Fully quantitative cardiovascular magnetic resonance myocardial perfusion ready for clinical use: a comparison between cardiovascular magnetic resonance imaging and positron emission tomography

Ischemic heart disease (IHD) is the most common cause of mortality and morbidity in the western world. Obstructive coronary artery disease (CAD) compromising myocardial perfusion is the main pathophysiological mechanism underlying stress-induced myocardial ischemia. The extent of stress-induced ischemia due to CAD has been shown to be of important prognostic value [1, 2]. Significant advances in both diagnosing and treating epicardial CAD have improved survival and reduced morbidity during the last decades in patients suffering from this disease [3‐5]. Most clinical methods used for diagnosing stress-induced myocardial ischemia focus on epicardial obstructive CAD and rely on qualitative or semi-quantitative assessment of the relative spatial perfusion distribution i.e. myocardial perfusion single photon emission computed tomography (SPECT), first-pass perfusion cardiac magnetic resonance (CMR) and non-dynamic cardiac positron emission tomography (PET). There are, however, data suggesting that the use of qualitative or semi-quantitative methods for assessing stress-induced ischemia have limited ability to depict the full extent of the atherosclerotic coronary disease [6]. Furthermore, recent studies have shown that full quantification of myocardial perfusion (MP) expressed in terms of ml/min/g myocardial tissue offers independent prognostic information in patients with CAD [7‐9].

Cardiovascular magnetic resonance (CMR) with a dual sequence, single contrast bolus approach has been shown to enable quantitative assessment of MP [10]. However, there have been challenges associated with this technique such as the non-linear relationship between signal intensity and gadolinium concentration, respiratory motion correction and gadolinium kinetic modelling. Recent technical developments have shown promising results to overcome these challenges [11]. The proposed CMR perfusion mapping method is implemented in-line, is fully automatic and is readily integrated into the clinical workflow via the Gadgetron framework [12]. Thus, this technique seems to be ready for clinical implementation, making quantitative perfusion available as a routine examination in patients with suspected or established ischemic heart disease. Even though this is a promising technique, clinical validation against an independent reference standard is still lacking.

Cardiac positron emission tomography (PET) with dynamic image acquisition during bolus injection of a radioactive perfusion tracer at rest and stress is currently considered to be the clinical reference method for quantitative assessment of MP. However, cardiac PET is of limited use as a wide spread routine clinical examination due to its limited availability and sometimes challenging logistics as well as due to the ionizing radiation associated with this technique.

Therefore, the aim of this study was to validate MP by a fully automatic, in-line, dual sequence, single contrast bolus CMR perfusion mapping sequence using cardiac 13N–NH3 PET as the reference method in stable CAD patients.

Patient characteristics are shown in Table 1. Typical images obtained by PET and CMR in a patient with right coronary artery (RCA) disease and stress-induced ischemia in the inferior LV wall are shown in Fig. 1. Myocardial infarction was detected in 4 of 21 patients. Infarct size in these 4 patients was 8 ± 5%. Of the 12 patients that underwent CMR and PET before angiography, 4 patients had signs of stenosis in left anterior descending coronary artery (LAD) of whom 3 underwent revascularization, 3 patients had stenosis in RCA of whom 2 underwent revascularization, 1 patient had stenosis in the left circumflex coronary artery (LCX) who underwent revascularization and 6 patients had no significant stenosis. Of the patients with CMR and PET 6 months after angiography 4 patients had stenosis in LAD of whom 3 had been revascularized, 2 had stenosis in RCA of whom 1 had been revascularized, 2 patients had stenosis in LCX of whom 1 had been revascularized and 2 patients had no significant stenosis. Complete records of the PET and CMR rate pressure products at both rest and stress were available for 12 patients. There were similar mean rate pressure products for PET and CMR both at rest (8932 ± 1563 vs 9189 ± 1756, p = 0.37) and at stress (10,945 ± 2170 vs 11,351 ± 2463, p = 0.53). Furthermore, the difference between stress and rest rate pressure products did not differ between PET and CMR (2014 ± 2554 vs 2162 ± 2120, p = 0.75) indicating similar hemodynamic responses to adenosine stress.

The present study shows that fully quantitative CMR MP mapping is ready for clinical use by validating the method against 13N–NH3 PET, which was used as the independent reference standard. There was strong correlation between the two methods with negligible bias for both global and regional MP. Thus, using CMR for quantifying MP is an attractive alternative to PET in the clinic. However, routine clinical implementation of quantitative CMR MP will rely on its importance on therapeutic decision making, which needs to be proven by large clinical outcome trials that use MP as an endpoint.

Quantification of MP and MPR in routine clinical practice will have important implications on the management of patients with established or suspected CAD. Even though a normal perfusion scan with qualitative or semi-quantitative evaluation of first-pass perfusion has been shown to predict low risk of future coronary events [22‐24], there remain patients in this group with adverse prognosis due to multi-vessel disease or non-obstructive CAD [25‐27]. Thus, even in the absence of obstructive CAD, such as for 3 of 12 patients in Table 2, patients with pathologic MP at stress and low MPR are most likely in need of medical attention. The pathophysiological mechanisms explaining a decreased MP in the absence of obstructive coronary artery disease are likely related to disease states shown to affect the coronary microvascular function, such as dyslipidemia [28, 29], hypertension [30], diabetes [31, 32], and smoking [33]. Significant vascular dysfunction related to these diseases may precede development of obstructive CAD [34], which makes quantification of MP and MPR potential diagnostic methods for identifying patients at risk of developing coronary stenosis. Furthermore, there are currently patients being treated with PCI in coronaries supplying myocardium with normal MPR [35]. Thus, these patients are put at risk associated with invasive treatment (endothelial injury, bleeding complications, risk of coronary artery dissection etc.) with little or no benefit to be gained from the procedure.

Using relative perfusion distribution, typically applied in routine clinical practice of nuclear cardiology, for diagnosing ischemic heart disease means that the counts in the perfusion maps are normalized to the highest regional counts. Thus, there might be variability in counts within normal myocardium due to variability of perfusion within the normal range. Hence, parts of the normal myocardium might have lower counts, potentially interpreted as ischemia and therefore compromising the diagnostic accuracy as recently shown by Kajander et al. [36]. They showed that the positive and negative predictive values for detecting or excluding invasively diagnosed fractional flow reserve (FFR)-positive coronary stenoses decreased significantly when using relative spatial perfusion distribution instead of quantitative MP.

One major challenge with quantitative perfusion imaging, including both CMR and PET in the present study, is to find a perfusion tracer with uptake kinetics that reflects the actual myocardial perfusion over a wide range of blood flows. It has previously been shown that 13N–NH3 uptake is approximately linearly proportional to flow at low MPs but that there is an underestimation of actual blood flow at high flows, which was compensated for by correcting for ammonia extraction [37]. The same underestimation has been shown for CMR and gadolinium-based extracellular contrast agents [38]. In the current approach [11], however, the gadolinium extraction fraction is estimated and used in calculating MP to avoid this underestimation of MP for CMR.

Imaging for both PET and CMR was initiated at the same time point (3 min) following the start of adenosine infusion. However, the duration for PET imaging and adenosine administration for the PET examination was significantly longer than for CMR. Thus, the adenosine response may differ somewhat [39] affecting the validation to some extent.

The results in the present study indicate that a dual sequence, single contrast bolus CMR perfusion imaging and 13N–NH3 PET have similar abilities in terms of quantifying myocardial blood flow over a wide range of flows. Thus, these techniques should be interchangeable and could therefore be used in accordance with local experience and expertise on the method of choice. A major limitation with 13N–NH3 PET is the cumbersome logistics with producing the tracer, requiring a cyclotron close to the imaging facility. Given the short half-life of 13N–NH3 (< 10 min) the tracer has to be produced and transported to the PET scanner within a limited time window. This requires advanced logistics and significant manpower, which also affects the cost per examination. With dual sequence, single contrast bolus CMR there are no such limitations. CMR offers superior characterization of myocardial morphology and function compared to 13N–NH3 PET due to its excellent spatial resolution. Furthermore, PET imaging is associated with ionizing radiation which CMR is not, making CMR the method of choice in younger patients and in situations where serial follow-up examinations is expected to be necessary, such as monitoring of treatment effects.

For wide spread clinical use of quantification of CMR-based MP, image acquisition, reconstruction and post-processing need to be both automated and standardized. The CMR technique described in the present study is highly automated and standardized with regards to imaging protocol, image reconstruction and post-processing of the data. The same requirements regarding automation and standardization is true for cardiac PET involving different perfusion tracers and different mathematical data modeling for quantification of MP built into different imaging software. The use of 13N–NH3 for quantification of MP and MPR has recently been shown to be stable across different software implementations [40].

Quantification requires knowledge of the amount of contrast agent in the myocardial tissue and the arterial input function (AIF) driving the delivery of this contrast agent. Accurate quantification is challenged by the lack of linearity between the measured signal and contrast agent concentration. The main sources of non-linearity and bias are: spatial signal variations caused by the sensitivity profiles of the surface coils, imperfect saturation of magnetization during contrast bolus passage, T2* decay (and signal loss) caused by high contrast agent concentrations in the blood pool, and the non-linear signal response inherent due to saturation recovery that depends on the parameters of the imaging protocol. The proposed dual sequence [11] is designed to address all of these issues. It has been proposed that some of the non-linearity of the AIF response curve can be mitigated by imaging the AIF during a separate injection of a bolus with lower concentration (the dual bolus approach) [41], but this approach has some practical drawbacks as it requires multiple injections and a longer overall acquisition time. Moreover, there are other potential bias sources with this approach, since changes in breathing, etc. between the two measurements may introduce new sources of variation. Consequently, it is desirable to acquire the AIF curve simultaneously with the tissue response curve.

There is a good agreement between MP quantified by CMR with inline perfusion flow mapping and 13N–NH3 PET in patients with suspected stable CAD. Thus, CMR is a viable alternative in clinical practice for quantification of MP.