Three-dimensional rotational angiography of the left atrium and the oesophagus: the short-term mobility of the oesophagus and the stability of the fused three-dimensional model of the left atrium and the oesophagus during catheter ablation for atrial fibrillation

Abstract

Aims

The objective of this study was to evaluate the mobility of the oesophagus and the stability of the three-dimensional (3D) model of the oesophagus using 3D rotational angiography (3DRA) of the left atrium (LA) and the oesophagus, fused with live fluoroscopy during catheter ablation for atrial fibrillation.

Methods and results

From March 2015 to September 2015, 3DRA of the LA and the oesophagus was performed in 33 patients before catheter ablation for atrial fibrillation. Control contrast oesophagography was performed every 30 min. The positions of the oesophagograms and the 3D model of the LA and the oesophagus were repeatedly measured and compared with the spine. The average shift of the oesophagus ranged from 2.7 ± 2.2 to 5.0 ± 3.5 mm. The average real-time oesophageal shift ranged from 2.7 ± 2.2 to 3.8 ± 3.4 mm. No significant shift was detected until the 90th minute of the procedure. The average shift of the 3D model of the LA and the oesophagus ranged from 1.4 ± 1.8 to 3.3 ± 3.0 mm (right–left direction) and from 0.9 ± 1.2 to 2.2 ± 1.3 mm (craniocaudal direction). During the 2 h procedure, there were no significant shifts of the model.

Conclusion

During catheter ablation for atrial fibrillation, there is no significant change in the position of the oesophagus until the 90th minute of the procedure and no significant shift in the 3D model of the LA and the oesophagus. The 3D model of the oesophagus reliably depicts the position of the oesophagus during the entire procedure.

Introduction

Radiofrequency ablation for atrial fibrillation is a recognized treatment for patients with drug refractory atrial fibrillation.¹ Catheter ablation involves the isolation of the pulmonary vein by placing circular radiofrequency lesions around the ostium of the pulmonary veins² in patients with persistent atrial fibrillation; these lesions can be supplemented by additional ablations, such as linear lesions in the left atrium (LA)³ or ablation of the fractionated atrial potentials.⁴ The emergence of an atrioesophageal fistula is a rare, but very serious complication.⁵ Atrioesophageal fistulas comprise <0.1% of all complications⁶ but are fatal in 70–80% of cases⁷ and are the cause of 16% of deaths associated with catheter ablation for atrial fibrillation.⁵ Atrioesophageal fistulas are caused by the application of radiofrequency energy along the posterior left atrial wall, leading to damage of the oesophagus, which is in close contact with this area.^8,9

The visualization of the relative position of the oesophagus and LA can contribute to a reduced risk of oesophageal damage. The mapping of the LA guided by 3D X-ray models of the LA provides objective information about the existing anatomy of the LA and the oesophagus.¹⁰ Contrast-enhanced computed tomography (CT) is a standard method used for the 3D imaging of the LA and the oesophagus.¹¹ The 3D rotational angiography (3DRA) of the LA is a new method equivalent to CT.^12–15 The results of the studies involving the preprocedural imaging of the position of the oesophagus by CT indicated that the oesophagus is a mobile structure, and preprocedural imaging of the position of the oesophagus, in many cases, does not reflect the position of the oesophagus during ablation.^16,17 A single image of the oesophagus obtained at the beginning of the ablation procedure may not be valid for the duration of the procedure, and the position of the oesophagus relative to the LA can vary significantly during an ablation procedure that lasts several hours.^16,18 The aim of our work was to verify the applicability of the periprocedural 3DRA of the LA and the oesophagus for imaging of the oesophagus during catheter ablation for atrial fibrillation. We sought to verify the short-term mobility of the oesophagus and the stability of the 3D model of the oesophagus fused with live fluoroscopy during catheter ablation for atrial fibrillation.

Methods

Patient population

This prospective study enrolled a group of 33 consecutive patients who were referred for catheter ablation for atrial fibrillation. In all patients, the ablation was carried out with the support of the 3DRA of the LA with contrasting imaging of the oesophagus. Patients with a history of iodine allergy or with impaired renal function [glomerular filtration rate <45 mL/s/1.73 m², as estimated using the Modification of Diet in Renal Disease (MDRD) formula] were excluded from the study. The institutional review board approved the study protocol, and written informed consent was obtained from all patients.

Rotational angiography imaging

Imaging was carried out using the Allura Xper FD 10 X-ray system (Philips Medical Systems, Inc., Best, The Netherlands) according to a left atrial protocol. The basic method of 3DRA involves the injection of contrast into the LA and the acquisition of the rotational image. After the opacification of the LA and the pulmonary veins, the C-arm was isocentrically rotated over 240° (120° right anterior oblique to 120° left anterior oblique) over 4.1 s with an X-ray acquisition speed of 30 frames per second. During rotational imaging, the patients were placed in a supine position with their arms in a natural position along the body, and a normal breathing was maintained.

Isocentring of the LA was obtained from the anteroposterior (AP) and left lateral X-ray views. The injection of contrast agent was carried out using a standard power injector (Mark V, Medrad, Inc., Indianola, PA, USA).¹⁹ A pigtail catheter was introduced through the transseptal sheath (Agilis NxT 8.5F, St. Jude Medical, St. Paul, MN, USA) into the LA. After the reduction of the cardiac output via rapid stimulation of the ventricles (frequency 230/min), the application of the contrast agent was initiated, and after a delay of 2 s, we commenced the rotation of the C-arm.²⁰ In total, 60 mL of contrast agent was injected at an injection velocity of 15 mL/s.

The oesophagus was visualized using the oral administration of 20–30 mL of barium sulphate contrast agent (Micropaque, Guerbet, Roissy, France) during the 3DRA of the LA.¹²

The 3DRA model of the LA was reconstructed offline using the EP Navigator workstation (EP Navigator 3.0, Philips Healthcare, Best, The Netherlands). Automatic segmentation was supplemented with manual segmentation when necessary. The 3D model of the oesophagus was manually segmented using the same workstation. The 3D model of the LA and the oesophagus was automatically registered using live fluoroscopy and was used for the navigation of the catheters during the entire procedure.

Current imaging of the position of the oesophagus

To verify the position of the oesophagus during the procedure, contrast oesophagography was performed approximately every 30 min. The oesophagus was opacified using the oral administration of 20–30 mL of barium sulphate contrast agent, and the localization of the oesophagus was recorded on an X-ray screen with the 3D model of the LA and the oesophagus (Figure 1).

Quantitative imaging analysis

For the purpose of our current study, we set a shift of the analysed structure of <5 mm as negligible. According to our experience, a shift of <5 mm is not significant to the operator during atrial fibrillation ablation. Data were evaluated offline by two experienced investigators. In an independent blinded manner, these investigators determined the positions of the oesophagus relative to the LA and measured the positions of the oesophagus and the LA. Interindividual variability was calculated.

At the beginning of the procedure, the positions of the 3D model of the oesophagus and the LA were determined in all patients immediately following the 3DRA and the fusion of the 3D models with live fluoroscopy. The positions of the oesophagus during contrast oesophagography were then determined. The position of the 3D model of the LA and the oesophagus was determined every 30 min during oesophagography. Measurements were performed using GIMP (a free program that enables measurements of JPG images, version 2.8.14, http://www.gimp.org/). The positions of both the 3D model of the oesophagus and the contrast oesophagograms were measured in the superior, middle, and inferior segments of the oesophagus; the spine served as a stationary reference structure. The superior segment of the oesophagus corresponded to the highest level of the 3D model of the LA, the inferior segment corresponded to the bottom level of the LA model, and the middle segment corresponded to the level between the upper and lower segments. The three vertebrae that were closest to the LA model created using the 3DRA were identified. The upper segment of the oesophagus was measured at the level of the upper corner of the lateral edge of the upper vertebra, the middle segment of the oesophagus was measured at the middle of the lateral edge of the middle vertebra, and the lower segment was measured at the level of the lower corner of the lateral edge of the lower vertebra. If the model of the oesophagus was not optimally depicted and oesophageal contours were not shown throughout the entire course of the oesophagus, we manually refilled the contours of the oesophagus before performing measurements. We measured the distances from the vertebrae to the lateral wall of the oesophagus and the width of the oesophagus in all segments, which allowed us to calculate the position of the centre of the oesophagus in every segment. Furthermore, the position of the 3D model of the LA and the oesophagus, relative to the closest vertebrae in the AP projection, was measured. We calculated the right–left shift of the model as the distance between the lateral upper corner of the first vertebra visible above the 3D model of the atrium and the point at which the left superior pulmonary vein and the LA auricle contour intersected. We measured the craniocaudal movement of the 3D model as the distance between the lateral upper corner of the first vertebra visible above the 3D model and the lowest point of the roof of the LA. For detailed measurements, see Figure 1.

Each measurement was repeated three times to reduce intra-individual variability, and the result was the average of these three measurements. The average of the two results measured by the two investigators was used for statistical analysis. In this way, the resulting distances between the oesophagus and the 3D model of the LA from the spine were recalculated into the actual distances in millimetres using a constant obtained by recalculating the known size of the ablation catheter with a diameter of 7 French.

Initially, we evaluated the oesophageal shift or the short-term mobility of the oesophagus during ablation [the shift of the current position of the oesophagus (actual contrast oesophagograms) in the right–left direction towards the input position]. The position of the 3D model of the oesophagus at the first measurement immediately after the fusion with live fluoroscopy was considered to be the entry position of the oesophagus.

Secondly, we calculated the real-time oesophageal shift—the shift of the oesophagus relative to the 3D oesophagus model in real time (the difference between the actual position of the oesophagus obtained by the contrast oesophagogram and the position of the 3D model of the oesophagus fused with live fluoroscopy during a specific measurement, which reflects any possible deviation of the position of the 3D model from the optimally registered model at the beginning of the examination).

Furthermore, the shift of the 3D model of the LA and the oesophagus was evaluated (the stability of the fused 3D model during ablation). The position of the 3D model during the first measurement immediately after the fusion with live fluoroscopy was taken as the input position of the correctly fused model. The right–left and craniocaudal movements of the model of the LA were evaluated, and the change in the position of the 3D model between individual measurements during the procedure was recorded. For details, see Table 2.

Catheter ablation

Ablation procedures were performed in a standard manner under light sedation (boluses of i.v. fentanyl and i.v. diazepam) using an irrigated tip catheter guided by the 3D electroanatomical mapping system EnSite Velocity (St. Jude Medical). Circumferential PV isolation confirmed using the 20-pole circumferential mapping catheter provided the basis for all procedures involving paroxysmal atrial fibrillation. Additional roof and mitral isthmus lines and coronary sinus ablations were performed in cases of persistent atrial fibrillation. Oesophageal imaging was taken into account for the creation of ablation lines in the posterior wall of the LA.

Ethics

This study complies with the Declaration of Helsinki. The research protocol was approved by the locally appointed ethics committee. The informed consent of the subjects has been obtained.

Statistical analysis

The baseline characteristics of the patients, the shift of the oesophagus, and the 3D oesophagus model in real time were analysed descriptively. Continuous variables are presented as the arithmetic means with standard deviations, and categorical variables are presented as the number (%) of patients. A one-sample t-test or its nonparametric alternative (Wilcoxon signed rank test), based on the distribution of data, was used to test whether a shift was significantly lower than 5 mm in each position and at each time point of measurement (30, 60, 90, and 120 min into the procedure). The results with P-values <0.05 were considered statistically significant.

Results

From March 2015 to September 2015, 33 patients undergoing catheter ablation of the LA were examined using 3DRA of the LA and the oesophagus. The majority of patients were males with an average age of 60 years who presented with no structural heart disease, a normal left ventricle function, and a slightly enlarged LA (see Table 1).

The average duration of catheter ablation was 112 ± 43 min, and an average of three contrasting oesophagograms were performed (range: 1–4). On average, contrasting oesophagograms were performed every 33 ± 10 min. The total execution time of the 3DRA of the LA and the oesophagus including the manual segmentation of the oesophagus was 9.92 min.

All measurements were not performed in all patients. Unfortunately, not all of the dimensions were measurable (unclear outline of the vertebra, oesophagus model non-segmented up to the level of the vertebra, overlap of the oesophagus model and vertebra). The number of measured values decreases with the increasing time of the procedure because some procedures had a shorter duration and were already complete at the time of some measurements (e.g. 120 min). A total of 13% of the values were unmeasurable during the study; for results of measured values, see Table 2.

Interobserver variability was 1.8 ± 1.5 mm. Intraobserver variability was 1.5 ± 1.3 mm.

The average width of the oesophagus was 18.8 ± 5.8 mm at the superior position, 19.5 ± 6.1 mm at the medium position, and 16.9 ± 4.6 mm at the inferior position.

The average shift of the position of the oesophagus during catheter ablation (average shift of the contrast oesophagograms during the individual measurements for superior, medium, and inferior positions of the measurement) ranged from 2.7 ± 2.2 to 5.0 ± 3.5 mm. The maximum shift of the oesophagus was 11.9 mm, and the minimum shift was 0.1 mm. A ≥3 mm shift of the oesophagus was present in 44.8% of patients, and a shift of the oesophagus ≥8 mm was present in 5% of the patients. The shift of the oesophagus during the procedure was significantly lower than 5 mm for measurements at 30, 60, and 90 min.

The average real-time oesophageal shift (shift of the virtual model of the oesophagus fused with live fluoroscopy at a given time in relation to the actual position of the oesophagus at the same measurement for the superior, medium, and inferior positions of the measurement) ranged from 2.7 ± 2.2 to 3.8 ± 3.4 mm. The real-time oesophageal shift was significantly lower than 5 mm for all measurements at 30, 60, 90, and 120 min with the exception of the measurement at 120 min in the inferior position.

The average shift of the 3D model of the LA and the oesophagus in the right–left direction ranged from 1.4 ± 1.8 to 3.3 ± 3.0 mm, and the shift in the craniocaudal direction ranged from 0.9 ± 1.2 to 2.2 ± 1.3 mm. The maximum shift of the 3D model was 11 mm in the right–left direction and 6.1 mm in the craniocaudal direction, and the minimum shift was 0 mm in both directions. A shift of the 3D model in the right–left direction ≥3 mm was present in 28.7% of patients, and a shift of ≥8 mm was present in 4% of patients. A shift of the 3D model in the craniocaudal direction ≥3 mm was present in 11.9% of patients, and a shift of ≥8 mm was not detected in any patient. The shift of the 3D model during the procedure was significantly lower than 5 mm for all measurements.

Details of the measurements are summarized in Table 2. For examples of the movement of the oesophagus and the 3D model of the LA and the oesophagus, see Figures 1 and 2.

Discussion

Our work yielded three major results. The first finding was that the oesophagus is a relatively stable structure within a few hours of the ablation procedure; its position in the posterior mediastinum relative to the heart does not significantly change. The non-significant results of the measurements at 120 min are probably due to the movement of the oesophagus after 2 h of surgery, as well as the small number of patients in this group (a range of 12–13 patients). With a small patient population, the average shifts ranged from 3.5 ± 3.0 to 5.0 ± 3.5 mm, which did not reach statistical significance. These results confirm the conclusions of Sherzer et al.,²¹ who reported the stable position of the oesophagus with minimal displacement in a group of 33 patients ablated under general anaesthesia for atrial fibrillation. Our results suggest that the oesophagus behaves similarly in patients ablated under light sedation. Our study did not confirm the results of Good et al. or Daoud et al.,^18,22 which described significant mobility of the oesophagus during these procedures. The cause of this inconsistency is not clear as both works compared two contrasting oesophagograms acquired during catheter ablation for atrial fibrillation conducted under light sedation.

A second notable finding was that when the actual mobility of the oesophagus and the shift of the model of the LA and the oesophagus during the ablation procedure were evaluated, the resulting shift was virtually identical to the displacement of the oesophagus and was not statistically significant. We conclude that the 3D model of the oesophagus reliably reflected the position of the oesophagus during a 2 h procedure.

The third chief result was that the automatic fusion of the 3D model of the LA and the oesophagus resulting from rotational angiography was reliable throughout the procedure, and there was no significant shift of the model towards the patient's heart, even after more than 2 h. The reliability of the fusion of the model with live fluoroscopy was dependent on a number of factors affected by the patient's cooperation during the procedure. The cooperation of patients in whom we performed ablation under light sedation was generally very good. We can conclude that the model of the LA and the oesophagus fused with live fluoroscopy reflects the position of these structures throughout the procedure with great precision. To achieve this reliability, it is not necessary to carry out the procedure under general anaesthesia. To the best of our knowledge, no previous studies have supported this conclusion.

The total execution time of the 3DRA of the LA and the oesophagus was virtually the same as in conventional clinical procedures performed at our department.²³ Three-dimensional rotational angiography constituted 8.8% of the total catheter ablation time.

Our findings are limited by the fact that we measured the lumen of the oesophagus, not the actual oesophagus, during oesophagography. According to our data of the long-term mobility of the oesophagus¹⁶ compared with the preprocedural CT of the heart and the oesophagus (static imaging of the oesophagus) and the periprocedural 3DRA of the LA and the oesophagus (dynamic imaging of the oesophagus during oesophagography), there were no significant differences between the oesophagogram and the CT imaging of the oesophagus. The measured difference between the width of the oesophagus during CT and 3DRA was minimal (average widths of 15.89 ± 4.03 and 17.13 ± 4.3 mm, respectively). This width was in the range of the widths of the oesophagus reported in studies of the detailed anatomy of the oesophagus and the LA using CT data (from 11 to 24 mm).^8,9,24 Swallowing the contrast agent in 3DRA appears to compensate for the thickness of the oesophageal muscle (which is usually <5 mm⁸), and the resulting oesophagogram approximates the actual oesophagus displayed by CT.

The next limitation is the relatively small number of measured values in the group of 120 min measurements and the relatively large portion of unmeasurable values during the oesophagus and LA model measurements. Despite that 13% of the values of the 90 min group were unmeasurable, the results of this group are statistically significant. The combination of a small number of procedures longer than 2 h and the presence of unmeasurable values led to non-significant results in the group of patients measured at 120 min. It appears that after 120 min, the oesophagus is more mobile, but the measured values are only slightly larger and do not exceed 5 mm on average.

The imaging of the oesophagus using barium sulphate contrast agent proceeded without complications. However, the use of barium is not without risk in a possibly injured oesophageal wall, and water-soluble contrast agents such as Gastrografine (meglumine diatrizoate) could be a safer alternative if available.

The final limitation of this study is the possibility that swallowing the barium contrast agent could stimulate the motility of the oesophagus and increase its mobility. However, this effect has not been demonstrated during the routine investigations of oesophageal mobility,²⁵ and our results did not indicate an increased mobility of the oesophagus due to the repeated swallowing of the contrast agent.

Conclusion

There was no significant change in the position of the oesophagus during the ablation procedure lasting up to 90 min. During procedures lasting up to 120 min, the average shift of the oesophagus was up to 5 mm. During the entire procedure, there was no significant shift in the 3D model of the LA and the oesophagus created using 3DRA, which was automatically fused with live fluoroscopy at the beginning of the procedure. The 3D model of the LA and the oesophagus fused with live fluoroscopy was very stable during the procedure. The 3D model of the oesophagus created at the beginning of the catheter ablation for atrial fibrillation reliably reflected the position of the oesophagus during the entire procedure and enabled us to monitor the position of the oesophagus behind the LA throughout the entire process.

Funding

This work was supported by Project No. LQ1605 from the National Program of Sustainability II and by Masaryk University specific research project MUNI/A/1270/2015 and by Masaryk University, Faculty of Medicine, Kamenice 5, 625 00 Brno, Czech Republic.

Conflict of interest: none declared.

References