Comparison between quantitative cardiac magnetic resonance perfusion imaging and [¹⁵O]H₂O positron emission tomography

Sixty patients with stable CAD referred on a clinical basis to the Amsterdam University Medical Centers, location VUmc, were prospectively enrolled. Exclusion criteria were presence of myocardial scar on late gadolinium enhancement, history of coronary artery bypass grafting, acute myocardial infarction, atrial fibrillation, significant valvular disease, heart failure, non-ischemic cardiomyopathy, renal insufficiency (eGFR < 45 mL/min), and contraindications to intravenous adenosine or CMR. All patients underwent [¹⁵O]H₂O PET followed by CMR within 7 days irrespective of the results of [¹⁵O]H₂O PET. Revascularization procedures and modifications to pharmacological therapy were not permitted in-between PET and CMR. All study procedures complied with the 1964 Helsinki declaration and its later amendments. The protocol was approved by the Medical Ethics Review Committee of the Amsterdam UMC, location VUmc. Written informed consent was obtained from all individual participants included in the study.

[¹⁵O]H₂O PET was performed using a Gemini TF 64 hybrid PET/CT scanner (Philips Healthcare, Best, the Netherlands). Patients were instructed to refrain from products containing caffeine or xanthine for 24 h prior to scanning. Images were first obtained during resting conditions and thereafter during vasodilator stress. The PET sequence has been described in detail previously [9]. Briefly, 370 MBq of [¹⁵O]H₂O was injected intravenously as a 5 mL bolus (0.8 mL/s), immediately followed by a 35 mL saline flush (2 mL/s). A dynamic PET emission scan of 6 min was started simultaneously with tracer administration. After a delay of 15 min, an identical PET sequence was performed during continuous infusion of adenosine through a second venous cannula at a dose of 140 μg/kg/min. Adenosine was started 2 min prior to PET scanning to ensure maximal vasodilation. To correct for photon attenuation and scatter, low-dose (10 mA) respiration-averaged computed tomography scans were obtained during normal breathing just before the rest scan and immediately after the stress scan. Post-processing of PET data was done by a single observer (PvD), who was blinded to all clinical and CMR data. Parametric images of rest and stress perfusion were generated in approximately 10 min using an in-house developed software package, Cardiac VUer [10]. Absolute MBF was quantified in mL per minute per g of perfusable tissue.

Again, patients were instructed to refrain from products containing caffeine or xanthine for 24 h prior to image acquisition. All CMR images were obtained on a 1.5-T whole body MR scanner (Magnetom Avanto, Siemens, Erlangen, Germany). Perfusion imaging was performed using a dual sequence, single bolus technique [11], implemented as a Siemens works in progress software by C. Glielmi. Perfusion images were acquired using an echo-planar imaging sequence in three parallel short-axis slices planned at the basal, mid, and apical levels. To assess the arterial input function, low-resolution turboFLASH images were obtained at the basal level using a sequence optimized for the high gadolinium concentration. Perfusion images were obtained every heartbeat for 50–70 cardiac cycles following intravenous injection of a 0.075 mmol/kg bolus of a gadolinium-based contrast agent (DOTAREM®, Guerbet, Villepinte, France). Patients were asked to hold their breath as long as possible and breathe slowly thereafter. In-plane respiratory motion of the heart was corrected using non-rigid registration [12]. Perfusion images were corrected for surface coil-induced signal inhomogeneities using a separate prescan normalization [13]. Typical in-plane resolution of the myocardial perfusion images was 2.5 × 2.5 mm, with a slice thickness of 10 mm (pre-pulse 90°, repetition time 5.6 ms, echo planar factor 4, echo time 1.1 ms, saturation time 110 ms, flip angle 18^°, matrix size 160 × 144, parallel imaging in the temporal direction [TGRAPPA] [14] factor 2). Perfusion imaging was performed first during vasodilator stress, which was induced by continuous infusion of adenosine using the same protocol as applied during PET. Rest perfusion images were obtained 15 min after stress imaging using identical scanning parameters and slice location. Left ventricular cardiac function was assessed in between stress and rest perfusion with steady-state free-precession cine imaging. Late gadolinium enhancement (LGE) was performed 12–15 min after rest perfusion using a 2D segmented inversion-recovery gradient-echo pulse sequence. Analysis of CMR data was done by a single observer (HE), who was blinded to all clinical and PET data. Post-processing of CMR perfusion images was performed in approximately 15 min using dedicated research software (MASS version 2017-Exp, Leiden, the Netherlands). A region of interest was placed in the LV blood pool of the image series obtained for the arterial input function. Care was taken to avoid inclusion of papillary muscles. Endocardial and epicardial contours were drawn manually on a single phase of each slice of the myocardial perfusion images. Subsequently, these contours were propagated to the other phases. Care was taken to avoid inclusion of blood pool or epicardial fat. Rest and stress MBF were quantified in mL per minute per g using Fermi function-constrained deconvolution, as described previously [15]. Cine and LGE images were analyzed using a commercially available software (QMASS version 7.6, Medis, Leiden, the Netherlands). Left ventricular (LV) end-diastolic volume, end-systolic volume, and ejection fraction were calculated from the cine images. LGE images were visually assessed in order to rule out presence of LV scar tissue.

Perfusion data were analyzed according to the 17-segment model of the American Heart Association (AHA) [16]. The apical cap (segment 17) was excluded from analysis since this segment was not in the imaging planes of the CMR perfusion acquisition. Myocardial segments were also excluded from analysis if either PET or CMR perfusion imaging was of insufficient quality. Global rest and stress MBF were calculated by averaging perfusion over all 16 segments. In addition, myocardial segments were allocated to the three vascular territories (LAD, left anterior descending; LCx, left circumflex artery; and RCA, right coronary artery) as follows: LAD, segments 1,2,7,8,13,14; LCx, segments 5,6,11,12,16; and RCA, segments 3,4,9,10,15. Rest and stress MBF were calculated for each vascular territory by averaging perfusion over the corresponding myocardial segments. Myocardial flow reserve (MFR) was defined as the ratio of stress to rest MBF and was calculated on a global as well as a regional level. Concordance between CMR and PET was assessed on a per-vessel basis. For [¹⁵O]H₂O PET, stress MBF ≤ 2.3 mL/min/g and MFR ≤ 2.5 were considered abnormal according to previously validated cutoff values for diagnosing hemodynamically obstructive CAD (i.e., fractional flow reserve ≤ 0.80) [17]. Receiver operator characteristic (ROC) curve analysis and the Youden index were used to define optimal cutoff values for CMR measurements of stress MBF and MFR (MBF_CMR and MFR_CMR).

Figure 2 displays the relationship between CMR and PET measurements of global myocardial perfusion. CMR-derived MBF (MBF_CMR) and PET-derived MBF (MBF_PET) obtained during vasodilator stress showed only moderate correlation (r = 0.41; P < 0.001) and poor to moderate inter-method reliability (ICC for absolute agreement = 0.40 [95% confidence interval (CI): 0.17 to 0.59]; P < 0.001). In addition, CMR-derived MFR (MFR_CMR) and PET-derived MFR (MFR_PET) showed a moderate correlation (r = 0.44; P < 0.001) and poor to moderate inter-method reliability (ICC for absolute agreement = 0.36 [95% CI: 0.09 to 0.58]; P < 0.001). Bland-Altman analysis demonstrated a mean bias of 0.2 ± 0.9 mL/min/g for stress MBF and − 0.5 ± 1.0 for MFR. Table 3 (top row) displays the mean values of global rest MBF, stress MBF and MFR as measured using CMR and PET. CMR measurements of global rest MBF were significantly higher than those obtained using PET (1.2 ± 0.3 vs. 0.9 ± 0.2 mL/min/g; P < 0.001), whereas global stress MBF did not differ between the techniques (3.1 ± 0.9 vs. 2.9 ± 0.8 mL/min/g; P = 0.14). Global MFR was significantly lower for CMR in comparison to PET (2.6 ± 0.7 vs. 3.2 ± 1.0; P < 0.001).

The relationship between CMR and PET measurements of regional myocardial perfusion is shown in Fig. 3. On a per vessel basis, stress MBF_CMR and stress MBF_PET showed only moderate correlation (r = 0.39; P < 0.001) and poor inter-method reliability (ICC for absolute agreement = 0.38 [95% CI: 0.25 to 0.50]; P < 0.001). In addition, only modest correlation (r = 0.36; P < 0.001) and poor inter-method reliability (ICC for absolute agreement = 0.30 [95% CI: 0.13 to 0.46]; P < 0.001) were present between MFR_CMR and MFR_PET. Bland-Altman analysis revealed a mean bias of 0.2 ± 1.0 for stress MBF and − 0.5 ± 1.2 for MFR. CMR demonstrated a tendency to underestimate MFR at higher values. Table 3 (bottom rows) lists the mean values of CMR and PET measurements of rest MBF and stress MBF and MFR. Rest and stress MBF were significantly higher for CMR compared with PET (1.2 ± 0.3 vs. 0.9 ± 0.2 mL/min/g; P < 0.001 for rest MBF and 3.1 ± 0.9 vs. 2.9 ± 0.8 mL/min/g; P = 0.014 for stress MBF). Conversely, MFR_CMR was significantly lower than MFR_PET (2.7 ± 0.9 vs. 3.2 ± 1.1; P < 0.001). [¹⁵O]H₂O PET demonstrated abnormal stress MBF and MFR in respectively 49 (27%) and 53 (29%) vascular territories. Figure 4 displays the ROC curves of quantitative CMR perfusion imaging for detecting abnormal stress MBF and MFR as defined by [¹⁵O]H₂O PET. Stress MBF_CMR displayed an area under the curve (AUC) of 0.72 (95% CI: 0.65 to 0.79) and had an optimal cutoff value of 2.35 mL/min/g. MFR_CMR had an AUC of 0.76 (95% CI: 0.69 to 0.83) and an optimal cutoff value of 2.25. Using these cutoff values, stress MBF_CMR and MFR_CMR were abnormal in respectively 35 (20%) and 48 (29%) vascular territories. CMR and PET were concordant in 137 (77%) vascular territories for stress MBF and in 135 (80%) vascular territories for MFR.

The present study lacks invasive confirmation of hemodynamically obstructive CAD. Although [¹⁵O]H₂O PET is considered to be the reference standard for quantification of myocardial perfusion, invasive measurements of fractional flow reserve are the preferred reference for diagnosing hemodynamically obstructive CAD and guiding revascularization. Therefore, our results urge for a new study comparing quantitative PET and CMR head-to-head against invasive measures of physiology. Secondly, resting flow is known to vary according to metabolic demand [37]. Some of the observed discrepancy between CMR and PET is therefore not attributable to differences in methodology but results from physiological fluctuations in rest flow. Finally, the sequence of imaging was similar in all patients as [¹⁵O]H₂O PET was always performed prior to CMR with a maximum delay of 7 days. Although medication and treatment were kept constant, physiological changes in myocardial perfusion may have occurred in-between PET to CMR.