Author + information
- Received June 23, 2009
- Revision received November 2, 2009
- Accepted November 6, 2009
- Published online April 1, 2010.
- Grigorios Korosoglou, MD⁎,⁎ (, )
- Stephanie Lehrke, MD⁎,
- Angela Wochele, RN⁎,
- Birgit Hoerig, RN⁎,
- Dirk Lossnitzer, MD⁎,
- Henning Steen, MD⁎,
- Evangelos Giannitsis, MD⁎,
- Nael F. Osman, PhD†,‡ and
- Hugo A. Katus, MD⁎
- ↵⁎Reprint requests and correspondence:
Dr. Grigorios Korosoglou, University of Heidelberg, Department of Cardiology, Im Neuenheimer Feld 410, Heidelberg 69120, Germany
Objectives This study sought to evaluate the diagnostic accuracy of strain-encoded cardiac magnetic resonance (SENC) for the detection of inducible ischemia during intermediate stress.
Background High-dose dobutamine stress cardiac magnetic resonance (DS-CMR) is a well-established modality for the noninvasive detection of coronary artery disease (CAD). However, the assessment of cine scans relies on the visual interpretation of wall motion, which is subjective, and modalities that can objectively and quantitatively assess the time course of myocardial strain response during stress are lacking.
Methods Stress-induced ischemia was assessed by wall motion analysis and by SENC in 80 patients with suspected or known CAD and in 18 healthy volunteers who underwent DS-CMR in a clinical 1.5-T scanner. Quantitative coronary angiography was used as the standard reference for the presence of CAD (≥50% diameter stenosis).
Results On a patient level, 46 of 80 patients (58%) had CAD, including 20 with single-vessel, 18 with 2-vessel, and 8 with 3-vessel disease. During peak stress, SENC correctly detected ischemia in 45 versus 38 of 46 patients with CAD (7 additional correct findings for SENC), yielding significantly higher sensitivity than cine (98% vs. 83%, p < 0.05). No patients were correctly diagnosed by cine and missed by SENC. During intermediate stress, SENC showed diagnostic value similar to that provided by cine imaging only during peak dobutamine stress (sensitivity of 76% vs. 83%, specificity of 88% vs. 91%, and accuracy of 81% vs. 86%; p = NS for all). Quantification analysis demonstrated that strain rate response is a highly sensitive marker for the detection of inducible ischemia (area under the curve = 0.96; SE = 0.01; 95% confidence interval: 0.93 to 0.99) that precedes the development of inducible wall motion abnormalities and already significantly decreases with moderate 40% to 60% coronary lesions.
Conclusions Using SENC, CAD can be detected during intermediate stress with similar accuracy to that provided by cine only during peak stress. By this approach, patient safety may be improved during diagnostic procedures within lower time spent (Strain-Encoded Cardiac Magnetic Resonance Imaging for Dobutamine Stress Testing; NCT00758654
- myocardial strain reserve
- strain-encoded cardiac magnetic resonance
- intermediate stress
- inducible ischemia
High-dose dobutamine stress cardiac cardiac magnetic resonance (DS-CMR) is a well-established diagnostic modality for the noninvasive detection of coronary artery disease (CAD). However, the assessment of cine images relies on the visual interpretation of regional wall motion, which depends on the experience of the readers. Especially with nonexpert magnetic resonance readers, the human eye primarily tracks the radial displacement of the myocardium with cine images, which is less sensitive than circumferential and longitudinal components for the detection of myocardial dysfunction (1–4).
Objective approaches for the quantification of myocardial strain during DS-CMR have been very limited so far. Strain-encoded cardiac magnetic resonance (SENC) has been previously proposed for the objective color-coded evaluation of regional myocardial strain in experimental and in clinical settings (5–7). The purpose of the present study was to investigate the ability of SENC to detect myocardial ischemia during intermediate stress in a cohort of subjects with suspected or known CAD. The results were compared with cine images, and quantitative coronary angiography was deemed as the standard reference for the presence of anatomically significant CAD.
Consecutive patients with suspected or known CAD (n = 131) were screened for inclusion in our study before clinically indicated coronary angiography. Patients were excluded for the following reasons: nonsinus rhythm (n = 5), electrocardiography signs or a history of previous myocardial infarction (n = 17), regional resting wall motion abnormalities (WMAs) or ejection fraction <55% (n = 8), severe arterial hypertension (>200/120 mm Hg) (n = 2), moderate or severe valvular disease (n = 2), or general contraindications to magnetic resonance examination (n = 8). Thus 89 patients were scheduled for DS-CMR. Age-matched healthy volunteers (n = 18) also underwent DS-CMR to acquire normal values for myocardial strain and strain rate response. All volunteers underwent laboratory testing before inclusion in our study. Exclusion criteria were history, symptoms, electrocardiographic signs, or biochemical findings indicative of cardiovascular disease, evidence of systemic hypertension (baseline blood pressure >140/85 mm Hg), diabetes, abnormal glucose tolerance, hyperlipidemia (low-density lipoprotein >130 mg/dl) or the presence of WMA at baseline or during DS-CMR. All procedures complied with the Declaration of Helsinki and were approved by our local ethics committee, and all patients gave written informed consent.
Cardiovascular magnetic resonance examination
Subjects were examined in a clinical 1.5-T whole-body Achieva system (Philips Medical Systems, Best, the Netherlands) using a 5-element cardiac phased-array receiver coil. Images were acquired at rest and during a standardized high-dose dobutamine/atropine protocol (6). A 4, 2, and 3-chamber and 3 short-axis views (apical, mid-ventricular, and basal) were used. Dobutamine was infused during 3-min stages at incremental doses of 10, 20, 30, and 40 μg/kg of body weight per min until at least 85% of the age-predicted heart rate was reached (220-age). If at peak infusion the target heart rate was not achieved, atropine was administrated in 0.25-mg increments up to a maximal dose of 2.0 mg. Stress testing was discontinued when the target heart rate was achieved or when one of the following occurred: extensive WMA in ≥2 adjacent segments, severe chest pain or dyspnea, decrease in systolic blood pressure of ≥40 mm Hg, arterial hypertension of ≥220/120 mm Hg, or severe arrhythmias.
In the absence of ischemia, failure to attain 85% of age-predicted maximal heart rate was considered as a nondiagnostic result.
A steady-state free-precession sequence was used to obtain cine images with 8-mm slice thickness. Typical parameters were sensitivity encoding factor of 2, field of view = 350 × 350 mm2, matrix size = 160 × 160, flip angle = 60°, repetition time/echo time = 2.8/1.4 ms, acquired voxel size = 2.2 × 2.2 × 8 mm3, image matrix = 288 × 288, and reconstructed voxel size = 1.2 × 1.2 × 8 mm3. The temporal resolution was 21 to 28 ms, and the total scan duration was 7 to 12 s. Cine images were acquired at baseline, and acquisitions were repeated during each stage, including the peak level of dobutamine stress.
The SENC pulse sequence is a modified spatial modulation of magnetization tagging pulse sequence, which provides the color-encoded visualization and quantification of myocardial strain. Technical details of this sequence are described elsewhere (5–7). SENC images were acquired at identical plane levels of that used for cine scans with 10-mm slice thickness. Typical parameters were field of view = 350 × 350 mm2, matrix size = 80 × 80, flip angle = 30°, repetition time/echo time = 25/0.9 ms, acquired voxel size = 4.4 × 4.4 × 10 mm3, image matrix = 256 × 256, and reconstructed voxel size = 1.4 × 1.4 × 8 mm3. Strain-encoded images were acquired at baseline, during 20 μg/kg of dobutamine infusion, during peak dobutamine/atropine administration, and 10 min after stress testing. Details on the temporal resolution and total scan duration are provided in Online Table 1.
Visual interpretation of cine and SENC images
For interpretation of wall motion, corresponding rest and peak stress cine images were displayed using View Forum software (Philips Medical Systems, Best, the Netherlands). Segmental wall motion was graded semiquantitatively using a 16-segment model according to American Heart Association guidelines (8) and a 3-point scale (0 = normal wall motion, 1 = hypokinesia, 2 = akinesia or dyskinesia), and inducible ischemia was considered present in cases of new WMA of ≥1 grade during stress (6).
Corresponding baseline and peak stress SENC images were displayed using Diagnosoft SENC (version 1.06, Diagnosoft, Palo Alto, California), a software package that allows the color-encoded interpretation of myocardial strain on SENC images. Similar to wall motion analysis, myocardial strain was graded semiquantitatively using a 3-point color scale (0 = normal strain corresponding to red myocardium, 1 = reduced strain corresponding to faded orange/yellowish myocardium, 2 = severely reduced or absent strain corresponding to white myocardial tissue on strain-encoded images). Inducible ischemia was considered present in case of strain reduction of ≥1 grade. Both cine and SENC images were interpreted visually by 2 independent observers (G.K. and D.L.), who evaluated images off-line after the termination of the stress studies. The evaluation was performed separately, and observers were blinded to all other data.
Quantitative analysis of circumferential and longitudinal strain with SENC images
Because the tagging modulation gradient is applied in the slice selection direction with SENC, quantification of circumferential strain was performed on 4-, 2-, and 3-chamber view, whereas quantification of longitudinal strain was performed on short-axis images. For each segment, the temporal course of regional myocardial strain was registered throughout the cardiac cycle, and quadratic interpolation was used to estimate the following parameters: 1) peak systolic strain (S), defined as the maximum strain during the cardiac cycle and expressed as a percentage; 2) peak systolic strain rate (SR), defined as the maximum strain rate during systole and expressed in 1/s; 3) strain reserve, defined as the relative increase in peak systolic strain during DS-CMR and calculated as follows: SReserve = Sdobutamine/Sbaseline; and 4) strain rate reserve, defined as the relative increase in peak systolic strain rate during DS-CMR and calculated as follows: SRReserve= SRdobutamine/SRbaseline.
Quantitative coronary angiography and comparison to cine and SENC images
Angiography was deemed as the standard reference for the detection of CAD and was obtained in all patients within 3 weeks from the DS-CMR study. The procedure was performed according to the angiographic guidelines, and at least 2 orthogonal views of every major coronary vessel and its side branches were acquired. Quantification of lumen narrowing was performed off-line using Centricity QCA (GE Medical Systems, Milwaukee, Wisconsin). Myocardial segments were assigned to coronary vessels according to American Heart Association guidelines (8) (see Online Appendix for details).
Analysis was performed using commercially available software (SPSS, version 12.0 for Windows, SPSS, Chicago, Illinois), and data are presented as mean ± SD. Agreement between the 2 observers interpreting cine and SENC was assessed using kappa statistics (9). Intra- and interobserver variability for quantification of strain were calculated by repeated analysis of 40 representative images. Differences in sensitivity, specificity, and accuracy were tested using exact 2-sided McNemar tests (10). Differences in strain and strain rate between normal, ischemic, and nonischemic segments at different time points and differences by stenosis severity (0%; 1% to 0%; 21% to 40%; 41% to 60%; 61% to 80%; and 81% to 100%) were assessed with clustered regression and using Bonferroni correction for multiple comparisons. The relation between SReserve and SRReserve with the stenosis severity was assessed by second-order polynomial regression analysis. Receiver-operator characteristics were used to estimate the accuracy of SReserve and SRReserve to predict ≥50% stenosis, and pairwise comparisons of areas under the curve were assessed (11). Differences were considered significant at p < 0.05.
Demographic and hemodynamic parameters
Diagnostic DS-CMR examinations (positive for ischemia or negative but with achievement of the target heart rate) were achieved in 80 of 89 patients. Nine patients were excluded from analysis for the following reasons: failure to achieve target heart rate in the absence of ischemia (n = 3), increase in systolic blood pressure >220 mm Hg (n = 1), discontinuation of the study on patient's request (n = 2), or repeated extrasystoles during stress, resulting in nondiagnostic image quality both for cine and SENC images (n = 3). In 80 patients with regular rhythm during baseline and stress, 17 segments were not interpretable either at baseline (n = 7; 0.5%) or during stress (n = 10; 0.8%) with SENC images. In all 18 healthy subjects, the target heart rate was achieved and diagnostic images were acquired. Overall, no severe adverse events were recorded. Demographics are summarized in Table 1.
Results of coronary angiography
Coronary angiography showed ≥50% stenosis in 80 of 240 coronary vessels (33%), including 32 vessels with left anterior descending, 27 with left circumflex, and 21 with right coronary artery lesions. On a patient level, 46 of 80 patients (58%) had CAD, including 20 with single-vessel, 18 with 2-vessel, and 8 with 3-vessel disease.
Detection of CAD by SENC versus cine imaging
During intermediate stress, SENC already detected inducible ischemia in 35 of 46 and correctly excluded CAD in 30 of 34 patients with normal strain response (sensitivity of 76%, specificity of 88%, and accuracy of 81%), whereas cine imaging showed low diagnostic value (sensitivity of 20%, specificity of 97%, and accuracy of 53%, p < 0.001 vs. SENC for sensitivity and accuracy). The diagnostic characteristics of SENC during intermediate stages were statistically similar to those provided by cine imaging only during peak dobutamine stress (Table 2). During peak stress, SENC correctly detected ischemia in 45 versus 38 of 46 patients with CAD (7 additional correct findings for SENC), yielding significantly higher sensitivity than cine (98% vs. 83%, p < 0.05). No patients were correctly diagnosed by cine and missed by SENC.
In a patient without CAD, circumferential peak systolic strain (Figs. 1A to 1F) remained constant, as indicated by black and green curved arrows in Figure 1G, whereas peak systolic strain rate increased stepwise, as indicated by black and green curved arrows in Figure 1H. Conversely, in a patient with CAD (Figs. 2A to 2H), SENC revealed a subtle, albeit clearly detectable strain defect in the anterior left ventricular (LV) wall (white arrow in Fig. 2D) already during intermediate stress, which increased during peak stress (white arrows in Fig. 2F). With cine images, ischemia was detected only during peak stress (white arrow in Fig. 2E). Quantitative analysis showed that longitudinal peak systolic strain decreased stepwise during stress, as indicated by black and green curved arrows in Figure 2I, whereas peak systolic strain rate remained unchanged, as indicated by the dotted red circle in Figure 2J. The presence of a 68% diameter stenosis in the left anterior descending artery was confirmed by angiography (Figs. 2K and 2L).
Quantitative assessment of circumferential and longitudinal strain response during DS-CMR
In normal and nonischemic segments, circumferential and longitudinal strain remained constant (hatched and dotted lines in Figs. 3A and 3B), whereas strain rate increased stepwise (hatched and dotted lines in Figs. 3C and 3D). Conversely, in ischemic segments, strain decreased stepwise (solid lines in Figs. 3A and 3B), whereas strain rate remained constant (solid lines in Figs. 3C and 3D) (p < 0.001 vs. nonischemic and normal).
During both intermediate and peak stimulation, significant correlations were observed between stenosis severity and circumferential SReserve and SRReserve (p < 0.001 for all) (Figs. 4A,4C, 4E, and 4G). Interestingly, segments with new WMA during peak stress on cine images were located at the bottom right corner (pink dots within the hatched red circles in Figs. 4E and 4G), whereas a considerable amount of segments, classified as “normal” by wall motion readings, had impaired SReserve and SRreserve (blue dots within the dotted red squares in Figs. 4E and 4G). SReserve decreased with >80% and >60% coronary lesions, respectively, during intermediate and peak stress, whereas SRReserve already decreased with >60% and with >40% lesions, respectively (i.e., 1 stage earlier) (Figs. 4B, 4D, 4F, and 4H). Similar findings were obtained for longitudinal strain (Online Fig. 1).
SRReserve provided significantly higher accuracy than strain for the detection of coronary lesions ≥50% than SReserve (green and pink arrows in Fig. 5, indicating significant increase in accuracy between strain and strain rate with intermediate and peak stress, respectively). Hereby, the highest accuracy was acquired by SRReserve during peak stress.
Observer agreement and variabilities
Agreement between observers interpreting wall motion on cine images and myocardial strain on SENC images was 87% (kappa = 0.78) and 84% (kappa = 0.75), respectively. SENC allowed for reproducible quantification of strain and strain rate, showing intra- and interobserver variabilities of 7.4% and 10.2% for strain and 8.7% and 11.1% for strain rate.
The results of our study demonstrate for the first time in the current literature the ability of a strain imaging technique to detect CAD during intermediate stages of inotropic stimulation with similar accuracy to that provided by conventional wall motion analysis only during peak stress. SENC can objectively assess the time course of regional myocardial strain and strain rate response during inotropic stimulation and provide a sensitive, noninvasive estimate of coronary stenosis severity.
Previous echocardiographic studies used tissue Doppler techniques to evaluate strain response during stress (12). However, to our knowledge no echocardiographic or magnetic resonance imaging studies have investigated thus far the ability of strain imaging to detect ischemia during intermediate stages. Recently, we demonstrated that the direct color-coded visualization of myocardial strain with SENC allows for the objective detection of subtle differences in regional myocardial strain (6). In the present study, we show for the first time that the reversible decrease of regional myocardial strain on SENC images precedes the development of WMAs, aiding the detection of CAD during intermediate stages of inotropic stimulation with sensitivity and accuracy, similar to that provided by cine imaging during peak stress. Thus the implementation of SENC in the clinical routine may help with the early detection of strain defects with intermediate stages, increasing patient safety during diagnostic procedures and simultaneously reducing time spent. Furthermore, newer, fast SENC techniques were recently shown to allow for quantification of myocardial strain within a single heartbeat, obviating the need of prolonged breathholds (5). Because cardiac abnormalities during stress procedures are of a transient nature, single heartbeat imaging might be able to detect wall motion disorders, which conventional 10-heartbeat or more segmented sequences might miss due to temporal averaging.
In the present study, quantitative analysis yielded different temporal patterns of strain and strain rate response between ischemic and nonischemic myocardium. Thus myocardial strain rate increased stepwise during stress (SRReserve ∼2) in nonischemic and remained constant in ischemic segments. On the other hand, strain remained constant in nonischemic segments (SReserve ∼1) and decreased stepwise during intermediate and peak stress in ischemic segments. Conversely, using magnetic resonance tagging, circumferential strain was previously shown to slightly increase (by 10% to 20%) during low-dose inotropic stimulation in normal segments (13). This mismatch may be attributed to differences in methodology used for the acquisition of SENC images, where in contrast to conventional tagging, the gradient is applied in the slice-selection direction, orthogonal to the imaging plane and not in the phase- or frequency-encoding direction (14). Furthermore, conventional SENC techniques fail to compensate for the through-plane motion of the heart (the motion in the slice-selection direction). For example, in long-axis images of the LV, long-axis shortening is observed as the base moves toward the apex during systole. This displacement can be ∼2 cm at baseline and increases during inotropic stimulation. However, because the imaged slice is fixed with respect to the magnet co-ordinate system, the images acquired at different heart phases do not always represent the same piece of myocardium. This effect may cause tissue misregistration when different SENC images are combined to create cardiac strain images. Because the displacement is higher during peak stress, the absence of through-plane motion correction is expected to underestimate strain values to a higher extent during stress, resulting in an underestimation of the calculated strain reserves. Through-plane motion tracking is, however, now available with new implementations of the SENC pulse sequence (15), which may allow for more accurate estimation of strain in future studies.
Emphasizing pathophysiologic aspects between myocardial strain response and lumen narrowing, the decrease of strain rate in our study was shown to precede that of myocardial strain in the ischemic cascade. Thus, SRReserve already decreased with intermediate (40% to 60%) coronary lesions during peak stress, representing a highly sensitive surrogate marker for regional ischemia. In a similar setting and using myocardial perfusion imaging, Cullen et al. (16) previously demonstrated that regions with 40% to 60% stenosis yield lower myocardial perfusion reserve compared with those with <40% stenosis, whereas even regions supplied by <40% coronary lesions have decreased myocardial perfusion reserve as compared with normal volunteers. Because the versatility of magnetic resonance allows for assessment of perfusion, wall motion, and myocardial systolic and diastolic strain within a single diagnostic stress test, and all without X-ray exposure for the patient, the accuracy of each of these separate markers for the detection of ischemia remains to be compared in future studies (Fig. 6).
Our study has some limitations. Patients with previous infarction were excluded in order to avoid potential problems in the detection of inducible ischemia resulting from hibernating or stunned myocardium. This limits the extrapolation of our findings to such patients. Although the total scanning time was similar for cine and SENC sequences, cine images generally had higher spatial and lower temporal resolution compared with SENC, which is a limitation. This may have accounted for the slightly lower specificity of SENC to detect inducible ischemia compared with cine imaging. In this regard, recent studies have used higher spatial resolution for cine imaging (17), which may further aid the detection of ischemia during stress. Furthermore, SENC is based on the stimulated echo acquisition mode technique, which inherently has a low signal-to-noise ratio. This makes SENC more susceptible to noise, which may cause the presence of “signal voids” in the reconstructed images (yellow arrowheads in Figs. 1 and 2). However, our initial observation is that the strain measurements are more robust against noise, even in regions darker than the rest of the myocardium. Thus, even in regions with such signal voids, the estimation of regional myocardial strain is feasible. Generally, increased noise with SENC images is expected to decrease the range but not the absolute values of estimated strain (which would in the vivo setting give the wrong the impression of inducible ischemia). Furthermore, observer agreement was slightly lower for SENC compared with cine imaging. This may be attributed to the extensive experience of the readers with wall motion readings during stress, so that SENC may have more potential for improving observer variability in less trained readers, where differences in color might be easier to interpret than the assessment of wall motion on cine images. In addition, relatively high variability in strain and strain rate was observed in segments without coronary stenosis. This may be attributed to: 1) regional heterogeneity in baseline (18) and recruitable myocardial strain (19) between different LV segments due to variations in transmural fiber orientation and local differences in ventricular morphology (20); and 2) to technical limitations with the current implementation of the SENC sequence in terms of temporal and particularly spatial resolution. In this regard, technically improved SENC sequences may reduce such variability and allow for more accurate quantification of regional strain in future studies. Because the color-encoded visualization of myocardial strain is not feasible with the View Forum (Philips Medical Systems) software, currently no clinical decisions can be made during stress testing based on SENC images. In future studies, however, the implementation of SENC post-processing in the routine workflow may have clinical potential if the results are available within a few seconds after image acquisition. Finally, although the assessment of early diastolic strain may offer enhanced sensitivity for the detection of ischemia during stress testing (21), in our study, we focused on the calculation of systolic strain, and diastolic strain values were not systematically measured, which is a limitation.
This is, to our knowledge, the first study to demonstrate that quantification of strain can detect inducible ischemia during intermediate stages of inotropic stimulation with similar accuracy to that provided by wall motion analysis during high-dose stress. Particularly, the quantification of strain rate response is a highly sensitive marker for stress-induced ischemia, which precedes the development of WMA and provides an excellent noninvasive measure of coronary stenosis severity.
For an expanded Methods section, supplementary table, and supplementary figure, please see the online version of this article.
Strain-Encoded Cardiac Magnetic Resonance for the Detection of Inducible Ischemia During Intermediate Stress
Dr. Osman is a founder and shareholder in Diagnosoft Inc., the software used for the analysis of the acquired SENC images.
- Abbreviations and Acronyms
- coronary artery disease
- dobutamine stress cardiac magnetic resonance
- left ventricular
- peak systolic strain
- strain-encoded cardiac magnetic resonance
- peak systolic strain rate
- wall motion abnormality
- Received June 23, 2009.
- Revision received November 2, 2009.
- Accepted November 6, 2009.
- American College of Cardiology Foundation
- Reant P.,
- Labrousse L.,
- Lafitte S.,
- et al.
- Chan J.,
- Hanekom L.,
- Wong C.,
- Leano R.,
- Cho G.Y.,
- Marwick T.H.
- Tanaka H.,
- Oishi Y.,
- Mizuguchi Y.,
- et al.
- Korosoglou G.,
- Lossnitzer D.,
- Schellberg D.,
- et al.
- Cerqueira M.D.,
- Weissman N.J.,
- Dilsizian V.,
- et al.
- Zweig M.H.,
- Campbell G.
- Voigt J.U.,
- Exner B.,
- Schmiedehausen K.,
- et al.
- Cullen J.H.,
- Horsfield M.A.,
- Reek C.R.,
- Cherryman G.R.,
- Barnett D.B.,
- Samani N.J.
- Garot J.,
- Bluemke D.A.,
- Osman N.F.,
- et al.
- Bogaert J.,
- Rademakers F.E.
- Ishii K.,
- Imai M.,
- Suyama T.,
- et al.