The study group of this feasibility study consisted of 45 subjects including 35 heart failure (HF) patients (65 ? 8 years, 22 men) with reduced systolic LV function (EF 30 ? 11%) and wide QRS complex ? 120 ms (mean QRS duration 158 ? 22 ms). Ten normal subjects without cardiac disease (42 ? 21 years, EF 70 ? 11%) with normal QRS duration (83 ? 5 ms on electrocardiogram) were evaluated for comparison. The inclusion criteria for the HF patients were as follows: 1) heart failure New York Heart Association (NYHA) class III or IV 2) known ischemic (n = 13) or non-ischemic cardiomyopathy (n = 22) with at least a moderately impaired LV function (EF ? 40%) as assessed by CMR and 3) wide QRS complex ? 120 ms with left bundle branch block or interventricular conduction delay. Patients with acute coronary syndromes were excluded from the study. The patients were recruited between October 2004 and November 2006.

The ischemic or non-ischemic etiology of the cardiomyopathy was assessed using invasive cardiac catheterization. We performed cine CMR in all patients and control subjects. In addition, we retrospectively analysed echocardiographic data of the HF patients.

Informed consent was obtained from each patient prior to the investigation in accordance with local Ethics Committee requirements.

Cine CMR was performed on a 1.5 T whole-body MR scanner (Intera, Philips Medical Systems, Best, The Netherlands) with electrocardiographic gating. Images were obtained using a five element cardiac synergy coil during repeated breath-holds (approximately 7 to 15 seconds). For cine imaging a standard SSFP imaging sequence (imaging parameters: echo time (TE) 1.8 ms, repetition time (TR) 3.6 ms, spatial resolution 1.8 ? 1.5 mm², slice thickness 8 mm) was used with a temporal resolution of 40 frames per RR-interval. Three long-axis images and a stack of 10 to 12 short axis images covering the complete LV were acquired.

The data set was transferred to a separate workstation and reconstructed off-line to a 4D volume data set. Quantitative analysis was performed using a software prototype for semi-automatic border detection (TomTec, Unterschleissheim, Germany). Endocardial borders were manually outlined in the three long-axis images at end-diastole and end-systole. Based on these initial contours LV casts are automatically calculated for each of the 40 time frames per RR interval providing a global volume-time-curve. Manual editing of endocardial borders in selected short-axis slices was possible. These changes in 2D images were subsequently adopted in the 3D data set. The cast were automatically subdivided into 17 subsegments resulting in 17 segmental volume-time curves (Fig. 1). The SDI was calculated from the time taken to reach minimum regional systolic volume for each segment as a percentage of the cardiac cycle. The SDI is defined as the standard deviation of these timings [¹⁰]. Higher SDI denotes increasing intraventricular dyssynchrony. In addition, global volumes and ejection fraction were calculated in each patient.

We retrospectively analysed echocardiographic data off-line using commercial software (Echopac 6.1, General Electric-Vingmed) in 32 of 35 (91%) HF patients. These patients received a comprehensive echocardiographic study using a commercially available ultrasonographic system (GE Vingmed, Vivid 7, Horten, Norway). Interventricular mechanical dyssynchrony was assessed by determination of aortic and pulmonary pre-ejection times. A difference of more than 40 ms between the pulmonary and the aortic pre-ejection times (dPEP) was considered as significant delay [¹³]. For assessment of intraventricular mechanical asynchrony, measurement of septal to posterior wall thickening delay (SPWMD) by M-mode echocardiography was performed using a cut-off of ? 130 ms indicating asynchrony [¹⁴]. In addition, in 20 patients (57%) tissue Doppler images from three apical views (2, 3 and 4 chamber views) were available and the standard deviation of the time to peak systolic motion in 12 segments (Ts-SD) according to Yu et al was determined [¹]. A cut-off of Ts-SD = 34.4 ms was considered as measure for the presence of asynchrony [¹].

Continuous variables are expressed as mean values ? standard deviation. Categorical data are presented as frequencies for the study population as well as for the control group and were compared using the Chi-Square test. The unpaired Student's t test was used for comparisons of continuous variables, i.e. for comparison of the ejection fraction, LV volume and SDI between patients and control subjects. The degree of association between SDI and EF as well as between SDI and echocardiographic parameters of asynchrony (Ts-SD, SPWMD, dPEP) was investigated by computing the Pearson's correlation coefficient. In addition, intraobserver and interobserver correlations were evaluated by using the Pearson's correlation coefficient.

Intraobserver variability of SDI was derived from manually outlined endocardial borders in the three long-axis images at end-diastole and end-systole from CMR of 10 randomly selected patients. Studies were performed by the same observer on 2 occasions 3 weeks apart. Interobserver variability was derived from the same subjects, performed by two blinded observers. Both intra- and interobserver variabilities were calculated as the standard deviation of the absolute difference between 2 measurements divided by the mean of both measurements, and expressed as a percentage. In Bland Altman analyses, the intraobserver and interobserver agreements, expressed in terms of the mean difference ? 2 standard deviation were evaluated.

The global significance level for all statistical tests conducted was chosen to ? = 0.05. Statistical analysis was performed with the use of SPSS 10.0 (SPSS Inc., Chicago, USA).

The clinical baseline characteristics of the patient population with heart failure and the healthy control subjects are demonstrated in Table 1.

Data analysis using the semi-automatic software was feasible in all patients. Manual correction of endocardial contours had to be performed in most of the patients (n = 39, 87%). The mean time required for complete analysis was 10 minutes. However, major corrections defined by an additional time required for manual editing of endocardial borders in selected short-axis slices > 5 minutes (range 6?15 minutes) were needed in only a smaller proportion of patients (n = 10, 22%). Eight (80%) of these patients belonged to the HF group. In patients with HF the EF was significantly lower (EF 30 ? 11%) compared to the control subjects (EF 70 ? 11%, p < 0.001). In addition, LVEDV and LVESV measured by CMR were larger in HF patients than in controls (Table 2). The mean SDI was significantly increased in patients with HF (14 ? 5%, range 6.0%?23.4%) compared to control subjects (4.9 ? 1.5%, range 3.2%?7.8%; p < 0.001). An example of segmental volume-time-curves for a patient with marked dyssynchrony and a normal control subject is shown in Fig. 2. A significant exponential correlation between the mean SDI and EF was observed (r = -0.84, p < 0.001; Fig. 3). The correlation between mean SDI and EF for patients with ischemic cardiomyopathy and non-ischemic cardiomyopathy was r = -0.59 and r = -0.76, respectively. In addition, a strong correlation between the QRS complex duration and SDI was noted (r = 0.67, p < 0.001).

Intraobserver and interobserver variability for SDI was 2.2% and 3.4%, respectively. In addition, intraobserver and interobserver correlation was r = 0.99 and r = 0.94, respectively. In Bland Altman analyses, the intraobserver and interobserver agreements, expressed in terms of the mean difference ? 2 standard deviation (upper and lower limits of agreement) were 0.02 (-1.87 to 1,91%) and 0.16 (-3.45 to 3.77%), respectively.

A significant association between the SDI assessed by CMR and the standard deviation of time to peak systolic motion of 12 LV segments (Ts-SD) determined by Tissue Doppler imaging was observed (r = 0.66, p = 0.002, n = 20, Fig. 4). However, no significant associations between the SDI and SPWMD (r = 0.31, p = 0.11, n = 29) as well as between the SDI and dPEP (r = 0.04, p = 0.82, n = 32) were noted.

The authors thank Dr. Ole A. Breithardt, MD for his assistance in the preparation of the manuscript.

Values are presented as mean ? S.D. or absolute values with percentage in parenthesis as appropriate.

Values are presented as mean ? S.D.

LVEDV = left ventricular end diastolic volume, LVESV = left ventricular end systolic. volume, EF = ejection fraction, SDI = systolic dyssynchrony index.