Author + information
- Received November 25, 2008
- Revision received April 17, 2009
- Accepted May 4, 2009
- Published online September 29, 2009.
- Matteo Bertini, MD*,†,
- Nina Ajmone Marsan, MD*,
- Victoria Delgado, MD*,
- Rutger J. van Bommel, MD*,
- Gaetano Nucifora, MD*,
- C. Jan Willem Borleffs, MD*,
- Giuseppe Boriani, MD, PhD†,
- Mauro Biffi, MD†,
- Eduard R. Holman, MD, PhD*,
- Ernst E. van der Wall, MD, PhD*,‡,
- Martin J. Schalij, MD, PhD* and
- Jeroen J. Bax, MD, PhD*,* ()
- ↵*Reprint requests and correspondence:
Dr. Jeroen J. Bax, Department of Cardiology, Leiden University Medical Center, Albinusdreef 2, 2333 ZA Leiden, the Netherlands
Objectives This study explored the effects of cardiac resynchronization therapy (CRT) on left ventricular (LV) twist, particularly in relation to LV lead position.
Background LV twist is emerging as a comprehensive index of LV function.
Methods Eighty heart failure patients were included. Two-dimensional echocardiography was performed at baseline, immediately after CRT, and at 6-month follow-up. Speckle-tracking analysis was applied to assess LV twist. The LV lead was placed preferably in a (postero)lateral vein, and at fluoroscopy, the position was classified as basal, midventricular, or apical. Response to CRT was defined as reduction of LV end-systolic volume ≥15% at 6-month follow-up. A control group comprised 30 normal subjects.
Results Peak LV twist in heart failure patients was 4.8 ± 2.6° compared with 15.0 ± 3.6° in the control subjects (p < 0.001). At 6-month follow-up, peak LV twist significantly improved only in responders (56%), from 4.3 ± 2.4° to 8.5 ± 3.2° (p < 0.001). The strongest predictor of response to CRT was the improvement of peak LV twist immediately after CRT (odds ratio: 1.899, 95% confidence interval: 1.334 to 2.703, p < 0.001). Furthermore, LV twist significantly improved in patients with an apical (from 4.3 ± 3.1° to 8.6 ± 3.0°, p = 0.001) and midventricular (from 4.8 ± 2.2° to 6.4 ± 3.9°, p = 0.038) but not with a basal (5.0 ± 3.3° vs. 4.1 ± 3.2°, p = 0.28) LV lead position. Similarly, LV ejection fraction significantly increased in patients with an apical (from 26 ± 7% to 37 ± 7%, p < 0.001) and midventricular (from 26 ± 6% to 33 ± 8%, p < 0.001) but not with a basal (26 ± 5% vs. 28 ± 8%, p = 0.30) LV lead position.
Conclusions An immediate improvement of LV twist after CRT predicts LV reverse remodeling at 6-month follow-up.
- heart failure
- cardiac resynchronization therapy
- left ventricular twist
- left ventricular reverse remodeling
- left ventricular lead position
The human heart has a specific helical arrangement of the myofibers with a right-hand orientation from the base toward the apex in the endocardial layers and a left-hand orientation in the epicardial layers. This spiral architecture of the myofibers leads to a left ventricular (LV) systolic wringing motion as a result of an opposite rotation of LV apex and base (1,2). The gradient between apex and base in the rotation angle along LV longitudinal axis is called twist and contributes significantly to LV systolic function, in addition to myocardial shortening and thickening (3–5).
In heart failure (HF) patients, LV twist is significantly reduced (6). Cardiac resynchronization therapy (CRT) is considered a major therapeutic breakthrough for HF patients, and recent large randomized trials have shown that CRT has beneficial effects on HF symptoms, LV systolic function, and survival (7,8). At present, minimal data are available about the effect of CRT on LV twist (9,10).
In the current study, the effect of CRT on LV twist was assessed using speckle-tracking echocardiography. Furthermore, the relationship between the change in LV twist and LV reverse remodeling at 6-month follow-up was investigated. Finally, the influence of the LV lead position on the improvement in LV twist and response to CRT was explored.
Study population and protocol
A total of 87 consecutive HF patients scheduled for CRT were prospectively included. According to current guidelines, the inclusion criteria were New York Heart Association (NYHA) functional class III to IV, sinus rhythm, left ventricular ejection fraction (LVEF) ≤35%, and QRS duration ≥120 ms (11). Etiology of HF was considered ischemic in the presence of significant coronary artery disease (>50% stenosis in ≥1 major epicardial coronary artery) on coronary angiography and/or a history of myocardial infarction or revascularization.
The clinical evaluation consisted of: 1) assessment of clinical status: NYHA functional class, quality of life (using the Minnesota Living With Heart Failure questionnaire) (12), and 6-min walk distance (13) at baseline and 6-month follow-up; and 2) assessment of LV volumes, function, dyssynchrony, and twist, using standard echocardiography and speckle-tracking analysis at baseline, within 48 h (immediately after CRT) and at 6-month follow-up.
In addition, 30 subjects without evidence of structural heart disease, frequency matched for age, sex, and body surface area, were included as a normal control group, selected from an echocardiographic database. These subjects were referred for the echocardiographic evaluation because of atypical chest pain, palpitations, or syncope without murmur.
All patients were imaged in the left lateral decubitus position using a commercially available system (Vingmed Vivid 7, General Electric-Vingmed, Milwaukee, Wisconsin). Standard 2-dimensional images were obtained using a 3.5-MHz transducer and digitally stored in cine-loop format; the analysis was performed offline using EchoPAC version 6.0.1 (General Electric-Vingmed).
From the standard apical views (4- and 2-chamber), LV volumes and LVEF were calculated according to the American Society of Echocardiography guidelines (14). At 6-month follow-up, patients were classified as echocardiographic responders based on a reduction ≥15% of left ventricular end-systolic volume (LVESV) (15).
Segmental wall motion was assessed according to the American Society of Echocardiography in order to evaluate the presence of scarred segments within ischemic HF patients (14). Akinetic and diskinetic segments (wall motion score 3 and 4) were classified as scarred segments (16).
The speckle-tracking software tracks frame-to-frame the movement of natural myocardial acoustic markers, or speckles, on standard gray-scale images. Speckles are randomly distributed and each region of the myocardium has a distinguishing pattern, a fingerprint. Furthermore, speckle-tracking analysis is angle independent and allows the evaluation of myocardial contraction/relaxation along the circumferential, longitudinal, and radial direction (17,18).
In the current study, speckle-tracking analysis was applied to evaluate LV dyssynchrony (based on radial strain analysis) and LV twist. Parasternal short-axis images were acquired at 3 distinct levels: 1) basal level, identified by the mitral valve; 2) papillary muscle level; and 3) apical level (the smallest cavity achievable distally to the papillary muscles, moving the probe down and slightly laterally, if needed). Frame rate ranged from 45 to 100 frame/s, and 3 cardiac cycles for each parasternal short-axis level were stored in cine-loop format for the offline analysis (EchoPAC, General Electric-Vingmed). The endocardial border was traced at an end-systolic frame, and the region of interest was chosen to fit the whole myocardium. The software allows the operator to check and validate the tracking quality and to adjust the endocardial border or modify the width of the region of interest, if needed. Furthermore, each short-axis image was automatically divided into 6 standard segments: septal, anteroseptal, anterior, lateral, posterior, and inferior. Aortic valve opening and closure were identified on pulsed-wave Doppler tracings obtained from the LV outflow tract.
LV Dyssynchrony Analysis
LV dyssynchrony was derived from the radial strain curves obtained from the papillary muscle short-axis view. As previously described, LV dyssynchrony was defined as the time difference of peak radial strain between the anteroseptal and posterior segments (19).
LV Twist Analysis
The speckle-tracking software calculates LV rotation from the apical and basal short-axis images as the average angular displacement of the 6 standard segments referring to the ventricular centroid, frame by frame. Counterclockwise rotation was marked as positive value and clockwise rotation as negative value when viewed from the LV apex. LV twist was defined as the net difference (in degrees) of apical and basal rotation at isochronal time points. For the calculation of LV twist, averaged apical and basal rotation data were exported to a spreadsheet program (Excel 2003, Microsoft Corp., Redmond, Washington) (Fig. 1)(20,21). The following measurements were derived: peak apical and basal rotation and peak LV twist.
Reproducibility of left ventricular end-diastolic volume (LVEDV), LVESV, LVEF, and peak LV twist was assessed on 20 randomly selected HF patients. Bland-Altman analysis was performed to evaluate the intraobserver and interobserver agreement repeating the analysis a few days later by the same observer and by a second independent observer. The results were expressed as absolute mean difference ± 2 SDs.
The intraobserver agreement for LVEDV, LVESV, LVEF, and peak LV twist were 7.4 ± 11.2 ml, 7.0 ± 10.1 ml, 1.9 ± 4.4%, and 0.2 ± 0.3°, respectively. The interobserver agreement for LVEDV, LVESV, LVEF, and peak LV twist were 12.9 ± 14.7 ml, 11.3 ± 13.9 ml, 2.5 ± 4.9%, and 0.7 ± 0.8°, respectively.
All patients received a biventricular pacemaker with cardioverter-defibrillator function (Contak Renewal 4RF, Boston Scientific, St. Paul, Minnesota; or InSync Sentry, Medtronic Inc., Minneapolis, Minnesota; Lumax 340 HF-T, Biotronik, Berlin, Germany). The right atrial and ventricular leads were positioned conventionally. All LV leads were implanted transvenously, and positioned preferably in a (postero)lateral vein. A coronary sinus venogram was obtained using a balloon catheter, followed by the insertion of the LV pacing lead. An 8-F guiding catheter was used to place the LV lead (Easytrak, Boston Scientific, or Attain-SD, Medtronic, or Corox OTW Biotronik) in the coronary sinus.
LV lead position
Target veins were lateral or postero-lateral veins. The LV lead position was determined using biplane fluoroscopy classification (22). In the right anterior oblique view and/or in the postero-anterior view, the distance between the coronary sinus/mitral plane and the cardiac apex was divided in 3 parts and the LV lead position was classified in 3 groups: basal, midventricular, and apical.
All continuous variables had a normal distribution (as evaluated with Kolmogorov-Smirnov tests). Summary statistics for these data are therefore presented as mean ± SD. Categorical data are presented as numbers and percentages. The paired ttest was used for the comparison between continuous variables at baseline and immediately after CRT and between baseline and at 6-month follow-up. The unpaired ttest was performed to compare continuous variables between normal control subjects and HF patients and between CRT responders and nonresponders. Chi-square/Fisher exact tests were computed to test for differences in categorical variables. Linear regression analysis was performed to determine the relations between LV twist, LVEF, and LV dyssynchrony. In order to identify independent determinants of LV twist, a multivariable linear regression analysis using the enter model was performed including as covariates LVEF and LV dyssynchrony. Linear regression analysis was used to assess the relation between the Δ (difference between immediately after CRT and baseline) peak LV twist and ΔLVEF. The differences in peak LV twist during follow-up in responders and nonresponders were assessed using analysis of variance (ANOVA) for repeated measurements. In order to identify variables related to a positive response to CRT, univariable and multivariable logistic regression analyses were performed including clinical (age, sex, etiology, QRS duration at baseline, and 6-min walk distance at baseline) and echocardiographic (LVESV at baseline, ΔLVESV, LV dyssynchrony at baseline, ΔLV dyssynchrony, peak LV twist at baseline, Δ peak LV twist) characteristics of the patients. Only significant (p < 0.05) univariable predictors were entered as covariates in the multivariable logistic regression analysis, which was performed using the entire model. Odds ratio and 95% confidence intervals were calculated. Model discrimination was assessed using c-statistic, and model calibration was assessed using Hosmer-Lemeshow statistic. The differences in peak LV twist and LVEF between the groups of patients with different LV lead position were assessed by 1-way ANOVA. All statistical tests were 2-sided, and a p value <0.05 was considered significant. The statistical software program SPSS version 14.0 (SPSS Inc., Chicago, Illinois) was used for statistical analysis.
Reliable speckle-tracking for rotation analysis was obtained in all normal control subjects and in 80 (92%) HF patients. Consequently, 7 (8%) patients were excluded from the study. Of the 80 HF patients enrolled, 9 did not complete the 6-month follow-up; 3 patients died of worsening HF, 1 had LV pacing switched off due to intolerable phrenic stimulation, 1 had CRT device explantation secondary to infection, and 4 were lost to follow-up. Therefore, data at baseline and immediately after CRT were collected for 80 patients and data at 6-month follow-up were collected for 71 patients. Baseline characteristics of normal control subjects and the HF patients are listed in Table 1.
LV twist baseline
As shown in Table 1, peak apical rotation, peak basal rotation, and peak LV twist were severely reduced in HF patients compared with normal control subjects: 2.4 ± 1.8° versus 9.4 ± 3.2° (p < 0.001), −3.3 ± 2.0° versus −6.1 ± 2.4° (p < 0.001), and 4.8 ± 2.6° versus 15.0 ± 3.6° (p < 0.001), respectively.
A significant relation (r = 0.53, p < 0.001) was observed between peak LV twist and LVEF in HF patients. This relation was stronger in nonischemic (r = 0.60, p < 0.001) than in ischemic HF patients (r = 0.34, p = 0.020) (Fig. 2A).Moreover, a modest relation (r = −0.33, p = 0.003) was observed between peak LV twist and LV dyssynchrony in HF patients. At multivariable linear regression analysis, LVEF (beta = 0.47, p < 0.001) and LV dyssynchrony (beta = −0.21, p = 0.032) were independent determinants of LV twist.
LV twist after CRT
Immediately After CRT
Immediately after CRT, peak LV twist increased from 4.8 ± 2.6° to 5.9 ± 3.2° (p = 0.007). In particular, Δ peak LV twist was strongly related to ΔLVEF (r = 0.83, p < 0 .001), and this relation was good in both nonischemic (r = 0.85, p < 0.001) and ischemic HF patients (r = 0.82, p < 0.001) (Fig. 2B). Furthermore, the relations between Δ peak LV twist and ΔLV dyssynchrony (r = −0.57, p < 0.001) and between ΔLV dyssynchrony and ΔLVEF (r = −0.63, p < 0.001) were good but less strong than the previous relation between Δ peak LV twist and ΔLVEF.
At 6-month follow-up, 40 of 71 (56%) patients were classified as echocardiographic responders to CRT (defined as a decrease in LVESV ≥15%). No significant differences in the baseline clinical characteristics were found between responders and nonresponders (Table 2).At 6-month follow-up, significant improvement in NYHA functional class (from 3.0 ± 0.5 to 2.0 ± 0.7, p < 0.001), quality of life (from 35 ± 23 to 20 ± 20, p < 0.001), and 6-min walk distance (from 306 ± 106 m to 363 ± 109 m, p < 0.001) were observed in CRT responders only (Table 2).
Baseline echocardiographic characteristics were also similar between the 2 groups, except for LV dyssynchrony (Table 3),which was larger in responders compared with nonresponders (182 ± 71 ms vs. 116 ± 83 ms, p = 0.003). A trend toward lower values of peak LV twist were noted in responders as compared with nonresponders (4.3 ± 2.4° vs. 5.4 ± 2.9°, p = 0.072). At 6-month follow-up, LV dyssynchrony improved in CRT responders (from 182 ± 71 ms to 60 ± 45 ms, p < 0.001), whereas in nonresponders LV dyssynchrony did not change (116 ± 83 ms vs. 136 ± 89 ms, p = 0.30) (Table 3). Importantly, within ischemic HF patients, CRT responders presented a significantly lower number of scarred segments at 2-dimensional echocardiography as compared with nonresponders (2.7 ± 0.9 vs. 4.2 ± 2.2, p = 0.016).
Concerning the rotational parameters, in responders peak LV twist progressively improved during follow-up (ANOVA p < 0.001), whereas in nonresponders a progressive deterioration of peak LV twist was noted (ANOVA p < 0.001) (Fig. 3).Particularly, both apical and basal rotation significantly improved in responders (from 2.3 ± 1.7° to 5.0 ± 3.0°, p < 0.001 and from −3.2 ± 2.2° to −4.3 ± 1.9°, p = 0.006), whereas only basal rotation significantly deteriorated in nonresponders (from −3.5 ± 1.7 to −2.1 ± 2.2, p = 0.001) (Table 3).
Prediction of LV reverse remodeling
At univariable logistic regression, LV dyssynchrony at baseline, ΔLV dyssynchrony, ΔLVESV, and Δ peak LV twist were significantly related to LV reverse remodeling at 6-month follow-up (Table 4).At multivariable logistic regression analysis, Δ peak LV twist was the strongest predictor of response to CRT at 6-month follow-up (odds ratio: 1.899, 95% confidence interval: 1.334 to 2.703, p < 0.001) (Table 4).
LV twist in relation to LV lead position
Considering the 71 patients with 6-month follow-up, 68 patients had the LV lead placed in a (postero)lateral vein and 3 in an anterior vein. The 3 patients with the LV lead positioned in an anterior vein were nonresponders at 6-month follow-up. Of the remaining 68 patients, the LV lead position was classified (from the right anterior oblique/postero-anterior view on fluoroscopy) as basal in 17 (25%), midventricular in 34 (50%), and apical in 17 (25%) patients. At baseline, peak LV twist was not significantly different between patients with apical, midventricular, and basal LV lead position (ANOVA p = 0.68). However, at 6-month follow-up, peak LV twist showed a significant improvement in patients with apical (from 4.3 ± 3.1° to 8.6 ± 3.0°, p = 0.001) and midventricular (from 4.8 ± 2.2° to 6.4 ± 3.9°, p = 0.038) LV lead position, whereas in patients with a basal LV lead position, peak LV twist did not change significantly (5.0 ± 3.3° vs. 4.1 ± 3.2°, p = 0.28) (Fig. 4A).Similarly, LVEF increased significantly in patients with an apical (from 26 ± 7% to 37 ± 7%, p < 0.001) and midventricular (from 26 ± 6% to 33 ± 8%, p < 0.001) but not with a basal (26 ± 5% vs. 28 ± 8%, p = 0.30) LV lead position (Fig. 4B).
Figure 5shows an example of a responder with the LV lead placed in an apical position of a postero-lateral vein and significant improvement in peak LV twist and LVEF after CRT (both immediately after CRT implantation and at 6-month follow-up).
The current study evaluated the effects of CRT on LV twist and provides new insights on the relationship between LV rotational mechanics, CRT response, and LV lead position. The main findings can be summarized as follows: 1) LV twist was significantly reduced in HF patients; 2) LV twist improved in responders and worsened in nonresponders to CRT; 3) the strongest predictor of LV reverse remodeling at 6-month follow-up was Δ peak LV twist (immediate change in LV twist after CRT); and 4) an LV lead placed in a (postero-)lateral vein with apical or midventricular position was associated with the greatest improvement of LV twist after CRT and with the highest response rate to CRT.
Relationship between LV twist and LV function
Several techniques have been applied for the assessment and quantification of LV twist. For this purpose, tagged cardiac magnetic resonance imaging and sonomicrometry are considered the gold standard, but the most recent speckle-tracking echocardiographic technique, used in the present study, demonstrated a good agreement with these imaging modalities (20,21). Previous studies, using both tagged cardiac magnetic resonance imaging and speckle-tracking analysis, suggested an important relation between LV twist and LVEF (4,9). Similarly, in the current study, the relation between LV twist and LV systolic function was good (r = 0.53, p < 0 .001), illustrating the potential role of LV twist as a comprehensive index of LV systolic function. Furthermore, the results of the present study highlight that the relation between LV systolic function and LV twist was stronger in nonischemic patients as compared with ischemic patients. A possible reason may be the presence of regional myocardial damage in ischemic patients, involving specifically the apex or the base with a different effect on LV twist (23).
Finally, LV twist was modestly related to LV dyssynchrony (r = −0.33, p < 0 .001), but at multivariable linear regression analysis, LV dyssynchrony was still independently related to LV twist. This finding points out that LV twist not only is a sensitive and thorough parameter of LV function, but also it may reflect the extent of LV (dys)synchrony.
Relationship between LV twist and CRT response
The effects of CRT on torsional mechanics were different in responders and nonresponders. A trend toward more reduced LV twist at baseline in responders as compared with nonresponders was observed. In the present study, a significant improvement of LV twist was observed in CRT responders and a significant worsening in nonresponders. In contrast, a previous study by Zhang et al. (10) did not show any significant increase of LV twist in responders to CRT. The different results may be related to sample size and population characteristics.
In the multivariable model, baseline LV dyssynchrony and an immediate improvement in LV twist after CRT were the only predictors of LV reverse remodeling at 6-month follow-up. The predictive value of LV dyssynchrony has been shown already in previous studies (19,24). The novelty of the present study is that CRT may (partially) restore LV twist, possibly by providing a more physiological electrical depolarization and mechanical contraction of the myofibers. Specifically, CRT partially restored LV torsional behavior in responders, by not only improving apical rotation but also basal rotation. In nonresponders, the deterioration of LV twist was mainly due to worsening of the basal rotation underscoring the influence of the basal level on LV twist (25).
Relationship between LV twist and LV lead position
Previous studies showed that HF patients treated with CRT showed the best hemodynamic improvement when the LV pacing lead was positioned in (postero)lateral veins (26). In the current study, 3 patients had the LV lead placed in an anterior vein, and none of them responded to CRT. The remaining 68 patients had the LV lead positioned in the (postero)lateral vein. In these patients, the optimal position of LV lead inside the target vein was explored. Patients with a midventricular or apical position had the largest systolic improvement, and showed a significant increase in LV twist, whereas patients with a basal LV lead position did not show improved systolic function and decreased in LV twist, confirming that the pacing site may influence torsional behavior of the LV (27). Similarly, a recent study by Helm et al. (28) reported that the optimal site of stimulation (although in a canine model of HF) was the LV free wall centered over the midapical part. This finding may be related to the fact that normal cardiac depolarization is directed from the apex toward the base (29), and an earlier activation of the LV basal region, altering the normal contraction pattern of the myofibers, may lead to a significant deterioration of LV twist. Another explanation for the findings may be related to the fact that the myocardial wall is thinner toward the apex (30,31); therefore, the epicardial LV lead in this position is closer to the Purkinje network. Consequently, pacing from this position may generate a cardiac pulse that spreads faster to the entire myocardium with a more physiological activation (32–34).
LV twist is reduced in HF patients and improves in patients who respond to CRT. Particularly, the change in LV twist immediately after CRT predicts LV reverse remodeling at 6-month follow-up.
Drs. Ajmone Marsan and Delgado are supported by a research grant from the European Society of Cardiology. Dr. Nucifora is supported by a research grant from the European Association of Percutaneous Cardiovascular Interventions. Dr. Schalij has received grants from Biotronik, Medtronic, and Boston Scientific. Dr. Bax has received grants from Medtronic, Boston Scientific, Biotronik, St. Jude Medical, Bristol-Myers Squibb Medical Imaging, Edwards Lifesciences, and GE Healthcare. Richard A. Grimm, DO, served as Guest Editor for this article.
- Abbreviations and Acronyms
- analysis of variance
- cardiac resynchronization therapy
- heart failure
- left ventricle/ventricular
- left ventricular end-diastolic volume
- left ventricular ejection fraction
- left ventricular end-systolic volume
- New York Heart Association
- Received November 25, 2008.
- Revision received April 17, 2009.
- Accepted May 4, 2009.
- American College of Cardiology Foundation
- Greenbaum R.A.,
- Ho S.Y.,
- Gibson D.G.,
- Becker A.E.,
- Anderson R.H.
- Torrent-Guasp F.,
- Kocica M.J.,
- Corno A.F.,
- et al.
- Buchalter M.B.,
- Weiss J.L.,
- Rogers W.J.,
- et al.
- Notomi Y.,
- Martin-Miklovic M.G.,
- Oryszak S.J.,
- et al.
- Zhang Q.,
- Fung J.W.,
- Yip G.W.,
- et al.
- Hunt S.A.
- Lipkin D.P.,
- Scriven A.J.,
- Crake T.,
- Poole-Wilson P.A.
- Lang R.M.,
- Bierig M.,
- Devereux R.B.,
- et al.
- Chung E.S.,
- Leon A.R.,
- Tavazzi L.,
- et al.
- Suffoletto M.S.,
- Dohi K.,
- Cannesson M.,
- Saba S.,
- Gorcsan J. III.
- Helle-Valle T.,
- Crosby J.,
- Edvardsen T.,
- et al.
- Notomi Y.,
- Lysyansky P.,
- Setser R.M.,
- et al.
- Delgado V.,
- Ypenburg C.,
- van Bommel R.J.,
- et al.
- Notomi Y.,
- Srinath G.,
- Shiota T.,
- et al.
- Butter C.,
- Auricchio A.,
- Stellbrink C.,
- et al.
- Notomi Y.,
- Popovic Z.B.,
- Yamada H.,
- et al.
- Helm R.H.,
- Byrne M.,
- Helm P.A.,
- et al.
- Sengupta P.P.,
- Khandheria B.K.,
- Korinek J.,
- et al.
- Bogaert J.,
- Rademakers F.E.
- Sengupta P.P.,
- Korinek J.,
- Belohlavek M.,
- et al.
- Peschar M.,
- de Swart H.,
- Michels K.J.,
- Reneman R.S.,
- Prinzen F.W.