Asymmetric collimation can significantly reduce patient radiation dose during pulmonary vein isolation^†

Abstract

Introduction

The conventional treatment of atrial fibrillation (AF) by radiofrequency ablation (RFA) uses fluoroscopic imaging to visualize both a circular catheter, inserted in each of the pulmonary veins (PVs), and an ablation catheter that is guided around the ostium for ablation. While fluoroscopic imaging provides a very high level of accuracy, it is associated with a considerable amount of X-ray radiation, potentially inflicting harm to the patient.

Electro-magnetic (EM) catheter tracking systems, like LocaLisa™(Medtronic), Carto™ (BioSense-Webster), and NavX/EnSite™ (St Jude Medical), have been introduced to reduce radiation dose for electrophysiology procedures, particularly for AF ablation. Several studies have shown that the use of these systems reduces radiation time for ablation of atrio-ventricular nodal reentry tachycardia (AVNRT), atrial flutter, and AF^1–3with a potential dose reduction ranging from 46 to 66%. However, successful AF ablation is largely reliant upon accurate definition of the anatomy of the left atrium (LA) and, thus, integration of pre-procedural 3D computed tomography (CT) reconstructions is frequently employed in conjunction with EM tracking systems. This increases patient radiation exposure by 1.5-fold⁴when compared with traditional fluoroscopy-guided procedures. The increase can be attributed not only to the CT image acquisition but also to the registration process that is needed to correctly align the CT image to EM coordinate system. In addition, the use of EM tracking systems may require the use of special purpose catheters which can increase the procedural cost substantially.

Because of its high registration accuracy, reliance on a single-imaging system, and cost-effectiveness, another approach has been developed that relies solely on integration of 3D image data with fluoroscopy (3D–2D overlay). The effective use of such an approach has been demonstrated for AVNRT^5,6and AF treatment.^7–9However, this approach results in an even greater radiation dose when compared with conventional fluoroscopy-based AF ablation.

Thus, all current systems employed for AF RFA involve substantial radiation doses. Nevertheless, operators should be aware of radiation risks and should make all efforts to reduce radiation dose. Relatively simple strategies such as verbal reinforcement of fluoroscopy times, effective collimation, minimizing the source–intensifier distance, and effective lead shield use have been applied¹⁰to effectively reduce radiation dose during routine clinical work.

Current fluoroscopy systems only allow symmetric collimation. This implies that the object of interest should always be central to the fluoroscopic image and therefore requires movement of the table position in addition to adjusting collimation. Table movement adds considerable complexity to the procedure in that 2D–3D overlay adjustments are required on each occasion which entails the risk of not using it. In addition, table movement may increase the risk of registration mismatches in the 2D–3D overlay. Finally, when performing the intervention following the conventional workflow, angiographic imaging should be repeated for each of the PVs. If the considerable complexity of table movement is not undertaken, then it is rarely possible to restrict the imaging window so as to encompass a single PV without irradiating a significant and unnecessary, wider region. Asymmetric collimation, on the other hand, can redirect X-rays to visualize a single PV and minimal adjacent structures. In a similar manner, asymmetric collimation is commonly used in oncologic radiation therapy to reduce harmful radiation for the patient and/or provide more flexibility in the ability to shield sensitive organs.^11,12

In this study, we examined the potential impact of introducing an asymmetric collimation technique in the electrophysiology lab for use in fluoroscopy-guided treatment of AF. We compared the effective dose for both optimal symmetric and asymmetric collimation approaches to assess the potential impact of a simple collimation measure on total patient radiation exposure.

Methods

Patients and procedures

This study consists of a retrospective simulation study and a prospective validation study which were performed on a series of 20 consecutive patients (17 males and 3 females, age 32–65) and 10 patients (8 males and 2 females, age 36–69), respectively. All patients underwent point-by-point radiofrequency AF ablation, guided by 3D–2D overlay guidance using custom locally developed software, Leuven Augmented Reality Catheter Ablation-software (LARCA).^6,5,8Clinical routine included the recording of patient parameters relevant for radiation dose calculations like age, sex, height, weight, and body mass index. In addition, procedural parameters included scope angle for the right and left anterior oblique (RAO and LAO) views, and procedural dose area product (DAP) for each view individually. The imaging tubes in RAO and LAO were individually adjusted based on the orientation of the His bundle and mitral plane annulus, with the LA centrally in the image.¹³

During the ablation the patient was kept under general anaesthesia with propofol and was mechanically ventilated. A femoral approach was used to introduce four catheters of which two were positioned in the LA (SL-0 sheaths, St Jude Medical, Daig Division Inc., Minnetonka, MN). Mapping of PV potentials was performed with a deflectable decapolar catheter with a distal ring configuration (Lasso, Cordis-Webster, Diamond Bar, CA). To accurately plan the target ablation lines, the LA was imaged by rotational angiography (syngoDynaCT, Siemens AG, Erlangen, Germany) using direct injection of diluted iodine contrast agent (Iomeron 400, Bracco, Milan, Italy) as described previously.¹⁴During imaging cardiac standstill was obtained by either adenosine administration or rapid ventricular pacing. Semi-automated segmentation was performed to obtain a 3D surface model of the LA by custom ‘in-house’ software (EPSegmenter).

Target ablation points were marked on the 3D surface model and shown together with a semi-transparent 3D model of LA and PVs overlaid on the live fluoroscopy images as depicted in Figure 1A. During ablation, the operator guided the ablation catheter to the target points by using the biplane overlay images. Ablation was performed with a standard EPT catheter using temperature feedback with a target temperature of 50°C and a maximum power output of 30 W or with an irrigated tip catheter at 20–30 W and 17 cc/min perfusion rate. The endpoint for PV isolation was the creation of bidirectional conduction block between the atrium and PV. This endpoint was reached by ablation of the circumference of each vein individually or of each ipsilateral vein pair. For each target line the collimation window was adjusted symmetrically around the centre of the LA, minimizing its surface but keeping all ablation targets around the individual PVs visible.

Figure 1

3D–2D fluoroscopy overlay views, treatment, and collimation scenarios. (A) Biplane overlay of a 3D left atrium model with target ablation points and fluoroscopic images as used during real-time ablation and offline in this study, to determine optimal symmetric/asymmetric collimation windows. The overlay shows the location of the right superior right inferior, left superior, and left inferior pulmonary vein. The figure shows a single collimation window in right and left anterior oblique for use throughout an entire procedure, as is still commonly used in many labs today (‘Sc 1’ in this manuscript). (B) Illustration of the different collimation and treatment scenarios used in this study depicted over an uncollimated fluoroscopic image with 3D surface and target ablation lines in overlay. Both the right and left anterior oblique views are shown with collimated regions visualized by grey semi-transparent rectangles. The remaining, visible part, entirely encompassing each vein or vein pair, determines the field size and is used as input to the Monte-Carlo simulations. When comparing symmetric to asymmetric collimation, a large discrepancy becomes apparent between the useful radiated tissue and the currently used symmetric collimation.

Open in new tabDownload slide

Asymmetric collimation approach

In a retrospective study, we compared three different collimation approaches (single symmetric, adjusted symmetric, and adjusted asymmetric) in combination with two common treatment scenarios (encircling of ipsilateral veins and per vein ablation). This results in the following five scenarios: An example of a single procedural collimation window (Sc 1) is given in Figure 1A. Asymmetric collimation was simulated assuming that each of the four collimation borders could be individually set while symmetric collimation was performed around the centre of the fluoroscopic view. All adjusted collimation scenarios, as visualized in Figure 1B, were such that all target ablation points of a single encircling (either around an individual PV or around a PV pair) remained visible during ablation of that line, and without repositioning the patient throughout the procedure.

1. a single collimation window for the entire procedure (Sc 1),

2. a symmetric collimation window adjusted for two ipsilateral veins (Sc 2),

3. a symmetric collimation window adjusted for each individual vein (Sc 3),

4. an asymmetric collimation window adjusted for two ipsilateral veins (Sc 4), and

5. an asymmetric collimation window adjusted for each individual vein (Sc 5).

In a prospective validation study, the effect of symmetric collimation (Sc 2) and asymmetric collimation (Sc 4) around each vein pair was measured by short fluoroscopy sequences consisting of 5–11 fluoroscopic frames recorded over an equal amount of time. Preliminary testing had shown that this number of frames yielded reproducible measurements. Symmetric collimation measurements were performed by centring the LA in each view and collimating around each vein pair. Asymmetric collimation was realized by repositioning the patient table such that each vein pair was centred in both RAO and LAO views and symmetrical collimation could be performed around the vein pair (since the current cath lab system does not allow asymmetric collimation yet). In addition, the time to centre and collimate around each vein pair was measured.

Dose calculation

Effective doses originating from fluoroscopic imaging were calculated by a Monte-Carlo technique for every patient in all collimation scenarios. Monte-Carlo simulation was performed using the PCXMC software suite (Radiation and Nuclear Safety Agency, Helsinki, Finland). This software package enables calculation of the effective dose based on the amount of exposure to X-ray radiation, imaging parameters like RAO and LAO angles, focus to skin distance and field size, and patient characteristics. The patient characteristics determine the size and shape of the modified hermaphrodite phantom, originally proposed by Cristy and Eckerman¹⁵and which is used in the organ dose simulations. Filtering was set with an aluminium filter of 2.5 mm thickness and a copper filter of 2.8 mm which are average values determined on a number of procedures. X-ray tube anode angles for the RAO and LAO plane were set, respectively, at 12.5 and 8.5°. The received radiation dose of each organ was converted to an effective dose by using the ICRP 103 tissue weighting factors.¹⁶

In a retrospective study, DAP for the RAO and LAO views during procedural fluoroscopy was computed by subtraction of the DAP values of acquisition images, which include DynaCT rotational angiography images, from the total procedural DAP value. Acquisition image DAP values were retrieved from the DICOM image header. We assumed that each PV required approximately the same fluoroscopy time for ablation. The radiation concentration at the skin entrance point for a specific PV was calculated by

in which DAP_iand DAP_tare the fluoroscopy DAP for, respectively, a specific PV and all PVs together; and A_iis the field size of the ith PV at skin entrance of the radiation beam. The radiation concentration can then be computed by dividing DAP_iby A_i. A_iwas derived from the source-to-skin distance, the source-to-image distance, and the field collimation width and height. The first two could be retrieved from the DICOM images while the latter two could be determined from an overlay image as depicted in Figure 1A. The overlay image visualizes the 3D model, generated from rotational angiography, the planned encirclement lines, and the real-time fluoroscopic images. The location of the target lines could easily be determined by observing the location of coloured spheres. Simulation of all collimation scenarios was conducted under the assumption of equal radiation concentration at the skin entrance and by modification of the field size.

In the prospective study, the effective dose per frame was simulated based on the quotient of the DAP value that was recorded in a stored DICOM fluoroscopy sequence and the number of recorded frames of each sequence. Other imaging parameters, including collimation window size and focal distance, were also retrieved from the stored DICOM header.

Statistical analysis

All summary values are reported by their mean and standard deviation. Data comparison is performed by a paired t-test.

Results

Patient and procedural characteristics are summarized in Table 1. All PVs in all patients could be isolated.

Retrospective study

Conventional treatment using a single collimation field for the entire procedure (Sc 1) would have resulted in an effective dose of 45 ± 31 mSv (DAP: 223 ± 149 Gy cm²). When treating both ipsilateral veins using a single ablation target the radiation dose would be 36 ± 25 mSv (DAP: 173 ± 115 Gy cm²) for symmetric collimation (Sc 2) or 15 ± 10 mSv (DAP: 73 ± 49 Gy cm²) for asymmetric collimation (Sc 4) as shown in Figure 1B. The procedural effective radiation dose received during fluoroscopy with symmetric collimation and a per vein approach (Sc 3), as it was clinically used in all 20 patients, was 24 ± 14 mSv (DAP: 120 ± 71 Gy cm²). Asymmetric collimation in a per vein approach (Sc 5) would result in an effective dose radiation of 5 ± 3 mSv (DAP: 23 ± 15 Gy cm²) (see Figure 2) i.e. a reduction by 80 ± 5% (P< 0.001) compared to Sc 3. Comparison of all adjusted collimation window approaches with the conventional single collimation window approach (Sc 1) leads to effective dose reductions of 22 ± 9% (Sc 2; P< 0.001), 43 ± 9% (Sc 3; P< 0.001), 66 ± 6% (Sc 4; P< 0.001), and 89 ± 2% (Sc 5; P< 0.001).

Figure 2

Effective dose for different collimation and treatment scenarios obtained from a retrospective simulation study (A) and a prospective clinical validation study (B). (A) Patient effective dose in mSv for the five scenarios studied in a retrospective study: a single symmetric collimation for the entire procedure (Sc 1), an ipsilateral symmetric collimation approach (Sc 2), a symmetric per vein collimation approach (Sc 3), an ipsilateral asymmetric collimation approach (Sc 4), and an asymmetric per vein collimation approach (Sc 5). Effective dose values are represented as mean ± 1 SD. (B) Patient effective dose of a single average fluoroscopic frame in μSv for Sc 2 and 4. Scenario 4 was obtained by repositioning of the patient table. Effective dose values are represented as mean ± 1 SD.

Open in new tabDownload slide

The proportional reduction in dose is similar for the RAO and LAO view when applying asymmetric collimation. This holds both for an ipsilateral vein approach (−58% RAO vs. −55% LAO, P= 0.22) and a per vein approach (−80% RAO vs. −79% LAO, P= 0.88). However, as illustrated in Figure 3, for all five collimation scenarios the effective dose originating from the LAO view is significantly larger than from the RAO view. Thus, asymmetric collimation has the largest absolute impact on the LAO view independent of the collimation and treatment approach.

Figure 3

Effective dose as function of projection view obtained from a retrospective simulation study (A) and a prospective clinical validation study (B). (A) Comparison of the effective dose of radiation originating from the right and left anterior oblique views in different collimate on scenarios. Although the proportional reduction in dose between right and left anterior oblique as a result of asymmetric collimation is independent of the treatment approach (both ipsilateral veins or each vein separately), the absolute values for the left anterior oblique view radiation is about double the right anterior oblique view values. Therefore, the implementation of an asymmetric collimation will yield the largest absolute benefit for left anterior oblique view fluoroscopy and physicians should be aware of the higher risk when using left anterior oblique imaging. (B) Results obtained in a prospective clinical validation study comparing right and left anterior oblique view doses for collimation Sc 2 and 4. Effective dose values represent dose of a single average fluoroscopic frame in μSv. These data confirm what was computed in the retrospective study (A).

Open in new tabDownload slide

Prospective study

Comparison of symmetric collimation around the LA centre and symmetric collimation around each of the ipsilateral vein pairs, thereby mimicking asymmetric collimation, resulted in a single frame effective dose of 1.8 ± 0.8 μSv (DAP: 4.37 ± 2.3 mGy cm²) and 0.7 ± 0.4 μSv (DAP: 1.83 ± 1.0 mGy cm²), respectively. Thus, in a clinical setting asymmetric collimation can realize an effective dose reduction of 56% (P< 0.001) when treating both ipsilateral veins by a single ablation line. This reduction is not significantly different from the result obtained in the retrospective study (P= 0.95).

Also the view dependency of the effective doses is confirmed in the clinical validation experiments: RAO and LAO reductions by the use of asymmetric collimation are not significantly different (51 vs. 57%, respectively, with P= 0.41).

The time required for centring one vein pair by repositioning the patient and for performing symmetric collimation around each ipsilateral vein pair, added up to 177 ± 116 s of which the largest part could be attributed to repositioning of the patient table.

Discussion

Our data, including both simulation and clinical measurements, indicate that the implementation of asymmetric X-ray beam shielding has the potential to substantially reduce radiation dose during PV isolation procedures. While it is technically feasible as shown in other applications and even hardware enabled by some modern systems, we are not aware of a study investigating or suggesting the possibility of asymmetric collimation for ablation. Using asymmetric collimation, the effective radiation dose from fluoroscopic imaging during PV isolation can be reduced to an average of 5 mSv, which is comparable with the effective dose received by a patient during coronary angioplasty (percutaneous coronary intervention).¹⁷

Although use of EM catheter tracking systems without any imaging integration has been shown to reduce fluoroscopy dose up to 66%,^1–3the combination of EM catheter tracking with 3D anatomy integration tends to increase radiation dose rather than decrease it while providing a highly detailed, more accurate 3D anatomical surface.¹⁸Della Bella et al.⁴reported an effective dose for conventional treatment of 44 mSv, including an acquisition of a venogram, while an image integration approach required 65 mSv radiation dose (of which 5 ± 1 mSv was attributed to CT imaging). The large difference in obtained fluoroscopy radiation dose compared with our results (44 vs. 24 mSv) could be explained by a difference in frame rate (for instance 6 fps instead of 3 fps), the use of a single collimation window, the additional radiation of the venogram, or a combination of these. Also, a recent randomized study did not show superiority concerning procedural outcome and radiation dose of electro-anatomic mapping vs. a 3D-fluoroscopy overlay approach.¹⁹Although this study reports lower fluoroscopy times than we, the effective dose may be similar because of the low fluoroscopy framerate used in our centre.

Additionally, radiation dose for 3D imaging should be considered. Depending on the 3D-imaging modality around 5 mSv for cardiac CT, around 6.6 mSv for rotational angiography²⁰and zero radiation dose for magnetic resonance imaging should be added. The combined radiation of both 3D imaging and fluoroscopy with asymmetric collimation would still result in a substantially reduced overall radiation dose and since all catheters of interest remain visible, it will most likely not affect procedure outcome. The 3D image overlay allows very precise collimation of only the target ablation points, which can be precisely determined from the 3D surface of the LA. Therefore, no visualization of other reference structures, like the spine, is needed.

Our results also indicate the importance of adjusting collimation as a function of the vein location. Indeed, even optimal symmetric collimation adjusted for each pair of veins or each individual vein can reduce dose by 22 and 43%, respectively, compared to using a single collimation for all veins.

Although the use of a wide fluoroscopic view on the heart to monitor pericardial effusion is no longer available when applying narrow field asymmetric fluoroscopic collimation, this complication could be detected by monitoring other vital signs, like heart rate, central venous pressure, and non-invasive blood pressure as is routinely performed in our centre.

In order to be used as described above, asymmetric collimation should be implemented in a way which requires minimal user intervention. This will also ensure that using it will result in a minimal increase in procedure duration. If it is properly integrated in the clinical workflow, we estimate that the increase will be <1 min.

The availability of an asymmetric collimation technique in the EP room and its associated radiation benefits may also have an impact on the choice of treatment strategy. Indeed, a number of studies have shown that a wide encircling of both PVs results in an improved outcome compared to the traditional ostial isolation of each vein individually,^21–24although there is no universal agreement.²⁵Our results show that effective dose of radiation by asymmetric collimation with individual PV ablation is three times lower than by encircling two ipsilateral PVs. These data need to be taken into account when weighing pros and cons of an ablation strategy. Another option for an approach encircling both ipsilateral PVs by one line could be to divide this line in a number of segments and to collimate asymmetrically for each segment. The impact of these alternatives on the success and effectiveness of PV ablation should be evaluated in further studies.

An alternative to asymmetric collimation would be to reposition the patient table for each PV, combined with simultaneous adaptation of the symmetric collimation window, as we performed in the prospective validation study. This approach does, however, require much more user intervention and more fluoroscopy during the repositioning. In the prospective study, an accurate centring around one vein pair took about 3 min which would add up to 12 min of additional time when centring around each individual vein. Additionally, it would also require re-registration for each vein, although that may become fully automated in the future. From an ergonomic and cost-effective viewpoint, this approach would be suboptimal. In case automated registration would fail, it would also increase the risk of severe harm to the patient since it could lead to applying ablation energy at undesired locations with risk of perforation or PV stenosis. Ideally, in the future, the imaging system could automatically suggest the collimation window on the basis of the treatment plan (i.e. target lines).

Limitations

We simulated asymmetric collimation by moving a symmetric beam towards a different location centred on a specific target area. It has been shown that asymmetric collimation may cause a different dose distribution than the corresponding symmetric dose distribution, which could result in an associated change in radiation dose.¹¹The values we obtained here may therefore differ slightly from the true obtainable reduction. However, implementation details of the filters and the automated exposure regulators on the imaging system may counteract these distortion effects.

In this study, we assumed an equal radiation time for each vein. In reality isolation of the left PVs may require a larger percentage of the procedure and imaging time. This can impact on the resulting absolute effective doses but will probably have a negligible effect on the obtained dose reductions.

Conclusion

Alternative collimation strategies during PVI show that implementing an asymmetric collimation approach can greatly diminish the effective dose for the patient. This decrease ranges from 66 to 89% when compared with a single collimation field for all veins, which is still commonly used in clinical practice, depending on whether ablation is performed on both ipsilateral PVs or on each individual PV.

Since asymmetric collimation can have an important impact on the reduction of harmful effects related to ablation treatment, imaging system manufacturers should prioritize its implementation and physicians should adopt it in their workflow when available.

Conflict of interest:The authors wish to make known the following disclosures. H.H. is a member of the scientific advisory board of Siemens Medical Solutions, Bayer, Daicchi-Sankyo, Merck, Sanofi-Aventis, Boehringer-Ingelheim and Bristol-Myers Squib, and receives unconditional research grants through the University of Leuven from St Jude Medical, Medtronic, Biotronik, and Boston Scientific Inc. S.D.B. and H.H. receive research funding through the University of Leuven from Siemens Medical Solutions for related topics. This research was academically driven and not supported by specific grants.

References

Author notes