Feasibility study of electrocardiographic and respiratory gated, gadolinium enhanced magnetic resonance angiography of pulmonary veins and the impact of heart rate and rhythm on study quality

101 consecutive patients referred for pre-catheter ablation MR imaging of the PVs and LA over an eight month study period were included in this prospective study. The study was approved by the Institutional Review Board and was compliant with the Health Insurance Portability and Accountability Act. The requirement for informed consent was waived because of the nature of the study. Baseline demographics, including HR and rhythm, were recorded for all patients at the time of CMR.

All patients were scanned on a commercial 3 T MRI scanner (TimTrio, Siemens, Erlangen, Germany). Electrocardiographic electrodes were positioned for optimal gating before the study. Conventional multiplanar scout images were obtained followed by a high resolution four chamber steady state free precession (SSFP) sequence (echo time : 1 ms, repetition time (TR): 24.24 ms; flip angle : 50 degree; field of view: 340 mm; readout bandwidth: 930 Hz/pixel; matrix size: 1.8 x 1.3 x 8.0 mm; slice thickness: 8 mm; parallel imaging with an acceleration factor of 3 was applied) was used to identify the cardiac phase corresponding to end-systole by visual assessment. The timing of this end-systolic phase was then entered as the trigger delay for acquisitions. The acquisition window duration was set at 100 ms for all patients, given that the end-systolic phase that lasts approximately 100–150 ms is more or less independent of heart rate and related RR interval [25]. Free breathing, contrast enhanced MRA using a ECG triggered, respiratory gated, inversion recovery prepared, 3D volume whole heart acquisition segmented gradient echo sequence (echo time: 1.4 ms, echo spacing: 3.22 ms, number of k space lines per cardiac cycle: 35, flip angle: 20 degrees, inversion time = 200 ms, readout bandwidth = 698 Hz/pixel, slice thickness: 1.2 mm, acquired voxel size: 1.3 × 1.3 × 1.2 mm3, 88–120 slices) was obtained, starting 45 seconds after initiation of an intravenous infusion of 0.15 mmol/kg of gadobenate dimeglumine (MultiHance®; Bracco Imaging SpA, Milan, Italy) at a rate of 0.3 ml/sec using a power injector (MEDRAD Inc., Warrendale, PA, USA), followed by 20 ml saline at the same rate. Gadolinium dimeglumine with a higher T1 relaxivity relative to other gadolinium chelates [26], a dosing regimen similar to that used in a smaller study [16], and the infusion protocol were all selected in an effort to maintain maximum blood pool enhancement for the duration of MRA acquisition. The FOV was selected to include the entire LA, LA appendage (LAA) and proximal PVs. An axial slab and a coronal image taken at end-expiration were used to position the navigator beam on the dome of the right hemi diaphragm to track the end expiratory position of the diaphragm to achieve respiratory gating, with an acceptance window of 5 mm. The position of the navigator bands was adjusted on an axial image to a location lateral to the proximal right pulmonary veins prior to acquisition. A parallel imaging technique with an acceleration factor of two was used to shorten acquisition times. An axial T1 weighted, fat saturated 3D gradient echo sequence (echo time: 1.26 ms, TR: 3.51 ms, flip angle: 10 degrees, readout bandwidth = 500 Hz/pixel; slice thickness: 4.0 mm) was performed after completion of the MRA to demonstrate the course of the esophagus relative to the pulmonary veins (Figure 1).

MRA datasets were imported into an electroanatomic mapping system (CartoMerge Image Integration Module, Biosense Webster Inc., Diamond Bar, CA USA). Using CartoMerge semi automated image integration software, a technician with over 10 years of experience, blinded to other results, segmented the LA and PVs. Segmentation quality was then assessed according to a 4 point score, similar to that used by Wagner et al. [14]: 1- poor segmentation quality due to inability to separate LA and PVs from adjacent structures; 2- moderate segmentation quality with incomplete separation of LA and PVs from adjacent structures; 3- good segmentation quality with near complete separation of LA and PVs, and 4- excellent segmentation quality with complete separation of LA and PVs achieved (Figure 3).

Complications related to catheter ablation of AF were defined as per consensus guidelines [29] and recorded with interval from procedure for all patients. Patients were followed with clinic evaluations, 12 lead electrocardiogram, and/or 24 hour Holter monitor to detect recurrent atrial arrhythmias. Atrial arrhythmias (including documented AF, atrial flutter, or atrial tachycardias) after a 3 month blanking period following ablation were defined as recurrences.

Continuous, normally distributed data are presented as mean ± SD. Continuous, non-normal data are presented as median with interquartile range (IQR). Categorical data are presented as percentages. Data are presented for entire patient cohort, patients in normal sinus rhythm (NSR) and patients in AF at the time of imaging. Continuous variables and binary variables are compared between NSR and AF patient cohorts using a two tailed Student’s t test and Fisher’s exact test, respectively. The Mantel Haenszel Chi Square test is used to test for group comparisons of image quality and segmentation quality scores. The image quality scores of readers 1 and 2 are compared using a paired t test, dichotomized as poor (quality score =1 or 2) or good (quality score = 3 or 4), and inter observer agreement for dichotomized quality scores between readers is determined by calculating the kappa statistic.

The crude relationship between dichotomized consensus image quality (score ≥3 good versus score ≤2 poor) and heart rhythm is presented as an unadjusted odds ratio (OR), with associated 95% confidence intervals (CI). To adjust for the effect of HR at the time of CMR, a logistic regression model with dichotomized consensus image quality as the dependent variable and HR and heart rhythm at time of CMR as predictor variables is used. Effect modification by HR on the association between heart rhythm and image quality is investigated using the same model with the inclusion of an interaction variable. To assess for non linear associations with HR, HR was tested in four formats: (i) continuous linear relationship, (ii) categorical relationship [3 categories: HR ≤ 55 bpm (n = 21), 55 < HR ≤ 90 bpm (n = 66), and HR > 90 bpm (n = 14)], (iii) log transformation of HR, and (iv) quadratic relationship. Multiple linear regression models were used to assess the relationship between categorical HR and rhythm with each of the SNRs and CNRs. A p value of <0.05 was considered significant. Analyses were performed using SAS 9.3® (SAS Institute Inc., Cary, NC, USA).

101 patients were included in this study. All patients had a history of persistent or paroxysmal AF. 35 (34.7%) patients were in AF at the time of CMR. Baseline characteristics for entire patient cohort, as well as by AF status at time of imaging are outlined in Table 1. Patients in AF at time of CMR had a significantly higher mean HR and lower mean systolic BP (SBP) at the time of imaging, lower mean left ventricular ejection fraction (LVEF) and larger mean LA diameter, as measured on a 3 chamber SSFP sequence at end-systole, than patients in NSR.

The mean ± SD acquisition time and acceptance rate were 7:51 ± 2:58 minutes and 54.4 ± 11.8%, respectively. The mean ± SD dose of gadobenate dimeglumine delivered was 29.8 ± 8.5 mls, over a mean ± SD infusion duration of 1:39 ± 0:28 minutes. 100 (99%) patients’ studies were considered diagnostic (consensus quality score > 1), and 91 (90.1%) were of good or excellent quality. Overall, mean ± SD consensus quality score was 3.40 ± 0.69. Among patients in NSR and AF, there were no significant differences in mean acquisition times, acceptance rates, quality scores, SNRs or CNRs, Table 2 and Figure 4.

The mean ± SD quality scores for reader 1 and 2 were 3.34 ± 0.70 and 3.31 ± 0.73, respectively (pooled t test p value = 0.57); the overall inter observer agreement for dichotomized quality scores assigned by these readers was substantial [30], (k = 0.68; 95% CI: 0.46, 0.90).

Heart rhythm was not significantly associated with dichotomized consensus image quality (crude OR = 0.44, 95% CI: 0.08, 2.19; p = 0.49). By logistic regression, neither HR adjusting for rhythm [OR = 1.03, 95% CI = 0.98,1.09; p = 0.22] nor rhythm adjusting for HR [OR = 1.25, 95% CI = 0.20, 7.69; p = 0.81] demonstrated significant association with dichotomized image quality. Further adjusting for LA diameter, LVEF, and SBP did not significantly alter results. Similarly, models fitted to allow for effect modification or non-linear associations with HR did not yield different results. Multiple linear regression models demonstrated that SNRs and CNRs were largely independent of categorical HR after adjusting for heart rhythm and vice versa; with the exceptions of SNR_PV and CNR_{PV/Myocardium} which demonstrate a negative association with HR < 55 bpm compared to a reference HR category (55 < HR ≤ 90 bpm), Table 3. Although mean CNR_{LA/myocardium} was higher than mean values of both CNR_{PV/myocardium} (p < 0.0001) and CNR_{LAA/myocardium} (p < 0.0001), mean values of CNR_{PV/myocardium} and CNR_{LAA/myocardium} were similar, (p = 0.51).

97 (96%) patients’ MRA datasets were segmented using CartoMerge semiautomated image integration software. Segmentation quality scores were good or excellent for 75 (77.3%) patients, with a mean ± SD score of 2.91 (+/− 0.63). There were no significant differences in segmentation quality scores between NSR and AF patient cohorts, Table 4.

94 of 101 (93.1%) patients included in this study proceeded to catheter ablation of AF. Three complications occurred in 2 (2.1%) patients: right phrenic nerve injury and heart block requiring permanent pacemaker insertion occurred at time of procedure in the same patient, and atrioesophageal fistula presented in another patient 17 days after ablation. Follow up data for recurrent arrhythmia were available for 91 of 94 (96.8%); 76 (83.5%) remained in NSR after median (IQR) follow up of 308 (87, 385) days following ablation. The median (IQR) interval from ablation to recurrence of atrial arrhythmias in 15 patients (16.5%) was 348 (191, 414) days.

Direct comparison of this 3D respiratory and ECG gated MRA sequence for imaging pulmonary veins to conventional contrast enhanced, breath held, ungated MRA is necessary, and is a major limitation of this study. While this study establishes the feasibility and high quality of 3D respiratory and ECG gated MRA for imaging pulmonary veins, the lack of a direct comparison precludes conclusions regarding superiority of one technique over another. Superiority over breath-held, ungated MRA will need to be investigated in future studies in order to justify longer acquisition times associated with 3D respiratory and ECG gated MRA. In addition, direct comparison to CTA would be informative. Such comparisons should include an assessment of registration errors during image integration into electroanatomic mapping systems. Certain patients, such as those with claustrophobia, may not tolerate longer acquisition times, but patient tolerance may be improved by the free breathing nature of this respiratory gated technique. Changes in LA and PV anatomy that occur during breath held imaging techniques may predispose to registration errors during image integration into electroanatomic mapping systems [15]; whether imaging in the free breathing state, in its similarity to the breathing pattern during electroanatomic mapping at catheter ablation, offers any advantage in terms of improvement in clinical outcomes is uncertain.