Fractional flow reserve measurement using dynamic CT perfusion imaging in patients with coronary artery disease

Abstract

Purposes

The objective was to evaluate the accuracy of a novel CT dynamic angiographic imaging (CT-DAI) algorithm for rapid fractional flow reserve (FFR) measurement in patients with coronary artery disease (CAD).

Materials and Methods

This retrospective study included 14 patients (age 58.5 ± 10.6 years, 11 males) with CAD who underwent stress dynamic CT myocardial perfusion scanning with a dual-source CT scanner. The included patients had analyzable proximal and distal coronary artery segments adjacent to the stenosis in the perfusion images and had corresponding invasive catheter-based FFR measurements for that stenosis. An in-house software based on the CT-DAI algorithm was used to compute FFR using the pre- and post- lesion coronary time-enhancement curves obtained from the stress myocardial perfusion images. The CT-DAI derived FFR values were then compared to the corresponding catheter-based invasive FFR values. A coronary artery stenosis was considered functionally significant for FFR value <0.8.

Results

The CT-DAI derived FFR values were in agreement with the invasive FFR values in all 15 coronary arteries in 14 patients, resulting in 100% per-vessel and per-patient diagnostic accuracy. FFR derived using CT-DAI (M = 0.768, SD = 0.156) showed an excellent linear correlation (R = 0.910, P < .001) and statistical indifference (P= .655) with that measured using invasive catheter-based method (M = 0.796, SD = 0.149). Bland-Altman analysis showed no significant proportional bias.

Conclusion

The novel CT-DAI algorithm can reliably compute FFR across a coronary artery stenosis directly from dynamic CT myocardial perfusion images, facilitating rapid on-site hemodynamic assessment of the epicardial coronary artery stenosis in patients with CAD.

Abbreviations

CAD = coronary artery disease; CCTA = coronary CT angiography; CT-DAI = computed tomography dynamic angiographic imaging; CTO = chronic total occlusion; FFR = fractional flow reserve; LAD = left anterior descending artery; LCX = left circumflex artery; MP = myocardial perfusion; RCA = right coronary artery; ROI = region of interest; TAG = transluminal attenuation gradient

Summary

A novel CT dynamic angiographic imaging analytic algorithm allows for rapid and accurate measurement of fractional flow reserve directly from the myocardial perfusion images of patients with coronary artery disease.

Key Results

• A novel CT dynamic angiographic imaging (CT-DAI) analytic algorithm enables fractional flow reserve (FFR) measurements based on the coronary time-enhancement curves sampled from the myocardial perfusion images.

• FFR derived using CT-DAI (M = 0.768, SD = 0.156) showed an excellent linear correlation (R = 0.910, P < .001) with that measured using invasive catheter-based method (M = 0.796, SD = 0.149).

• FFR derived using CT-DAI in 15 coronary arteries demonstrated 100% accuracy in detecting functionally significant coronary artery disease compared to the invasive catheter-based assessment.

Introduction

Non-invasive functional evaluation plays an important role in the management of patients with coronary artery disease (CAD). In recent years, there is increasing clinical evidence that suggests CT can provide accurate functional assessment of CAD through fractional flow reserve (FFR) or myocardial perfusion (MP) measurement.^1–5 However, the approach of the respective functional assessments has their own limitations. While dynamic CT perfusion imaging can offer absolute myocardial blood flow quantification for a reliable differentiation between ischemic and non-ischemic myocardial tissues, the perfusion level in the myocardium may be discordant with the physiological condition in the epicardial coronary arteries when microvascular disease is present.⁶

On the other hand, CT-FFR measurement can offer a more specific functional evaluation of individual coronary artery stenoses.⁷ However, the existing CT-FFR approach based on fluid dynamics simulation from static CT image data is computationally intensive, and its diagnostic accuracy may be affected by blooming artifacts arising from dense calcification in the coronary arteries.⁸ In a substudy of the PROMISE trial, 33% of the coronary CT angiography (CCTA) cases were excluded from CT-FFR analysis by the core laboratory due to inadequate image quality.⁹ While only less than 10% of the CCTA cases were excluded by the core laboratory in the more recent FACC study, the reported per-patient accuracy of CT-FFR for detecting functionally significant CAD was below 75% in patients with a Agatston Score above 399.¹⁰ This is similar to the per-patient diagnostic accuracy reported in a substudy of the NXT trial for patients whose Agatston Score exceeded 416.¹¹

To overcome these limitations, a novel CT dynamic angiographic imaging (CT-DAI) algorithm that can derive FFR directly from dynamic myocardial perfusion images has been developed. This approach uses the temporal changes in contrast signal (time-enhancement curve) across a coronary stenosis following a bolus contrast injection to estimate the hemodynamics associated with the stenosis. According to the indicator-dilution principle, the pre- and post- lesion coronary blood flow can be estimated based on the area under the respective time-enhancement curves. Because the relationship between flow velocity and pressure of blood (incompressible fluid) is governed by Bernoulli’s principle, the post-lesion pressure can be estimated once the flow velocity is known, from which FFR can be sequentially derived. Theoretically, the proposed CT-DAI algorithm can rapidly compute a FFR value for a coronary stenosis with or without heavy calcification, since brute-force simulation of fluid dynamics and in-stenosis dynamic contrast signal are not required for the FFR computation. This study sought to evaluate the accuracy of the novel CT-DAI algorithm for FFR measurement compared to corresponding FFR values derived using the invasive catheter-based method in patients with CAD.

Methods

Patients

We retrospectively reviewed the clinical and CT data of 14 patients who had known or suspected CAD and who underwent dynamic CT myocardial perfusion scanning with a second-generation dual-source CT scanner (Somatom Definition Flash, Siemens Healthineers, Erlangen, Germany) between November 2016 and January 2019 at 3 hospitals and institutes. The included patients were part of a prospectively acquired cohort reported in a previously published study, which compared the diagnostic accuracy of dynamic and static CT myocardial perfusion measurement for functional assessment of CAD.¹² A total of 14 cases were selected from the 42 cases available in the cohort for the CT-DAI derived FFR analysis based on the following criteria (Figure 1): (1) both the proximal and distal coronary artery segments across the stenosis of interest were covered sufficiently by the dynamic CT myocardial perfusion scan. The distal segment was required to be at least 1 cm beyond the stenosis; (2) had invasive coronary angiography of the artery where the stenosis of interest was located; (3) underwent invasive FFR measurement for the stenosis of interest. All 14 included patients had undergone cardiac catheterization after the dynamic CT myocardial perfusion scan was performed.

Figure 1.

Patient inclusion flow chart.

CT image acquisition

The prospectively ECG gated dynamic myocardial perfusion scans were acquired with a second-generation dual-source CT scanner (Somatom Definition Flash, Siemens Healthineers, Erlangen, Germany) at the following settings: 100 kV tube voltage, 670 ± 5 mA tube current, 350 ms gantry rotation speed. Approximately 7.5 cm of the heart in the z-axis was imaged 10-14 times with an axial shuttle mode. In each perfusion study, maximum hyperemia was induced by an intravenous infusion of adenosine at 140 µg/kg/min prior to the perfusion scan. At 3 minutes after the start of adenosine infusion, a dynamic perfusion scan was initiated after 50 mL of non-ionic iodinated contrast solution (Iohexol 350 mgI/mL, Bonorex; Central Medical Service, Seoul, Korea) was injected intravenously at 5 mL/s followed by 30 mL of saline flush. The perfusion source images acquired at end-systole over 10-14 time points were reconstructed at 1.5 mm slice thickness with a filtered back projection algorithm dedicated to iodine beam hardening correction.¹³

FFR calculations using dynamic CT perfusion images

The detailed calculation algorithms are provided in Supplemental Materials. In brief, the proposed CT-DAI algorithm for FFR measurement consisted of 3 main steps (Figure 2): First, the pre-lesion and post-lesion volumetric flow rates were computed based on the indicator-dilution principle and the pre-lesion and post-lesion coronary time-enhancement curves sampled from the dynamic myocardial perfusion images. The volumetric flow rates were then converted to the flow velocities based on the lumen radius at the pre-lesion and post-lesion sampling sites. Second, the pre-lesion and post-lesion flow velocities were used to estimate the pressure difference between the 2 sampling sites according to the Bernoulli’s principle. Third, the pressure difference between the 2 locations and the patient’s arterial blood pressure measured with a pressure cuff were used to derive FFR across the coronary lesion.

Figure 2.

Schematic illustration of non-invasive fractional flow reserve (FFR) measurement with the CT-DAI algorithm. CT-DAI = CT dynamic angiographic imaging.

Image analysis

The dynamic contrast-enhanced CT images from each myocardial perfusion scan were analyzed with an in-house software based on the CT-DAI algorithm discussed above. To compute FFR, a single reader placed a region of interest (ROI) within the coronary artery lumen proximal to the stenosis and another ROI 1 cm distal to the lower border of that stenosis to obtain a pair of pre-lesion and post-lesion coronary time-enhancement curves. Depending on the size of the arterial lumen, each sampling ROI was either a block of 2 × 2 or 3 × 3 pixels. The sampled coronary time-enhancement curves were denoised prior to the FFR calculation. Sampling of the post-lesion coronary time-enhancement curve was also repeated at 1.5  and 2.0 cm distal to the lower border of the same stenosis. The post-lesion sampling range (1-2 cm distal to the lower border of the stenosis) was in accordance with the standardized CT-FFR interpretation protocol recommended by an independent group of physicians with years of clinical experience of CT-FFR.^14,15

Catheter based FFR measurement

Each patient underwent standard coronary angiography through either femoral or radical arterial access and received appropriate heparinization. Iodinated contrast solution was injected through a 5 or 6F coronary guiding catheter. FFR was assessed with a 0.014″ pressure wire (Radi Medical Systems, St Jude Medical Inc., United States), which was advanced through the guiding catheter and its pressure recording was equalized with that obtained from the fluid-filled coronary guiding catheter at the origin of the left main or right coronary artery (RCA). The pressure wire was then advanced beyond the stenosis in question such that the measuring point was at least 1 cm beyond the stenosis itself. Approximately 100-200 mg of intracoronary nitroglycerin was administered. Subsequently, adenosine was infused intravenously at 140 µg⋅kg⁻¹⋅min⁻¹ and measurement of distal coronary pressure and proximal coronary/aortic pressure were recorded simultaneously. FFR was calculated as the ratio of distal mean pressure to proximal coronary/aortic pressure.

Myocardial perfusion measurement

To investigate whether FFR measurement with the proposed CT-DAI algorithm provided incremental diagnostic value to CT myocardial perfusion measurement for CAD functional assessment, the qualitative and quantitative CT myocardial perfusion assessment reported in a previously published study were reviewed. The sensitivity, specificity, positive predictive value and negative predictive value of the respective perfusion assessments were determined for the patients included in this study. Myocardial perfusion color maps of each included patient were generated from the corresponding dynamic contrast-enhanced image set using the Syngo.via software (Siemens Healthineers, Erlangen, Germany). The software derived absolute myocardial blood flow based on the maximum slopes of the myocardial time-enhancement curves fitted during a deconvolution analysis.¹⁶ A coronary territory was considered ischemic if the lowest perfusion value in any myocardial segment within this territory was below 1 mL/min/g. Qualitative myocardial perfusion assessment was performed by 2 experienced Cadiologists based on visual interpretation of the dynamic contrast-enhanced images. A coronary territory was considered ischemic if any myocardial segment within this territory was hypo-enhanced for more than 6 consecutive time points in the image set.¹²

Transluminal attenuation gradient measurement

Transluminal attenuation gradient (TAG), defined as the mean decrease in contrast enhancement along a coronary artery in CCTA images, was proposed as an alternative to FFR for hemodynamic evaluation of coronary stenosis without the need of complicated computation as for FFR.¹⁷ Previous studies have shown inconsistent diagnostic values of TAG for functional assessment of CAD.^18–20 To determine whether the proposed CT-DAI algorithm offered a superior diagnostic value than the TAG approach, TAG was derived for each coronary artery included for the present study from the respective CCTA images as described previously.^17,18 The centerline of each coronary artery was reconstructed from the CCTA images using the Volume Viewer software (GE Healthcare, Waukesha, United States), and the resulting cross-sectional lumen view of the coronary artery were used for TAG analysis. A set of circular ROIs were placed along the coronary artery at ∼5-mm intervals until at the distal segment where the lumen diameter was ∼2 mm.²⁰

The CT number (in Hounsfield Unit) at each sampled point along the coronary artery was plotted against the distance relative to the first sample point in the proximal coronary segment. The plot was fitted with a linear regression and the resulting slope was used as the TAG value of that artery.

Statistical analysis

All statistical analyses were performed with the SPSS software (version 29; IBM, Armonk, NY). The FFR values measured using the CT-DAI algorithm and using pressure wires during cardiac catheterization were compared using linear regression analysis (Pearson correlation coefficient and p value for correlation), and Bland-Altman analysis. The mean of the FFR values acquired with the 2 modalities were also compared with a one-way within-subject ANOVA to determine if the difference reached statistical significance. The 95% confidence intervals (CI) of the FFR values measured using CT-DAI were calculated from the following equation:

$$\mathrm{CI}=\overline{x}\pm 1.96·\frac{s}{\sqrt{n}}$$

where $\overline{x}$ and s are the mean and standard deviation of the FFR values measured using CT-DAI, respectively, and n is the number of FFR measurement in our study. The per-vessel and per-patient diagnostic performance (sensitivity, specificity, positive predictive value, and negative predictive value) of CT-DAI for detecting functionally significant CAD were evaluated using binary logistic regression analysis with the FFR values measured at 1 cm distal to the lower border of the stenosis. The optimal cut-off FFR value was determined with the following equation:

$$\mathrm{logit}\mathrm{ }p=\ln \left(\frac{p}{1-p}\right)=ax+b$$

where p is the predicted probability and has a value between 0 and 1, a is the constant and b is the slope of the regression model determined from the regression analysis. The optimal cut-off FFR value was calculated at the 50th percentile (P = .5). Similarly, the diagnostic performance of qualitative myocardial perfusion, quantitative myocardial perfusion, TAG were determined using logistic regression.

Algorithm availability

Our software code is unavailable to the public. Readers that are interested in testing our software are welcome to contact the corresponding author.

Results

Cohort

Based on the inclusion criteria defined above, 15 coronary arteries from 14 patients (all South Koreans) were selected for the evaluation of diagnostic accuracy of the CT-DAI method for detecting functionally significant coronary artery stenosis. The mean age of the included patients was 58.5 years (SD = 10.6; 11 males and 3 females) and their mean weight was 75.8 kilograms (SD = 13.9). The average time between the CT and catheter-based FFR measurements was 12.7 days (SD = 8.9). Table 1 provides a summary of the clinical information of the selected patients, including the total coronary artery calcium score, the severity of luminal narrowing in each coronary artery, age, sex, and body-mass index. The effective dose associated with the stress dynamic myocardial perfusion scan of each patient was also provided.

Table 1.

Clinical and demographic information of the patients selected for the comparison in FFR measurement between the CT-DAI based method and invasive catheter based method.

Study number	Age (years)	Sex	Body mass index (kg/m2)	Coronary artery calcium score (HU)	Effective dose of stress perfusion scan (mSv)
1	53	M	26.2	33.9	9.7
2	46	M	29.8	238.9	11.4
3	73	F	25.9	288.4	10.8
4	64	M	25.1	94.5	11.7
5	51	M	27.1	0	9.1
6	61	M	31.9	30	8.2
7	73	M	24.1	224	9.1
8	76	F	25.2	450	9.9
9	52	M	32.7	0	10.7
10	44	M	28.4	0	10.1
11	66	M	22.6	67.6	8.3
12	57	M	26.4	140.2	9.9
13	47	M	27.0	377.1	9.9
14	56	F	21.6	0	10.0

All 14 patients included for this study were South Koreans (3 from site 1, 5 from site 2, 6 from site 3).

Abbreviation: FFR = fractional flow reserve.

Diagnostic accuracy for functional stenosis

Figures 3-5 show the proximal and distal time-enhancement curves across a coronary artery stenosis sampled from dynamic myocardial perfusion images in 3 patients with varying grades of epicardial coronary artery stenosis. With the CT FFR threshold set to 0.794 as determined by binary logistic regression, the CT-DAI based FFR assessment at 1 cm distal to the lower border of the stenosis agreed with the invasive catheter assessment in all 15 coronary arteries, thus both the per-vessel and per-patient diagnostic accuracy of CT-DAI were 100% (Tables 2 and 3).

Figure 3.

A 46-year-old male with a 50%-70% narrowing in the right coronary artery (RCA). The plaque in the RCA was moderately calcified. The FFR values acquired with the dynamic CT and catheter methods were 0.72 and 0.74, respectively. (A) 3D view of the RCA. (B) Location of the pre-lesion sampling site (single-head arrow) and plaque (double-head arrow) in the curve reformatted coronary CT angiography (CCTA) image. (C) Location of the post-lesion sampling site (single-head arrow) in the curve reformatted CCTA image. The post-lesion sampling site was ∼1 cm distal to the plaque (double-head arrow). (D) Coronary angiography of the RCA. (E) Pre-lesion sampling site (arrow) in the axial perfusion images. (F) Normalized time-enhancement curve sampled from the site shown in (E). (G) Post-lesion sampling site (arrow) in the axial perfusion images. (H) Normalized time-enhancement curve sampled from the site shown in (G). FFR = fractional flow reserve; RCA = right coronary artery.

Table 2.

Comparison of coronary artery disease (CAD) functional assessment between qualitative myocardial perfusion, quantitative myocardial perfusion, transluminal attenuation gradient (TAG) and FFR derived with the CT-DAI algorithm.

Patient	Assessment in myocardium						Assessment in epicardial coronary arteries						Qualitative assessment of stenosis severitya
	Qualitative perfusion			Quantitative perfusion			TAG			FFR (CT-DAI)
	LAD	LCX	RCA	LAD	LCX	RCA	LAD	LCX	RCA	LAD	LCX	RCA	LAD	LCX	RCA
1		1			1			1			1		–	4	–
2			0			0			0			1	–	–	3
3	0			0			1			0			2	–	–
4			0			0			0			0	–	–	3
5	0			1			1			1			4	–	–
6	0		1	1		1	0		0	0		1	3	–	4
7	1			1			1			1			3	–	–
8			0			0			0			0	–	–	0
9			0			0			0			0	–	–	3
10	1b	0		1b	1		0b	0		1b	1		5b	4	–
11			0			0			0			0	–	–	2
12			0			0			0			0	–	–	3
13			0			0			0			0	–	–	3
14			0			0			0			0	–	–	3

“0” represents non-significant CAD and “1” represents significant CAD identified by each assessment method. The functional assessment highlighted in gray represents a discordant diagnosis with the invasive catheter-based FFR method.

^aStenosis severity in each coronary artery was assessed with the following scales: 0 = no narrowing; 1 = 0%-25% narrowing; 2 = 25%-50% narrowing; 3 = 50%-70% narrowing; 4 = 70%-99% narrowing; 5 =  CTO.

^bThe CTO artery was excluded for the diagnostic performance evaluation shown in Table 3 since it had no invasive FFR measurement.

Abbreviations: CT-DAI = CT dynamic angiographic imaging; FFR = fractional flow reserve; LAD = left anterior descending; RCA = right coronary artery.

Table 3.

Per-vessel and per-patient diagnostic performance of qualitative myocardial perfusion, quantitative myocardial perfusion, TAG and FFR derived with the CT-DAI algorithm for detecting functionally significant coronary artery disease.

	Per-vessel analysis				Per-patient analysis
	Qualitative Perfusion	Quantitative Perfusion	TAG	FFR (CT-DAI)	Qualitative Perfusion	Quant Perfusion	TAG	FFR (CT-DAI)
Sensitivity	75%	83.3%	50%	100%	50%	83.3%	50%	100%
Specificity	100%	88.9%	87.5%	100%	100%	100%	87.5%	100%
PPV	100%	83.3%	75%	100%	100%	100%	75%	100%
NPV	75%	88.9%	70%	100%	75%	88.9%	70%	100%
Accuracy	80%	86.7%	73.3%	100%	78.6%	85.7%	71.4%	100%

Abbreviations: CT-DAI = CT dynamic angiographic imaging; FFR = fractional flow reserve; NPV = negative predictive value; PPV = positive predictive value; TAG = transluminal attenuation gradient.

Figure 4.

A 53-year-old male with a 70%-99% non-calcified stenosis in the left circumflex (LCx). The FFR values obtained with the dynamic CT and catheter methods were 0.38 and 0.40, respectively. (A) 3D view of the LCx. (B) Location of the pre-lesion sampling site (top arrow) and plaque (bottom arrow) in the curve reformatted CCTA image. (C) Location of the post-lesion sampling site (bottom arrow) and plaque (top arrow) in the curve reformatted CCTA image. (D) Coronary angiography of the LCx. (E) Pre-lesion sampling site (arrow) in the axial perfusion images. (F) Normalized time-enhancement curve sampled from the site shown in (E). (G) Post-lesion sampling site (arrow) in the axial perfusion images. (H) Normalized time-enhancement curve sampled from the site shown in (G). CCTA = coronary CT angiography; FFR = fractional flow reserve; LCx = left circumflex.

Figure 5.

A 51-year-old male whose left anterior descending (LAD) artery was 70%-99% stenosed. The FFR value measured with the dynamic CT and catheter methods were 0.72 and 0.71, respectively. (A) 3D view of the LAD. (B) Location of the pre-lesion sampling site (right arrow) and plaque (left arrow) in the curve reformatted CCTA image. (C) Location of the post-lesion sampling site (bottom arrow) and plaque (top arrow) in the curve reformatted CCTA image. (D) Coronary angiography of the LAD. (E) Pre-lesion sampling site (arrow) in the axial perfusion images. (F) Normalized time-enhancement curve sampled from the site shown in (E). (G) Post-lesion sampling site (arrow) in the axial perfusion images. (H) Normalized time-enhancement curve sampled from the site shown in (G). CCTA = coronary CT angiography; FFR = fractional flow reserve.

Two coronary arteries were excluded for evaluation of the diagnostic accuracy at different sampling locations distal to stenosis. In one of the excluded cases (study number 1 in Tables 1 and 2), the coronary artery bifurciated immediately after the 1 cm post lesion sampling location, and consequently the 1.5 and 2 cm post lesion sampling locations fell into one of the bifurciated branches. In the other excluded case (study number 4 in Tables 1 and 2), an independent stenosis was observed between the lower border of the first stenosis and the 1.5/2 cm post lesion sampling location. In the remaining 13 coronary arteries, the per-vessel and per-patient diagnostic accuracy of CT-DAI remained unchanged between 1 and 2 cm distal to the lower border of the stenosis.

FFR measurements

Linear regression analysis of the FFR values measured by the CT-DAI and catheterization was performed in only 12 of 15 coronary arteries. The exact FFR value measured by the pressure wire was not recorded in 3 coronary arteries (study number 3, 6 left anterior descending [LAD] and 10 left circumflex [LCx] in Tables 1 and 2), with only above or below the 0.80 threshold being recorded. The mean and standard deviation of FFR measured with the CT-DAI method was 0.768 and 0.156, with a 95% confidence interval of 0.728-0.808, and not statistically different (P = .655) from the values measured by invasive catheterization method (M = 0.796, SD = 0.149). The linear regression coefficient (R) of the 2 sets of FFR measurements was 0.910 (P < .001, Figure 6). The Bland-Altman analysis revealed no significant proportional bias (See online supplementary material for a color version of this Figure S7).

Figure 6.

Linear regression plot of the FFR values measured with the dynamic CT (y-axis) and catheter (x-axis) methods. FFR = fractional flow reserve.

Comparison with other CT functional measurements

Tables 2 and 3 summarize the results of CAD functional assessment with qualitative myocardial perfusion, quantitative myocardial perfusion, TAG deriving from CCTA images, and FFR deriving from myocardial perfusion images with the CT-DAI algorithm. On both per-vessel and per-patient analyses, FFR measurement with CT-DAI yielded the highest accuracy for CAD functional assessment compared to other CT functional assessment methods.

Discussion

To the best of our knowledge, this is the first study that demonstrates the feasibility of FFR measurement directly from dynamic CT myocardial perfusion images. Our findings revealed that the FFR measurements with the novel CT-DAI algorithm were highly accurate compared to the gold standard invasive catheter FFR measurements for assessing the functional severity of coronary stenosis. Furthermore, the FFR values derived with CT-DAI showed excellent correlation with the FFR values measured with the pressure wires during cardiac catheterization.

As opposed to the conventional CT-FFR technology which is based on static image data (time-invariant cardiac images) acquired at rest coupled with simulation of the hyperemic hemodynamics, the FFR analysis with CT-DAI was based on dynamic data (time-variant images) acquired during hyperemia. Since no intensive computer simulation was involved, the FFR value associated with each coronary artery stenosis was computed in just a few seconds. The superior computational speed offered by this new approach may facilitate a wider application of CT-FFR. Additionally, the proposed technology can be used in conjunction with myocardial perfusion imaging for a comprehensive functional assessment of CAD at the macro and microvascular levels simultaneously. In the patients with known or suspected microvascular disease, such concomitant assessment would provide valuable insights on the underlying cause of discordance between epicardial and microvascular hemodynamics from a single CT imaging dataset.

In this study, the accuracy of quantitative myocardial perfusion measurement for CAD functional assessment was comparable to that reported in a recent meta-analysis.²¹ FFR measurement with CT-DAI demonstrated a superior accuracy to qualitative and quantitative myocardial perfusion measurements, suggesting that CT-DAI may provide incremental diagnostic value to the more established dynamic myocardial perfusion approach. Additionally, the TAG analysis further illustrated that the proposed CT-DAI algorithm is not simply a variant of the TAG approach but is based on a more solid imaging physics and physiological basis, thus leading to a more reliable functional assessment. Patient 10 in Tables 1 and 2 had an intermediately stenosed LCx and his LAD had a chronic total occlusion (CTO). While the TAG analysis suggested the LAD artery was not functionally significant, the CT-DAI assessment revealed the FFR at 1 and 2 cm distal to the stenosis was 0.57 and 0.41, respectively (average FFR value was 0.49 between 1 and 2 cm distal to the stenosis). This finding agreed with the FFR values (<0.5) reported in the coronary arteries with CTO measured using invasive catheter based method,^22–24 suggesting that CT-DAI may be useful to inform treatments in CTO via non-invasive assessment of the magnitude of collateral circulation in occluded or near occluded coronary arteries.

None of the patients included in this study had myocardial perfusion imaging acquired at rest, and as such, the diagnostic accuracy of CT-DAI without vasodilator administration could not be evaluated. Our future studies will investigate whether the proximal/distal flow across a stenosis measured at rest with CT-DAI can be a useful imaging biomarker to determine the functional severity of coronary artery stenosis.

This study had several limitations. First, only 15 coronary arteries were available for analysis against the clinical gold standard. The dynamic CT myocardial perfusion scans were acquired with a second generation dual-source CT scanner, which only covered ∼7.5 cm of the heart in dynamic perfusion imaging. As a result, many coronary arteries from the 42 patients were excluded from analysis because the proximal segment of the artery was not sufficiently covered by the perfusion scan. Second, the selected coronary arteries had only moderate degree of calcification within the arterial lumen, which was insufficient for a comprehensive evaluation on the capability of the CT-DAI algorithm for accurate FFR measurement in more heavily or diffusely calcified coronary arteries. The proposed technology relies on the dynamic temporal information proximal and distal to a stenosis to compute FFR, and should be theoretically unaffected by the blooming artifacts emanating from heavy lumen calcification since in-stenosis temporal information is not required for the FFR computation using CT-DAI algorithm.

In conclusion, the novel CT-DAI algorithm can reliably compute FFR across coronary artery stenosis directly from dynamic CT myocardial perfusion images. This may facilitate rapid on-site hemodynamic assessment of the epicardial coronary artery stenosis in patients with CAD.

Author contributions

Aaron So, PhD (Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Writing—original draft, Writing—review & editing) Ki S. Choo, MD (Data curation, Investigation, Project administration, Writing—review & editing) Ji W. Lee, MD (Data curation, Investigation, Project administration, Writing—review & editing) Yun-Hyeon Kim, MD (Data curation, Investigation, Project administration, Writing—review & editing) Mustafa Haider, MSc (Formal analysis, Investigation, Validation, Writing—review & editing) Mahmud Hasan, MSc (Investigation, Software, Validation, Writing—review & editing) Serag El-Ganga, MSc (Formal analysis, Investigation, Validation, Writing—review & editing) Akshaye Goela (Data curation, Formal analysis) Patrick Teefy, MD (Investigation, Supervision, Writing—review & editing) Yeon H. Choe, MD (Data curation, Investigation, Project administration, Supervision, Writing—review & editing)

Supplementary material

Supplementary material is available at Radiology Advances online.

Funding

This work was supported by the Canadian Institutes of Health Research (CIHR) Project Grant 169036, the Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant RGPIN-2016-06565, and the NSERC Idea to Innovation Grant 567667-21.

Conflict of interest

A.S. receives a U.S. patent (US12059283B2) for the CT-DAI technology discussed in the manuscript. Please see ICMJE form(s) for author conflicts of interest. These have been provided as supplementary materials.

Data availability

Data generated or analyzed during the study are available from the corresponding author by request.