To compare myocardial viability assessment accuracy of cardiac magnetic resonance imaging (CMR) compared to [¹⁸F]-fluorodeoxyglucose (FDG)- positron emission tomography (PET) depending on left ventricular (LV) function.

One-hundred-five patients with known obstructive coronary artery disease (CAD) and anticipated coronary revascularization were included in the study and examined by CMR on a 1.5T scanner. The CMR protocol consisted of cine-sequences for function analysis and late gadolinium enhancement (LGE) imaging for viability assessment in 8 mm long and contiguous short axis slices. All patients underwent PET using [¹⁸F]-FDG. Myocardial scars were rated in both CMR and PET on a segmental basis by a 4-point-scale: Score 1 = no LGE, normal FDG-uptake; score 2 = LGE enhancement < 50% of wall thickness, reduced FDG-uptake ( ≥ 50% of maximum); score 3 = LGE ≥ 50%, reduced FDG-uptake (< 50% of maximum); score 4 = transmural LGE, no FDG-uptake. Segments with score 1 and 2 were categorized “viable”, scores 3 and 4 were categorized as “non-viable”. Patients were divided into three groups based on LV function as determined by CMR: Ejection fraction (EF), < 30%: n = 45; EF: 30%-50%: n = 44; EF > 50%: n = 16). On a segmental basis, the accuracy of CMR in detecting myocardial scar was compared to PET in the total collective and in the three different patient groups.

CMR and PET data of all 105 patients were sufficient for evaluation and 5508 segments were compared in total. In all patients, CMR detected significantly more scars (score 2-4) than PET: 45% vs 40% of all segments (P < 0.0001). In the different LV function groups, CMR found more scar segments than PET in subjects with EF< 30% (55% vs 46%; P < 0.0001) and EF 30%-50% (44% vs 40%; P < 0.005). However, CMR revealed less scars than PET in patients with EF > 50% (15% vs 23%; P < 0.0001). In terms of functional improvement estimation, i.e., expected improvement after revascularization, CMR identified “viable” segments (score 1 and 2) in 72% of segments across all groups, PET in 80% (P < 0.0001). Also in all LV function subgroups, CMR judged less segments viable than PET: EF < 30%, 66% vs 75%; EF = 30%-50%, 72% vs 80%; EF > 50%, 91% vs 94%.

CMR and PET reveal different diagnostic accuracy in myocardial viability assessment depending on LV function state. CMR, in general, is less optimistic in functional recovery prediction.

Left ventricular (LV) dysfunction due to myocardial ischemia is one of the most common manifestations in chronic coronary artery disease (CAD), but does not necessarily represent non-viable, irreversibly injured tissue[1-4]. Although large multicenter studies such as the positron-emission-tomography (PET) and recovery following revascularization PARR-2 trial[5,6] or the surgical treatment for ischemic heart failure STICH-trial have been performed, even today there is no general consensus as to when assigning patients to either optimized medical treatment alone or to a combination of medical treatment plus revascularization procedures [(percutaneous coronary intervention or coronary artery bypass graft (CABG) surgery)][7-12]. Arriving at the optimal management strategy for these patients is a complex multifactorial process that includes not only viability but also processes such as ischemia and remodeling[4,7,8,13-15].

Currently, several imaging modalities are used for the evaluation of myocardial viability, each of them assessing different myocardial features: e.g., low-dose dobutamine stress echocardiography (DSE), nuclear techniques such as PET or single-photon-emission-computed-tomography (SPECT) as well as cardiac magnetic resonance imaging (CMR). Traditionally, nuclear techniques were regarded as gold standard for viability testing owing to their high sensitivity and negative predictive value (NPV) [e.g., fluorodeoxyglucose (FDG)-PET 92% and 87%, respectively][7,12,16,17]. For this purpose mainly three tracers are used evaluating cell membrane integrity, perfusion and intact mitochondria ([²⁰¹Tl] thallium- or [^99mTc] technetium-SPECT)[18] or maintained metabolism ([¹⁸F] FDG-PET)[2].

With recent advances of the hard and software, especially the introduction of non-breath-hold sequences and arrhythmia rejection protocols[19,20], CMR has become a versatile alternative to nuclear imaging, coming along with an excellent spatial resolution[17]. While late gadolinium enhancement (LGE) allows visualization of the transmural extent of the scar by achieving signal intensity differences of nearly 500% between irreversibly injured and viable myocardium[14,18,21], dobutamine stress CMR analyzes contractile reserve of dysfunctional myocardium similar to DSE[2,7]. Although the low specificity of LGE-CMR is well known, which is mainly attributable to the variable functional recovery in segments with LGE covering 25%-75% of the wall[1,14,22], its general high diagnostic accuracy for detecting myocardial scars has been proven in different studies[2,7,14,22-25]. However, so far, the value of CMR and PET has not been defined considering different LV function states. The aim of this study therefore was to evaluate the diagnostic accuracy of LGE-CMR viability assessment and PET and to compare them in patient groups with different LV functions.

All patients successfully finished the study protocol, therefore complete data sets of 105 subjects underwent analysis. Depending on the long axis diameter, the LV was covered in CMR by 8 to 14 short axis slices. The 3D PET data set was then separated accordingly into the same number of slices yielding a total of 5508 segments. Mean CMR volumetric measures were: EDV, 198 ± 69 mL (range, 63-386 mL) and ESV, 137 ± 66 mL (range, 24-316 mL), resulting in a mean EF of 34% ± 14% (range, 9%-78%). Forty-five patients had an EF < 30%, 44 patients 30%-50%, and 16 patients > 50%.

Scar assessment by CMR and PET depending on LV function

As demonstrated by Figure 1, CMR detected myocardial scars (score 2-4) in 45% of all segments, while PET depicted scars in 40% of all segments (P < 0.0001). Inter-observer agreement (Cohen’s Kappa) between CMR and PET in scar detection was 0.39 (fair to moderate). Analysis of the different patient groups revealed that CMR found more scars than PET in subjects with EF < 30% (55% vs 46%; P < 0.0001) and EF 30%-50% (44% vs 40%; P < 0.005). However, CMR revealed less scars than PET in patients with EF > 50% (15% vs 23%; P < 0.0001). Statistical values (sensitivity, specificity, PPV, NPV, and accuracy) within the 3 different patient groups are summarized in Table 2.

Open in New Tab   Full Size Figure   Download Figure

Figure 1 Histogram showing the frequency of scar detection in cardiac magnetic resonance imaging (grey bars) and positron emission tomography (black bars). In total, cardiac magnetic resonance imaging (CMR) found scars in 45% of all segments compared to PET in 40%. CMR depicted significantly more scars in patients with severely (EF < 30%) and moderately (EF, 30%-50%) impaired left ventricular function. However, PET suggested more scars in EF > 50% group. PET: Positron emission tomography; EF: Ejection fraction.

	Sensitivity	Specificity	PPV	NPV	Accuracy
All patients	69%	71%	62%	77%	70%
EF< 30%	79%	66%	66%	79%	72%
EF 30%-50%	64%	69%	58%	74%	67%
EF > 50%	25%	89%	41%	79%	74%

Detection of myocardial scar (score 2-4) by cardiac magnetic resonance imaging (CMR) as test parameter compared to positron emission tomography (PET). Analysis was done for the whole patient collective as well as in the three patient subgroups with different ejection fraction (EF). In patients with moderately or severely compromised EF CMR outperformed PET. NPV: Negative predictive value; PPV: Positive predictive value.

Viable segments which can be expected to improve after revascularization (score 1 and 2) were detected by CMR in 72% (3949/5508) compared to non-viable segments (score 3 and 4) in 28% (1559/5508). For PET, viability of segments was declared in 80% (4396/5508) and non-viability in 20% (1112/5508). CMR and PET significantly differed in depicting viable and non-viable segments (P < 0.0001). Inter-observer agreement (Cohen’s Kappa) between CMR and PET in functional recovery estimation was 0.48 (moderate). Analysis of the different patient groups Analysis of the different subgroups revealed that CMR judged segments as viable in 66% in patients with severely compromised LV function, in 72% in subjects with moderately reduced LV function and in 91% in patients with non-compromised LV function. For PET, these values were 75%, 80%, and 94% respectively. Comparison of CMR and PET showed that CMR declared significantly less segments as viable than PET in patients with severely or moderately reduced LV function (for all, P < 0.0001). In patients with uncompromised LV function no statistical significance was evident between both modalities after Bonferroni correction (P = 0.03).

Table 3 provides statistical values for CMR in the estimation of functional recovery (MRI score 1 and 2) compared to PET. Comparison of Tables 2 and 3 reveals that overall performance of CMR was better in Table 3, when small scars (< 50% transmurality) were excluded. Figure 2 gives an example of a transmural scar detected both by CMR and PET as well as of small subendocardial scars that were found by CMR but overseen in PET. Figure 3 shows diagrams of the volumetric measures in relation to the total extent of scar (mean scar score).

Open in New Tab   Full Size Figure   Download Figure

Figure 2 A sixty-seven-year-old man with severe coronary artery disease and history of myocardial infarction. A: Short axis inversion-recovery gradient-recalled echo cardiac magnetic resonance imaging (CMR) image of the mid- to apical-portion of the left ventricular shows a small area of transmural late gadolinium enhancement (LGE) in the inferolateral wall (broad arrow). CMR viability scores: Anterior: 2; anterolateral: 2; inferolateral: 4, inferior: 2; B: The positron emission tomography (PET) image of the corresponding slice reveals an uptake defect (broad arrow) in the same segment suggesting a transmural scar. PET viability scores: anterior: 2; inferolateral: 4. Small subendocardial scars with LGE in CMR (A) in the anterolateral and inferior wall (small arrows) were overseen in PET.

	Sensitivity	Specificity	PPV	NPV	Accuracy
All patients	72%	83%	51%	92%	81%
EF < 30%	74%	79%	55%	90%	78%
EF 30%-50%	73%	82%	50%	93%	81%
EF > 50%	30%	92%	19%	95%	89%

The contingency table shows the detection of tissue with potential functional recovery (score 1 and 2) by cardiac magnetic resonance imaging (CMR) as test variable compared to positron emission tomography (PET). Calculation was performed for the whole patient collective as well as separately for the three patient subgroups with different EF. Except for patients with not compromised left ventricular function sensitivity, specificity, PPV, NPV and accuracy were higher in CMR. EF: Ejection fraction; NPV: Negative predictive value; PPV: Positive predictive value.

Open in New Tab   Full Size Figure   Download Figure

Figure 3 Correlation of left ventricular volumes and function to the MRI-derived extent of myocardial scar (summarized as mean magnetic resonance imaging score per patient). A: End-diastolic volumes (EDV); B: End-systolic volumes (ESV); C: Ejection fraction (EF).

Results from myocardial viability testing play an important role in clinical decision making especially in patients with impaired myocardial function, who might require invasive treatment. Albeit several studies reported on patients who benefited from restoration of blood flow despite pre-interventional proof of non-viable tissue and without post-operative functional recovery consequently[11,28,29], general consensus exists that information on the transmural extent of myocardial scar is of importance because it holds a close relationship with recovery of segmental function at follow-up[3]. We, therefore, sought to evaluate the diagnostic accuracy of LGE-CMR viability assessment and PET in patient groups with different LV functions. The main findings were: (1) in subjects with severely or moderately reduced LV function, CMR detects considerably more myocardial scars than PET; (2) scars, which are only seen in PET in patients with non-compromised LV function are probably false-positive results or artifacts; and (3) in patients with impaired LV function (EF < 50%), CMR demonstrates more non-viable myocardium compared to PET and is generally less optimistic concerning functional recovery after revascularization procedures.

Within the last decade CMR has more and more replaced nuclear imaging techniques in myocardial viability assessment. In particular, CMR seems to be useful in identifying patients, who most likely will not benefit from coronary revascularization[11]. Kühl et al[11] demonstrated that none of the segments which were classified as viable by PET/SPECT and nonviable by CMR showed functional recovery 6 mo after revascularization procedures, while 42% of segments described as viable by CMR and non-viable by PET/SPECT still improved[7].

To our knowledge, the present study is the first to compare CMR and PET differentiated in groups depending on LVEF. In patients with severely or moderately reduced LVEF, our study revealed a high sensitivity, specificity and NPV, indicating that CMR is good in detecting myocardial scars. Especially, the high NPV is of great importance in decision-making: if CMR indicates non-viable tissue, low likelihood for functional improvement after revascularization would be expected, which might prevent patients from unnecessary invasive procedures and potential peri-interventional risks. Moreover, CMR demonstrated considerably more nonviable myocardial tissue than PET with better sensitivity, PPV, NPV and accuracy (excluding small scars with < 50% transmural extent, score 2; Tables 2 and 3). One reason for that most probably is the significantly higher spatial resolution of CMR compared to PET, which is of benefit in finding subendocardial scars and in analyzing the thinned myocardial walls in subjects with severely reduced LVEF. Another explanation might be related to the fact that FDG-uptake represents viability, so that small amounts of viable cells lead to visible FDG-uptake indicating viability[11,29,30], although structural changes may already be present (e.g., expansion of extracellular spaces), leading to altered gadolinium kinetics. In addition to that, PET evaluates myocardial viability semiquantitatively: FDG-uptake in a given segment is expressed relative to the segment with maximum FDG-uptake. As a consequence, in thinned myocardium already small rims of reduced FDG-uptake may decrease relative percentage of FDG-uptake to below the threshold value set for viability, albeit viable tissue exists[11], resulting in false-negative evaluation of myocardial viability.

PET detected more scars in subjects with noncompromised LVEF (> 50%), similar to what Klein et al[30] described in their study. These surplus segments seen in PET only are most likely false-positive results or artifacts: As LVEF is only marginally or not impaired, only small amounts of non-viable subendocardial tissue are expected. Larger scars would have had more impact on myocardial function. However, a considerable number of non-viable cells is needed for detectable reduction of the relative FDG-uptake below the threshold-value considered for viability, which seems less probable in small scars[31-33]. Owing to the lower spatial resolution of PET it appears unlikely that these small scars were depicted by PET with higher sensitivity than by CMR[3,17]. This is further underlined by studies describing that more than half of the small subendocardial scars depicted by CMR cannot be delineated in PET[30]. Moreover, a minimum of 2 g irreversibly injured myocardium can be detected by LGE-CMR compared to a minimum tissue of 10 g required in PET[33]. Because of that recent studies have denominated LGE-CMR as method of choice for small subendocardial unrecognized myocardial infarction scars[34,35]. Other shortcomings of PET are its radiation exposure, the long examination time and the allocation of appropriate tracers[11]. CMR has evolved as a valuable alternative to PET for evaluation of viable and infarcted myocardium by different techniques (morphology, edema, function, perfusion and scar) in a single examination[2]. Estimated examination time for a complete CMR work-up is 30-60 min. Whether the recently emerging combination of PET and CMR (PET/CMR-hybrid) might be an alternative to CMR alone in assessment of myocardial viability will be the task of future studies[36,37].

We are aware of the following limitations of our study: First, the possibility of anatomical misalignment between different imaging modalities cannot be excluded. Within the last years, hybrid PET/CT-systems had increasingly replaced single PET-scanners. However, as published in a review by Anagnostopoulos et al[38] the impact of PET/CT-imaging on clinical outcomes in patients suffering from CAD is still unclear. They therefore recommended that until further studies are performed anatomical or functional imaging should be done sequentially by cardiac CT or PET depending on the pre-test probability of CAD[38]. In patients with higher probability PET should be the first line modality as its ability to guide patient management decisions regarding revascularization or medical treatment has been shown[39]. Therefore, we think that our results can be regarded representative, even though we are aware of the limitations of comparing semi-quantitative assessment of radiotracer uptake in PET as comparator for evaluating LGE-CMR. Furthermore, we defined myocardial scar as area with decreased FDG-uptake, even though reduced FDG-uptake might also be caused by stunned myocardium. The most appropriate approach would have been to perform both a functional and perfusion scan, in which stunned myocardium would have been detected as area with reduced FDP-uptake but normal perfusion (perfusion-metabolism reverse mismatch)[33]. However, for the given clinical indication in our patients the functional scan was sufficient and therefore no perfusion data was available. Another technical limitation might be that we did not use a glucose-insulin clamp to standardize the glucose metabolism within the whole myocardium. Like others[36,37], in our institution PET imaging is done under fasting condition and oral administration of two doses of acipimox and 75 g glucose prior to FDG application. Therefore, we believe the number of false-negative segments to be negligible. However, the lack of clinical approval of acipimox in the United States hampers representativeness of our data abroad. Moreover, no correction technique was applied for PET. And finally, a detailed segment-to-segment attenuation comparison was only partly done in our study, but would have been of interest to get deeper information on exact differences between PET and CMR regarding scar detection, e.g., considering the location of scar keeping in mind that the exact delineation of the inferior LV segments is often impaired in PET[30,40]. This has to be investigated by further studies.

In conclusion, our study demonstrates differences in diagnostic accuracy of CMR and PET differentiating between patient groups according to LV ejection fraction. Advantages of CMR compared to PET were found in detecting scars and non-viable tissue in subjects with severely or moderately reduced LVEF. CMR is generally less optimistic concerning functional recovery after revascularization procedures, which is of great importance in clinical decision-making.