large number of diverse clinical cases and achieve a relatively high overall accuracy rate. This requires the models to have high stability of most patients and rapidly calculation in a short time, which outweighs the need for precise calculations of some special cases. Currently, there are some the FFR calculation models that can be applied clinically by doctors rapidly are all based on Murray’s Law (15). However, the specific values of the relationship between the flow rate and pipe diameter are not consistent in various studies. In this paper, a proportional relationship between the flow rate and pipe diameter to the third power is used to determine the flow distribution between branches (10).

Estimation of the boundary conditions using CTP images

Image registration

The MBF images from CTP provide total blood flow to the left ventricular myocardium (Figure 3A,3B). To determine MBF at each myocardial voxel on cardiac CTA images, we use registration techniques to align CTP with CTA images. The MBF voxels are then mapped onto the CTA images. The registration operation consists of two stages. First, a rigid registration method based on mutual information (16) is used to align the initial position of the heart. Second, a non-rigid registration method is used to finely adjust the myocardial part. Regarding non-rigid registration, the SyN method that is based on diffeomorphism is used (17), and local cross correlation is used as the similarity measure. The registered images are shown in Figure 3C.

Figure 3 Workflow of get artery flow from CTP images. (A) The principle of MBF calculation from CTP images; (B) the MBF of myocardium; (C) the fusion display of registered MBF and CTA images; (D) the display of MBF and coronary artery tree by 3 dimensions (3D) rendering. AIF, arterial input function; CTA, computed tomography angiography; CTP, computed tomography perfusion; HU, Hounsfield unit; MBF, myocardial blood flow; TAC, tissue attention curve.

After registration, the mapping relation between coronary arteries obtained from CTA images and the MBF of myocardium obtained from CTP images is determined and can be as the base of more calculation. Figure 3D shows the relation between coronary arteries and the myocardium.

Method for optimizing the boundary conditions using CTP images

In this article, we investigated two methods for estimating the coronary flow as the boundary conditions using myocardial perfusion information: (I) inlet flow optimization: the total coronary inlet flow is estimated by MBF from CTP images, and the outlet flow is calculated according to the method described in Section “Outlet boundary conditions”; and (II) inlet & outlet flow optimization: the flow rate of each branch outlet is estimated based on the MBF value in the myocardial area covered by the coronary artery branch, and the inlet flow rate can then be calculated by the sum of the outlet flows.

Method for optimizing coronary inlet flow using CTP images. In this section, CTP images are used to calculate the MBF distribution of the left ventricular myocardium and sum all the blood flow of the left ventricular myocardium to obtain the blood supply of the left ventricular myocardium. From this result, it can be estimated that the total coronary flow is 1.5 times the left ventricular myocardial flow as mentioned in Section “Inlet boundary conditions”.

qcoro=1.5∗MBFall_MBF_sum[3]

This method can more directly obtain the total flow information of the coronary artery. It is worth noting that before CTP image scanning, adenosine is needed to increase coronary blood flow, while it is not required before CTA image scanning. Therefore, it is necessary to convert the coronary blood flow calculated based on CTP images to the rest state by the hyperemia coefficient according to adenosine dose before the CFD boundary condition inlet flow calculation (18).

Method of optimizing the inlet and outlet blood flow of the coronary artery branches using CTP images. This method is guided by the MBF and the blood supply area analysis of every coronary artery branch to optimize the CFD outlet boundary conditions. After registration, the correspondence between the CTP image and the CTA image is obtained. Based on this correspondence, the MBF value can be matched to the CTA image, and each voxel of the left ventricular myocardium in the CTA image can be used to obtain the corresponding MBF value.

Afterward, the blood supply area of each coronary branch is calculated in the myocardium. First, structural analysis is conducted on the coronary artery tree, which is divided into independent coronary segments according to the branching structures. Then, using the region growth algorithm and starting from each coronary artery segment, growth at different rates is carried out to gradually cover all myocardial regions. The growth rate is determined by the artery diameter. It is based on the hypothesis that the thick coronary artery supplies a larger area than the thin artery. The growing areas of different coronary segments are labeled differently as shown in Figure 4, and they can be used to determine the blood supply area for each coronary segment. Finally, the sum of the MBF values for each area is calculated as the basis for traffic weight allocation between different coronary branches. The blood flow distribution of the target vessel is proportional to its weight of the total flow.

Figure 4 Schematic diagram of the blood supply area analysis.

After successful blood supply area analysis of the arteries, the corresponding flow is adjusted based on the weighted average ratio of coverage volume and MBF. In practical, an empirical equation is used to get better performance as follow:

q1q2=(d1d2)m(MBF1MBF2)n[4]

In this equation, q₁ and q₂ refer to the flows of two branches, MBF₁ and MBF₂ refer to the diameters the branches, and refer to the sum of MBF values in the corresponding blood supply area, and m and n refer to empirical parameters. In this article, m is set to 3.0, and n is set to 0.5 to get better performance.

Some of the coronary artery branches do not supply blood to the left ventricular myocardium, and the flow of these artery branches are still distributed according to the Murry’s law as described in section “Outlet boundary conditions”. Due to the small proportion of left ventricular myocardial blood supply from the right coronary arteries (RCAs), the significant variability of the RCAs among individuals, and most of the clinical FFR measurements are performed on the left coronary artery, we did not apply the optimization for the outlet boundary conditions using MBF to the RCAs. After calculating the outlet flow of coronary artery branches, the inlet flow can then be calculated by the sum of all the outlet flows.

Mesh generation and CFD solution

After the segmentation of the coronary arteries, the coronary arteries and the aorta were meshed using the Delaunay method. We employed a multi-scale meshing approach. Since the aorta is not the primary object of simulation, we used a coarser mesh with a maximum element size of 2.0 mm to save computational resources. For the coronary arteries, a finer mesh was applied, with a maximum element size of 0.3 mm. This is because the resolution of the CT images is at least 0.3 mm, and using a smaller mesh size would not provide greater accuracy but instead could introduce artifacts due to the voxel size of the image, affecting the simulation. The CFD calculations were performed using the open-source software OpenFOAM. The CFD solution method of this algorithm is implemented on the SIMPLE (semi-implicit equation of pressure coupled equation) algorithm, which is based on the finite volume method commonly used in CFD. When iterating in the flow field, the velocities on adjacent grids do not affect each other, and the pressure correction is limited to only local areas, resulting in less computational complexities and faster calculation speeds. During the iterative process, the computation was considered to have converged and was terminated when the difference in pressure values between two consecutive iterations was less than 10⁻⁵ (Appendix 1).

Statistical analysis

Statistical analysis was conducted using the Python programming software (version 3.8.5; https://www.python.org/). Continuous variables were presented as mean ± standard deviation or median (interquartile range). To verify the accuracy of the results, we compared the FFR values obtained by the CFD-based method with the FFR values obtained by the pressure guide wire measurement, using the latter values as the gold standard. We used the mean absolute error to evaluate the difference between the calculated value and the gold standard value. FFR values greater than 0.8 were considered negative, and FFR values less than or equal to 0.8 were considered positive. We calculated statistical parameters such as the accuracy, sensitivity and specificity to determine the difference in the positives and negatives between the calculated CT-FFR values and the gold standard values. Youden’s index was used for the cutoff value calculation. P<0.05 was considered statistically significant (Appendix 1).

Results

Data processing

We collected images from 60 cases with both cardiac CTA and CTP images for this study. Among them, 7 cases were excluded due to image blurring, severe artifacts, or the presence of other severe faults described above, which affected the discrimination of the coronary arteries and prevented the accurate segmentation of the coronary arteries. three cases underwent stent implantation, and three cases were excluded due to unclear recordings of the FFR measurements, such as not recording measurement branches. In the end, a total of 47 cases (31 male and 16 female; age, 60.66±8.07 years) that included 100 pressure measuring positions with pressure guide wires were used. Among them, there are 65 negative measurement values and 35 positive measurement values (Table 1).

Inlet flow and outlet flow ratio calculation

The coronary inlet flows of the 47 cases were calculated. The mean coronary inlet flow calculated by myocardium mass is 175.7 mL/min. The mean value of the MBF sum of the left myocardium is 185.46 mL/min. The mean inlet flow calculated by MBF is 155.8 mL/min, less than the value by mass. The comparison of every case is shown in Figure 5A.

Figure 5 Comparison of coronary inlet flows by myocardium mass and MBF, and outlet flow ratios by Murry’s law and MBF in blood supply area. (A) Compare coronary inlet flow by Mass and MBF; (B) compare outlet flow ratio of LADs by Murry’s law and MBF in blood supply area; (C) compare outlet flow ratio of LCXs by Murry’s law and MBF in blood supply area; (D) compare outlet flow ratio of RIs by Murry’s law and MBF in blood supply area. LADs, left anterior descending branch; LCXs, left circumflex branch; RIs, ramus intermedius branch.

The flows of three primary branches of the left main branch were compared instead, which named left anterior descending branch (LADs), left circumflex branch (LCXs) and ramus intermedius branch (RIs). The RCAs were not analyzed as mentioned in section “Method for optimizing the boundary conditions using CTP images”. Just some of the patients (17 among 47) have RIs. There is no FFR pressure measurements on RIs, but the flows of RIs could affect the flow ratios of LADs and LCXs and the flow of RIs are also be analyzed. The flows of LADs, LCXs and RIs included all the flows of the downstream branches, and the sum of the flow ratio of the three primary branches is equal to the left main branch, which is set to 1.0. The mean flow ratios of LADs, LCXs and RIs are 0.487, 0.431 and 0.082 by Murry’s law, while those are 0.529, 0.420 and 0.050 by MBF in blood supply area. It shows that the LADs branch supplied more blood flow ratio than their size, and it may explain why ischemia occurs on LAD more frequently. The flow ratio of LADs, LCXs and RIs of every case are shown in Figure 5B-5D. Some of the ratio of RIs are 0, because some of the cases do not have RI.

Optimizing the FFR results by the new boundary conditions

Table 2 shows that the method of optimizing the coronary inlet and outlet flow based on CTP has a significant improvement in the performance of the CFD-based FFR calculation algorithms. The diagnostic performance at the patient level can be seen in Table S1. Among them, the inlet flow optimization method based on CTP improves the accuracy by 5% and specificity by 7.6% compared to traditional method. Furthermore, the inlet & outlet optimization method based on CTP and blood supply area analysis further improved the accuracy by 1% to 94% while increasing the sensitivity to 100%. However, at the same time, the specificity decreased. This indicates that the MBF value calculated based on CTP images reflects the actual heart blood flow conditions of patients better when used to estimate the coronary inlet flow compared to traditional myocardial mass estimation. The specific case analyses for the computational results under different boundary conditions are detailed in the Appendix 1.

Discussion

Feasibility of obtaining MBF through CTP for flow distribution analysis

Obtaining the total coronary artery blood supply through CTP information is intuitive. Notably, the MBF information itself also has certain errors, which may come from the fitting error of the tissue attention curve (TAC) and arterial input function (AIF) curves or from the error of the device side scanning. However, in situations where it is difficult to obtain blood flow information through other noninvasive methods (such as CT-FFR), MBF information still has certain guiding significance for CT cardiac-related hemodynamic simulation.

According to the experimental results, optimizing the calculation of CO information through MBF information showed a significant improvement (from 88% to 93%) to calculate FFR. From this result, it can be inferred that the CO calculated by this method is more reasonable than that estimated by the myocardial mass for the following reasons:

• The distribution of blood flow within the myocardial region is not uniform, and simple estimations based on factors such as the volume and mass are not sufficiently accurate. For patients with myocardial ischemia or local myocardial necrosis, the MBF value better reflects local MBF and thus is more useful for reference.
• Although we determined the range of relevant hyper parameters for calculating the CO through myocardial volume and mass through research and made multiple attempts, it is still uncertain whether the formulas and hyper parameters used can reflect the impact of different myocardial masses on the CO. In contrast, the MBF information can provide a more intuitive reference.
• The significant increase in specificity (from 86.2% to 93.8%) indicates that the total CO calculated by the MBF information can reduce the proportion of false positives in some data, thereby achieving an overall accurate improvement.

The impact of registration

The method of optimizing outlet flow through MBF distribution relies on the myocardial registration algorithm. When the difference between CTP and CTA myocardial imaging is too significant, there may be an inability to register. When the registration deviation is large, it may reduce the accuracy of the blood supply area analysis, which may have a certain impact on the optimization method proposed in this article.

Impact of the blood supply area analysis on the results

As mentioned in Section “Method for optimizing the boundary conditions using CTP images” regarding the blood supply area analysis, the blood supply area of each branch of the coronary artery needs to first be calculate. An example of the myocardial blood supply area is shown in Figure 4.

By conducting blood supply area analysis to optimize the exit boundary conditions, the myocardium is divided into blood supply regions corresponding to the coronary segmentation. This optimization method has the following characteristics:

• Calculating the blood supply area instead of artery diameter to estimate the blood flow distribution of artery branches reduces the impact caused by the difficulty in accurately segmenting the end of coronary arteries. The growth of regions can reflect the actual influence of blood vessels on the blood supply area of the myocardium to some extent.
• The main blood vessels can occupy the dominant position in flow distribution because in the blood supply area analysis, the main blood vessels typically affect larger myocardial regions.
• The notable rise in sensitivity is due to MBF information being more reliable than myocardial volume for estimating CO. The attempt to divide the blood supply area has also led to a deeper integration of the MBF information in CFD calculations, thereby providing more reasonable hydrodynamic analysis results.
• We confirmed the guiding role of MBF information in this study. However, due to the small dataset (the gold standard of 100 cases at the vascular level), the overall plan is only in the first step verification stage, and it cannot be confirmed whether the blood supply area analysis is similar to the objective blood supply logic of coronary vessels. However, the significant improvement on this dataset also preliminarily proves that implementing the blood supply area analysis may be more meaningful than using Murray’s law alone for boundary condition setting.

Limitations

• A larger amount of data is needed, and the validation of large datasets may lead to more universal patterns. The reason for the small data in this study is detailed in the Appendix 1.
• The accuracy of registration needs to be improved, as CTP and CTA scans typically use different scanning strategies. A more automated high-precision registration tool can make the optimization method proposed in this article smoother and improve the efficiency of clinical analysis processes.
• More accurate coronary image and segmentation quality are needed, and more accurate blood supply area analysis methods need to be developed. The specific reasons can be found in the Appendix 1.
• Previous study has shown that Nitroglycerin administration before CCTA can enhance the diagnostic performance of CT-FFR (19). Future optimization of the scanning process, combined with the findings of this study, will be explored to determine its potential to further enhance the diagnostic performance of CT-FFR.

Conclusions

The results of this article reflect the correction effect of the CTP information on CFD calculation. The use of CTP-specific MBF as an optimization condition for coronary hemodynamic calculations can provide certain reference values, including the total blood flow and various outlet boundary conditions. This method can improve the accuracy of CFD simulation and mitigate the impact of simulation uncertainty caused by patients’ personalized information.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STARD reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-24-2172/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-24-2172/dss

Funding: This work was supported by the Beijing Natural Science Foundation (grant No. Z210013, 2021), National Natural Science Foundation of China (NSFC) National Science Fund for Distinguished Young Scholars (No. 82325026), Chinese Academy of Medical Sciences Innovation Fund for Medical Sciences (CIFMS) (No. 2023-I2M-C&T-A-004), the National High Level Hospital Clinical Research Funding (Nos. 2022-PUMCH-B-027 and 2022-PUMCH-A-097), and the Non-profit Central Research Institute Fund of Chinese Academy of Medical Sciences (No. 2024-RC320-03).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-24-2172/coif). Y.W. reports that this study receives funding from Beijing Natural Science Foundation (grant No. Z210013, 2021), National Natural Science Foundation of China (NSFC) National Science Fund for Distinguished Young Scholars (No. 82325026), Chinese Academy of Medical Sciences Innovation Fund for Medical Sciences (CIFMS) (No. 2023-I2M-C&T-A-004), the Non-profit Central Research Institute Fund of Chinese Academy of Medical Sciences (No. 2024-RC320-03), and the National High Level Hospital Clinical Research Funding (Nos. 2022-PUMCH-B-027 and 2022-PUMCH-A-097). P.Q., F.J., J.G., and X.W. were an employee of Shanghai United Imaging Healthcare Co., Ltd. Y.Y., C.X., L.Z., M.W., Y.W. were employees of Department of Radiology, Peking Union Medical College Hospital. Y.Z. and Z.W. were employees of Beijing United Imaging Research Institute of Intelligent Imaging. P.D. was an employee of Beijing United Imaging Intelligence. D.W. was an employee of Shanghai United Imaging Intelligence. The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by Ethics Committee of Peking Union Medical College Hospital (No. I-22PJ012) and individual consent for this retrospective analysis was waived.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Danad I, Szymonifka J, Twisk JWR, Norgaard BL, Zarins CK, Knaapen P, Min JK. Diagnostic performance of cardiac imaging methods to diagnose ischaemia-causing coronary artery disease when directly compared with fractional flow reserve as a reference standard: a meta-analysis. Eur Heart J 2017;38:991-8. [Crossref] [PubMed]
2. Xu R, Li C, Qian J, Ge J. Computed Tomography-Derived Fractional Flow Reserve in the Detection of Lesion-Specific Ischemia: An Integrated Analysis of 3 Pivotal Trials. Medicine (Baltimore) 2015;94:e1963. [Crossref] [PubMed]
3. Otake H, Taylor CA, Matsuo H, Tanaka N, Akasaka T. Noninvasive Fractional Flow Reserve Derived From Coronary Computed Tomography Angiography - Is This Just Another New Diagnostic Test or the Long-Awaited Game Changer? Circ J 2017;81:1085-93. [Crossref] [PubMed]
4. Coenen A, Lubbers MM, Kurata A, Kono A, Dedic A, Chelu RG, Dijkshoorn ML, Gijsen FJ, Ouhlous M, van Geuns RJ, Nieman K. Fractional flow reserve computed from noninvasive CT angiography data: diagnostic performance of an on-site clinician-operated computational fluid dynamics algorithm. Radiology 2015;274:674-83. [Crossref] [PubMed]
5. Rajiah P, Cummings KW, Williamson E, Young PM. CT Fractional Flow Reserve: A Practical Guide to Application, Interpretation, and Problem Solving. Radiographics 2022;42:340-58. [Crossref] [PubMed]
6. van Assen M, De Cecco CN, Eid M, von Knebel Doeberitz P, Scarabello M, Lavra F, Bauer MJ, Mastrodicasa D, Duguay TM, Zaki B, Lo GG, Choe YH, Wang Y, Sahbaee P, Tesche C, Oudkerk M, Vliegenthart R, Schoepf UJ. Prognostic value of CT myocardial perfusion imaging and CT-derived fractional flow reserve for major adverse cardiac events in patients with coronary artery disease. J Cardiovasc Comput Tomogr 2019;13:26-33. [Crossref] [PubMed]
7. Nous FMA, Geisler T, Kruk MBP, Alkadhi H, Kitagawa K, Vliegenthart R, et al. Dynamic Myocardial Perfusion CT for the Detection of Hemodynamically Significant Coronary Artery Disease. JACC Cardiovasc Imaging 2022;15:75-87. [Crossref] [PubMed]
8. Schrauwen JTC, Coenen A, Kurata A, Wentzel JJ, van der Steen AFW, Nieman K, Gijsen FJH. Functional and anatomical measures for outflow boundary conditions in atherosclerotic coronary bifurcations. J Biomech 2016;49:2127-34. [Crossref] [PubMed]
9. Murray CD. The Physiological Principle of Minimum Work: I. The Vascular System and the Cost of Blood Volume. Proc Natl Acad Sci U S A 1926;12:207-14. [Crossref] [PubMed]
10. Sharma P, Itu L, Zheng X, Kamen A, Bernhardt D, Suciu C, Comaniciu D. A framework for personalization of coronary flow computations during rest and hyperemia. Annu Int Conf IEEE Eng Med Biol Soc 2012;2012:6665-8. [Crossref] [PubMed]
11. Hansen JT, Koeppen BM, Netter FH. Netter’s atlas of human physiology. Elsevier Healthe; 2002.
12. Choy JS, Kassab GS. Scaling of myocardial mass to flow and morphometry of coronary arteries. J Appl Physiol (1985) 2008;104:1281-6. [Crossref] [PubMed]
13. Zhang M, Keramati H, Gharleghi R, Beier S. Reliability of characterising coronary artery flow with the flow-split outflow strategy: Comparison against the multiscale approach. Comput Methods Programs Biomed 2025;263:108669. [Crossref] [PubMed]
14. Liu J, Wang X, Li B, Huang S, Sun H, Zhang L, Sun Y, Liu Z, Liu J, Wang L, Zhao X, Wang W, Zhang M, Liu Y. Non-Invasive Quantification of Fraction Flow Reserve Based on Steady-State Geometric Multiscale Models. Front Physiol 2022;13:881826. [Crossref] [PubMed]
15. Cortés C, Liu L, Berdin SL, Fernández-Corredoira PM, Zhang R, Schäfer U, López M, Diarte JA, Tu S, Gutiérrez-Chico JL. Agreement between Murray law-based quantitative flow ratio (μQFR) and three-dimensional quantitative flow ratio (3D-QFR) in non-selected angiographic stenosis: A multicenter study. Cardiol J 2022;29:388-95. [Crossref] [PubMed]
16. Maes F, Loeckx D, Vandermeulen D, Suetens P. Image registration using mutual information. In: Paragios N, Duncan J, Ayache N, (eds). Handbook of Biomedical Imaging. Springer, Boston, MA; 2025.
17. Avants BB, Epstein CL, Grossman M, Gee JC. Symmetric diffeomorphic image registration with cross-correlation: evaluating automated labeling of elderly and neurodegenerative brain. Med Image Anal 2008;12:26-41. [Crossref] [PubMed]
18. Wilson RF, Wyche K, Christensen BV, Zimmer S, Laxson DD. Effects of adenosine on human coronary arterial circulation. Circulation 1990;82:1595-606. [Crossref] [PubMed]
19. Zhou Y, Zhao N, An Y, Ma W, Han L, Song L, Yang W, Gao Y, Lu B. Impact of nitroglycerin on machine-learning fractional flow reserve in coronary computed tomography (CT)-angiography. Quant Imaging Med Surg 2024;14:6856-68. [Crossref] [PubMed]