- Email: [email protected]
Biplane three-dimensional augmented fluoroscopy as single navigation tool for ablation of atrial fibrillation: Accuracy and clinical value
Biplane three-dimensional augmented fluoroscopy as single navigation tool for ablation of atrial fibrillation: Accuracy and clinical value Joris Ector, MD,* Stijn De Buck, MSc, PhD,† Wim Huybrechts, MD,* Dieter Nuyens, MD, PhD,* Steven Dymarkowski, MD, PhD,‡ Jan Bogaert, MD, PhD,‡ Frederik Maes, MSc, PhD,† Hein Heidbüchel, MD, PhD* From the Departments of *Cardiology, †Electrical Engineering, and ‡Radiology, University Hospital Gasthuisberg, University of Leuven, Leuven, Belgium. BACKGROUND We developed new methods for real time biplane integration of three-dimensional (3D) left atrial models with fluoroscopic images to assist in catheter ablation of atrial fibrillation (AF). OBJECTIVE The purpose of this study was to quantitatively assess the accuracy of 3D fluoroscopy integration and to evaluate its clinical value when used as a single navigation tool for AF ablation. METHODS Sixty patients underwent AF ablation under biplane fluoroscopic guidance after selective angiography of the four pulmonary veins. Computed tomography [CT]-based 3D models were integrated in the fluoroscopic framework using visual matching and landmark-based registration approaches. Integration accuracy was quantitatively assessed according to registration approach and different CT acquisition parameters (electrocardiogram [ECG] gating, respiratory phase). In 30 of the 60 patients (3D⫹ group), the integrated 3D model was used for real time 3D-augmented fluoroscopic catheter navigation, and the effects on procedural parameters and patient radiation dose were evaluated. RESULTS Landmark-based registration resulted in superior 3D fluoroscopy integration accuracy compared with the visual matching approach (P ⬍.001 for alignment error and alignment score). The
Introduction Preprocedural imaging and three-dimensional (3D) reconstruction of the left atrium (LA) and pulmonary veins (PVs) is performed in the majority of centers before atrial fibrillation (AF) ablation procedures.1 The detailed patient-specific anatomical information can help achieve a more efficient and successful ablation and may prevent procedure-related complications. In addition, nonfluoroscopic 3D mapping
Hein Heidbüchel is a member of the scientific advisory board of Biosense Webster, Inc., and St. Jude Medical, Inc. The other authors have no conflicts of interest to declare. Stijn De Buck is funded through the Institute for the Promotion of Innovation by Science and Technology in Flanders of the Flemish government (project OZM 080511). Address reprint requests and correspondence: Joris Ector, M.D., Cardiology, University Hospital Gasthuisberg, Herestraat 49, B-3000 Leuven, Belgium, Europe. E-mail address: [email protected]. (Received January 24, 2008; accepted March 18, 2008.)
effects of ECG gating and respiratory phase during CT acquisition on integration accuracy were small and clinically irrelevant. The use of 3D-augmented fluoroscopy in the 3D⫹ group was gauged as extremely helpful by the operator. It resulted in a significant reduction of fluoroscopy time (61 ⫾ 18 minutes vs. 77 ⫾ 26 minutes; P ⫽ .009) and a trend toward shorter procedure duration (230 ⫾ 67 minutes vs. 257 ⫾ 58 minutes; P ⫽ .06) versus conventional procedures. The systematic use of nongated cardiac CT in the 3D⫹ group resulted in an important reduction in total effective patient radiation dose due to CT⫹fluoroscopy (4 ⫹ 14 ⫽ 18 ⫾ 8 mSv vs.17 ⫹ 16 ⫽ 33 ⫾ 13 mSv; P ⬍.001). CONCLUSIONS Biplane 3D-augmented fluoroscopy can be used as a safe and accurate stand-alone method to guide AF ablation procedures. The use of nongated cardiac CT substantially reduces total patient radiation dose without a relevant reduction in integration accuracy. KEYWORDS Atrial fibrillation; Ablation; Image integration; Fluoroscopy; Computed tomography (Heart Rhythm 2008;5:957–964) © 2008 Heart Rhythm Society. All rights reserved.
systems allow the integration of this 3D anatomical information during the ablation procedure to directly guide catheter navigation and ablation. Recently, integration of 3D models of the LA and PVs has also been applied to fluoroscopic imaging. In 3D-augmented fluoroscopy, computed tomography (CT)/magnetic resonance imaging (MRI)-based 3D models are shown as a semitransparent overlay on fluoroscopic images, using specific calibration and registration procedures.2,3 When applied to fluoroscopy in a single plane, this method provides a two-dimensional (2D) projection of fluoroscopic catheter positions on the integrated 3D overlay but does not allow determination of 3D catheter position unless continuosly changing the fluoroscopic view angles. Monoplane 3Daugmented fluoroscopy has therefore only been reported as an additional tool used in combination with electroanatomic mapping to guide AF ablation procedures.4
1547-5271/$ -see front matter © 2008 Heart Rhythm Society. All rights reserved.
doi:10.1016/j.hrthm.2008.03.024
958 Given the limitations of monoplane 3D-augmented fluoroscopy, our research group developed new methods for real time biplane integration of 3D LA models with fluoroscopic images. In this work, the integration accuracy of our approach is quantitatively assessed, and the influence of different registration methods and preprocedural imaging parameters is evaluated. Moreover, biplane 3D-augmented fluoroscopy was used as a single navigation tool during AF ablation to evaluate its clinical value and possible drawbacks, such as the associated patient radiation dose.
Methods Study population Our study included 60 patients (54 men and 6 women, age 53 ⫾ 7 years) undergoing a first ablation procedure because of symptomatic, drug-refractory AF. AF was paroxysmal in 41 (68%) and persistent in 19 (32%) patients. Structural heart disease was present in only three patients (hypertrophic cardiomyopathy in one, myocardial infarction in two). Mean patient weight, height, and body mass index were 83 ⫾ 11 kg, 179 ⫾ 9 cm, and 26.1 ⫾ 2.7 kg/m2 respectively.
Preprocedural imaging and 3D reconstruction All patients underwent cardiac CT on the day before the ablation procedure with a 64-slice multidetector scanner. The first 31 patients underwent cardiac CT with retrospective electrocardiogram (ECG) gating on a Siemens Somatom Sensation 64 scanner Siemens AG, Medical Solutions, Forchheim, Germany. The 29 following patients underwent non-ECG-gated cardiac CT on a Philips Brilliance 64 scanner (Philips Medical Systems, Eindhoven, The Netherlands), to examine the influence of ECG gating on 3D integration accuracy and patient radiation dose.5 Sixty to 80 mL of iodine contrast agent (Iomeron 400, Bracco, Milan, Italy) were administered intravenously in all patients at a rate of 5 mL/s to obtain contrast opacification of the LA and PVs. To examine the influence of respiration during preprocedural imaging on 3D fluoroscopy integration accuracy,6 CT acquisition was performed during inspiration in 11/31 ECG-gated acquisitions and in both inspiration and expiration in 13/29 nongated CT acquisitions. All other patients underwent CT imaging during expiration only. The resulting axial images were imported into a 3D reconstruction software package (Amira 4.1, Mercury Computer Systems SAS). Intensity-based segmentation of the LA and PVs was performed to reconstruct a detailed 3D LA-PV surface model.
Ablation procedures and 3D model integration General procedural aspects Ablation procedures consisted of Lasso-guided (Lasso, Biosense Webster Diamond Bar, CA) electrical isolation of the four PVs. All procedures were performed under fluoroscopic guidance with a Siemens Coroskop C biplane image intensifier system. The system was routinely set to pulsed
Heart Rhythm, Vol 5, No 7, July 2008 fluoroscopy at 3 frames per second and a cine frame rate of 12.5 frames per second to minimize radiation exposure. Standard 30° right anterior oblique (RAO)/60° left anterior oblique (LAO) fluoroscopic view angles were adjusted to the intracardiac catheter positions as outlined before,7 resulting in angles of 43° ⫾ 9° in the RAO view and 50° ⫾ 8° in the LAO view. All procedures were performed under general anesthesia with propofol and mechanical ventilation. Transseptal catheterization was performed using two transseptal sheaths (SR 0, Daig Corp, Minnetonka, MN). Intravenous heparin was given to maintain an activated clotting time of 250 –350 seconds during the procedure. Selective angiography of all four PVs was performed before ablation by manual injection of 10 mL of contrast agent during apnea. Ablation was performed using temperature feedback with a target temperature of 50°C and a maximum power output of 30 W. The endpoint for PV isolation was the creation of bidirectional conduction block from atrium to PV and vice versa. The presence of adenosine triphosphate (ATP)-induced dormant conduction was systematically evaluated and ablated using additional radiofrequency applications when present. Linear ablation lesions from the left inferior PV toward the mitral valve and in the roof of the LA were applied in 22 patients (37%) if AF was persistent and/or still inducible after PV isolation. The inferior flutter isthmus between the tricuspid ring and inferior caval vein orifice was ablated in 42 (70%) of 60 patients during the same procedure. Integration of 3D model with biplane fluoroscopic images Integration of the CT-based 3D models of the LA and PVs with biplane fluoroscopic images was achieved on a personal computer with custom in-house developed software (LARCA, Leuven Augmented Reality Catheter Ablation).2,8 Fluoroscopic images were captured from the Siemens Coroskop system to the personal computer as a standard composite video signal and A/D converted. After removal of nonlinear distortion, angiographic images of the four PVs were combined in each plane into a single angiographic image by taking the minimum of corresponding pixels of the four images (Figure 1). This combined angiographic image served as reference for the registration of the 3D LA model into the fluoroscopic framework. Two different registration strategies were evaluated (Figure 2): Visual matching In a first step, a calibration of the fluoroscopic images is computed after manually marking seven electrodes with known interelectrode distances on the coronary sinus catheter in the RAO and LAO planes. This allows the 3D model to be introduced into the fluoroscopic images in its correct dimensions. In a second step, a rigid registration (translation and rotation) of the 3D model is used to visually align it with the angiographic reference image of the PVs in each plane. The initial rotation of the 3D model is determined by the fluoroscopic view angles. Additional manual translation and rotation is achieved by
Ector et al
Biplane 3D-Augmented Fluoroscopy for Ablation of AF
959
Figure 1 Construction of a combined angiographic reference image for biplane 3D fluoroscopy image integration. Selective angiographic images of the four PVs are acquired during apnea and combined by taking the minimum of corresponding pixels of the four images in each imaging plane. The position of the coronary sinus catheter is used to select images acquired immediately before onset of atrial contraction. RSPV: right superior PV, RIPV: right inferior PV, LSPV: left superior PV, LIPV: left inferior PV.
means of a 3D SpaceMouse (3Dconnexion Inc., Los Gatos, CA; Figure 2, upper panels). Landmark-based registration This approach is based on manual marking of corresponding landmarks on the combined angiographic image and the 3D model seen from a corresponding view angle. A variable number of correspondences is marked only at clearly identifiable landmarks such as PV
bifurcations and PV ostia (Figure 2, lower panels). Based on the marked correspondences, a 3D-2D registration function is automatically computed that minimizes the distances between the projection of landmarks on the 3D model and their correspondences on the angiographic image.9 A minimum of three corresponding landmarks is needed to calculate the 3D-2D registration function.
Figure 2 Different registration approaches used for 3D-augmented fluoroscopy. The upper panel is an illustration of the visual matching registration approach. After calibration of the fluoroscopic images by marking seven electrodes on the coronary sinus catheter, a rigid registration is performed by manually translating and rotating the integrated 3D model with a 3D spacemouse to align it with the angiographic reference image. The lower panel is an illustration of the landmark-based registration approach. A number of correspondences are marked at clearly identifiable landmarks on the angiographic image (left) and on the 3D model (right) shown from a corresponding view angle. A 3D-2D registration function is then automatically computed to register the 3D model with the angiographic reference image.
960
Heart Rhythm, Vol 5, No 7, July 2008
Figure 3 Quantitative assessment of 3D fluoroscopy integration accuracy. A: The alignment error is measured for each PV as the distance in millimeters between the PV ostium on the angiographic image (blue broken line) and its position on the 3D model after integration. In this case, the alignment error for the right superior PV is 5 mm in the RAO plane. B: Alignment scores are given to each PV to reflect the 3D angiographic integration accuracy in its more distal course. An alignment score of 0 reflects perfect alignment of the main PV and all side branches. An alignment score of 1 reflects angular deviation of one PV side branch but perfect alignment of main PV and all other side branches; an alignment score of 2 reflects angular deviation of two or more PV side branches; an alignment score of 3 reflects angular deviation of both the main PV and two or more side branches. The deviation of the distal PVs is illustrated for different scores by the yellow broken lines on an angiographic image of the right superior PV in the LAO plane. RSPV: right superior PV.
The accuracy of the 3D fluoroscopy integration process was quantitatively assessed by measuring for each PV the alignment error (in millimeters) between the PV ostium on the angiographic image and its position on the integrated 3D model (Figure 3A). Moreover, an alignment score (0 –3) was given to each PV to describe the 3D angiographic alignment of the PV in its more distal course. Different alignment scores are defined and illustrated in Figure 3B. The global in-plane alignment error and score were defined as the sum of alignment errors respectively scores of all four PVs (for each imaging plane). Integration accuracy in a plane was very stringently defined to be “optimal” if global in-plane alignment error and score were both ⱕ4.
Clinical use of biplane 3D-augmented fluoroscopy to guide AF ablation After performing the integration of the 3D model with the biplane fluoroscopic images, real time fluoroscopy could be shown together with a semitransparent 3D overlay in an
augmented reality view on the PC. In the last 30 patients (3D⫹ group), the 3D-augmented view was displayed on a dual monitor in the intervention room and was used to guide ablation. The intervention was facilitated by showing 3D ablation target lines on the integrated 3D models (Figure 4 and movie in online data supplement). Integration in the 3D⫹ group was based on the landmark-based registration approach. In the first 30 patients (2D group), ablation target sites were defined by angiography and Lasso position, and real time visualization of fluoroscopic catheter position relative to the 3D model was not available during the procedure. Procedures in the 2D and 3D⫹ group were compared regarding fluoroscopy time, procedure duration, and total radiofrequency ablation time (minutes) and radiofrequency ablation energy (Joules) needed to achieve complete PV isolation. All procedures were performed by the same experienced operator (⬎150 procedures), which excludes a potential learning curve bias.
Ector et al
Biplane 3D-Augmented Fluoroscopy for Ablation of AF
961
Figure 4 Screenshot from the software for biplane 3D-augmented fluoroscopy taken during ablation of the right inferior PV. The integrated 3D model of the LA is shown as an orange semitransparent surface in RAO and LAO views after landmark-based registration. The anterior part of the LA is clipped away to clearly show the PV ostia and posterior LA surface. Yellow dotted circles indicate the circumference of the PV ostia and are used as ablation target lines. The purplemarked point indicates the current position of the ablation catheter in the augmented fluoroscopy views. The corresponding 3D position of the ablation catheter is shown on the LA surface in the 3D viewer. This process is illustrated for different ablation catheter positions in the accompanying movie file.
Evaluation of patient radiation dose The effective patient radiation dose was calculated both for the ablation procedure and for the preprocedural CT imaging. During AF ablation, the radiation dose was quantified with dose-area product (DAP) meters incorporated in the fluoroscopy unit. These DAP meters use an air-ionization chamber mounted just beyond the X-ray collimators and integrate exposure over the entire image field. DAP was recorded as a total for both X-ray tubes. The conversion of these DAP values to effective patient radiation dose was done with conversion factors previously calculated and published for AF ablation procedures in our center.10 Different conversion factors were used for normal weight, overweight, and obese patients. The effective radiation dose due to cardiac CT examinations was calculated by multiplying the CT dose-length product with the conversion factor for chest examinations (0.017 mSv mGy⫺1 cm⫺1) according to the European guidelines on quality criteria for CT.11
Statistical analysis Summary values are given as mean ⫾ standard deviation or median [interquartile range (IQR)] for not normally distributed values. The Shapiro-Wilk W-test was used to test for normality. Comparisons of not normally distributed data were performed with a Wilcoxon matched-pairs test for dependent samples and a Mann-Whitney U-test for independent samples. Normally distributed data were compared with a paired or unpaired t-test for dependent or indepen-
dent samples, respectively. P ⱕ.05 was considered statistically significant. The local ethics committee approved this study, and informed consent was obtained from all patients. The authors had full access to the data and take responsibility for its integrity. All authors have read and agree to the manuscript as written.
Results 1. Integration accuracy of biplane 3D-augmented fluoroscopy Alignment errors and alignment scores were measured for all four PVs in both RAO and LAO planes in 57 of 60 patients. In the other three patients, an angiography of one of the PVs was of insufficient quality to calculate the integration accuracy for that PV. Registration and evaluation of integration accuracy was based on the three other veins in these patients. Total time needed to perform the visual matching registration was 130 ⫾ 45 seconds versus 161 ⫾ 31 seconds for landmark registration in both planes (P ⬍.001). When the visual matching registration approach was limited to calibration of the fluoroscopic images and translation of the 3D model without additional manual rotation, large alignment errors occurred in both planes (global in-plane alignment error RAO: 12 [IQR 6 –23] mm, score 7 [range 5– 8]; LAO error 16 mm [IQR 7–22] mm, score 6 [IQR 3– 8]; Table 1). Additional manual rotation using the 3D mouse reduced the global RAO alignment error and score to 7 (IQR 3–10) mm and 4
962
Heart Rhythm, Vol 5, No 7, July 2008
Table 1 Integration accuracy of 3D-augmented fluoroscopy according to registration approach and preprocedural image acquisition characteristics RAO
Visual matching: Translation only Translation ⫹ rotation Landmark registration: ECG gating: ⫹ ⫺ Respiration: Expiration Inspiration
LAO
Alignment error
Alignment score
n
Median
[IQR]
Median
[IQR]
% Optimal integration
Alignment error
Alignment score
Median
[IQR]
Median
[IQR]
% Optimal integration
60 60 60
12 6.5 2*
[6–23] [3–10] [0.5–5]
7 4 3*
[5–8] [3–6] [2–4]
11.7 35.0 63.3*
15.5 4 2*
[7–22] [2–8] [0–4]
6 3 1*
[3–8] [1–4] [0–3]
8.3 48.3 80.0*
31 29
2§ 2
[0–4] [2–6]
3 3
[2–3] [2–5]
77.4§§ 48.3
2 2
[0–3] [0–4]
2 1
[0–3] [0–3]
77.4 82.8
49 24
2 2
[1–5] [1–4.5]
3 3
[2–4] [2.5–5]
59.2 62.5
1 3
[0–3] [0–4.5]
1 2
[0–3] [0.5–3]
83.7 70.8
Global alignment error and alignment score in right-anterior oblique (RAO) and left-anterior oblique (LAO) views for different registration strategies. Landmark-based registration results in significantly better integration accuracy than a visual matching registration approach (* marks P-values ⱕ0.001 vs. visual matching approach using translation and rotation). Landmark-based registration results in an ‘optimal’ integration in 63% and 80% of patients in RAOand LAO-views respectively. CT-acquisition during inspiration vs. expiration and ECG-gated CT-acquisition vs. no gating had no clinically relevant effects on integration accuracy (§P-value ⫽ .05, §§P-value ⬍.001 vs. non-gated CT-acquisition).
(IQR 3– 6) mm, respectively, and the LAO alignment error and score to 4 mm (IQR 2– 8) mm and 3 (IQR 1– 4), respectively. This corresponded to an “optimal” registration accuracy in 35% of patients for the RAO plane and 52% in the LAO plane. The landmark registration approach was based on manual marking of 11 ⫾ 2 corresponding landmarks in both the RAO (range 5–16) and LAO (range 7–16) planes. The RAO global alignment error and score were 2 mm (range 0.5–5 mm) and 3 (range 2– 4), respectively, versus an alignment error of 2 mm (range 0 – 4 mm) and score of 1 (range 0 –3) in the LAO plane. An optimal registration accuracy was achieved with this registration approach in 63% of patients in the RAO plane versus in 80% in the LAO plane (P ⫽ .04). The landmark-based registration approach therefore significantly reduced alignment errors and scores in both imaging planes when compared with the visual matching registration approach (P ⬍.001 for all accuracy parameters). The effects of ECG gating and respiratory phase during CT acquisition on integration accuracy were small and clinically irrelevant when landmark-based registration is used (Table 1). Only the alignment error in the RAO plane showed a borderline significant reduction for ECG-gated versus non-gated CT acquisitions, leading to a higher percentage of optimal integration accuracy in the RAO plane (77% vs. 48%; P ⫽ .001). A change in cardiac rhythm status (from sinus rhythm to AF or vice versa) between preprocedural CT and per-procedural angiography occurred in eight patients and did not result in a decreased alignment error or alignment score in either imaging plane; neither was image integration accuracy reduced in patients in AF during CT acquisition (n ⫽ 10).
2. Clinical utility of 3D-augmented fluoroscopy The use of 3D-augmented fluoroscopy in the 3D⫹ group was gauged as extremely helpful by the operator during
catheter navigation and ablation. The optimal ablation trajectory, as shown by the 3D target lines on the integrated 3D models, could often not be inferred from PV angiography or Lasso position and therefore contributed to a safer ablation procedure. Procedures in the 3D⫹ group were associated with a significantly shorter fluoroscopy time than procedures in the 2D group (61 ⫾ 18 minutes vs. 77 ⫾ 26 minutes; P ⫽ .009), and there was a trend toward shorter total procedure duration (230 ⫾ 67 minutes vs. 257 ⫾ 58 minutes; P ⫽ .06). These reductions occurred despite a higher rate of additional linear lesions in the 3D⫹ group (13/30 vs. 9/30 patients) and an identical rate of cavotricuspid isthmus ablations in both groups (21/30 patients). Complete bidirectional PV isolation was achieved in 117 (98%) of 119 targeted veins in the 2D group and 118 (98%) of 120 targeted veins in the 3D⫹ group. The subjective feeling of superior lesion deployment based on 3D target lesions was not associated with a significant reduction in the total radiofrequency ablation time/energy needed to achieve complete PV isolation (ablation time 57.2 ⫾ 28 minutes vs. 57.5 ⫾ 23 minutes; P ⫽ .94; ablation energy 65,887 ⫾ 32,129 J vs. 69,592 ⫾ 26,132 J; P ⫽ .63 for 3D⫹ and 2D groups, respectively). There were no procedure-related complications in the 3D⫹ group. Complications in the 2D group consisted of a ⬎50% PV stenosis of the right-inferior PV in one patient and an arteriovenous fistula in the groin in another patient.
3. Radiation dose associated with 3D-augmented fluoroscopy There were no significant differences in the DAP values and patient effective doses associated with procedures in the 3D⫹ and 2D group (DAP 78.6 ⫾ 35.0 Gy cm2 vs. 88.5 ⫾ 56.8 Gy cm2, P ⫽ .41; effective dose 14.2 ⫾ 9.7 mSv vs. 16.2 ⫾ 9.7 mSv; P ⫽ .33, in 3D⫹ and 2D groups, respectively).
Ector et al
Biplane 3D-Augmented Fluoroscopy for Ablation of AF
There was a very important and significant difference in the patient effective dose associated with ECG-gated versus non-gated preprocedural cardiac CT (17.3 ⫾ 5.2 mSv vs. 4.4 ⫾ 3 mSv; P ⬍.001). The use of nongated cardiac CT for biplane 3D-augmented fluoroscopy in the 3D⫹ group therefore resulted in a large decrease in the combined patient radiation dose from procedure plus CT scan. Whereas the total patient radiation dose in the 2D group was 34.6 ⫾ 12.7 mSv, patients in the 3D⫹ group received a total effective dose of 18.8 ⫾ 8.4 mSv (P ⬍.001; 54% of the dose in the 2D group), which is comparable to the effective dose of the ECG-gated CT scan alone in the 2D group.
Discussion This is the first report describing the use of biplane 3Daugmented fluoroscopy as a single navigation tool for AF ablation. Moreover, our work contains the first quantitative evaluation of the integration accuracy of 3D-augmented fluoroscopy based on angiographic images of the four PVs for completeness. Our data show the importance of the registration approach itself for achieving accurate and useful 3D fluoroscopy image integration. A landmark-based registration approach significantly improved integration accuracy in both imaging planes when compared with a visual matching approach based on free manual translation and rotation of 3D LA models. For each registration approach, integration in the LAO plane was more accurate than in the RAO plane. This finding is probably caused by the higher superposition of angiographic structures in the RAO plane, complicating optimal 3D fluoroscopy registration. However, the observed global alignment error of 2 mm in the RAO plane still compares favorably to the other registration approaches in any imaging plane and is certainly acceptable for clinical use. In contrast to the important influence of registration strategy, the use of ECG gating and the respiratory phase during preprocedural CT acquisition had only a minor impact on the 3D fluoroscopy integration accuracy. ECGgated CT only resulted in a borderline significant (but clinically irrelevant) improvement of the integration alignment error in the RAO plane but not in the LAO plane. Given the high patient radiation doses associated with ECG-gated cardiac CT, the possibility of using nongated CT acquisitions without appreciably affecting integration accuracy is an important advantage for 3D-augmented fluoroscopy. The total patient effective radiation dose due to preprocedural nongated CT and 3D-augmented biplane fluoroscopy during the ablation procedure in our study was 18.8 ⫾ 8.4 mSv. This combined effective dose came close to the effective dose estimated for ECG-gated cardiac CT acquisition alone (17.3 ⫾ 5.2 mSv). The use of biplane 3D-augmented fluoroscopy as a single navigation tool for AF ablation therefore most likely does not increase the patient radiation dose when compared with ablation procedures of any kind that are preceded by ECG-gated CT acquisition. Rather than a true positive argument for using 3D-augmented fluoros-
963
copy, these results should encourage the use of nongated cardiac CT or other low-dose alternatives (such as cardiac MRI or CT with prospective ECG gating12) as preprocedural imaging tools before AF ablation. Although data on the accuracy of 3D image integration based on ECG-gated versus ECG-non-gated CT are currently not available for 3D electroanatomic mapping systems, some indirect evidence supports the use of nongated CT for accurate 3D image integration in this setting as well.13 The use of 3D-augmented fluoroscopy was felt to be very helpful and provided additional 3D information to guide catheter navigation and ablation during the procedures. The significant decrease in fluoroscopy time and trend toward shorter procedure times found in our study, however, need to be interpreted cautiously as they are based on nonrandomized data. Nevertheless, the same findings were noted in a study from Sra et al.,4 in which 50 consecutive patients were randomized to AF ablation procedures guided by electroanatomic mapping alone or in combination with monoplane CT fluoroscopy image integration. These investigators also reported a significant decrease in procedure duration and fluoroscopy times in the group with CT fluoroscopy guidance. A prospective randomized evaluation of the procedural characteristics and clinical outcome of biplane 3D-augmented fluoroscopy as a single navigation tool for AF ablation is currently ongoing in our center. An advantage of our implementation is that it has been developed independently in an academic environment and can be used with any type of fluoroscopy system and images from any type of CT or MRI unit. New techniques such as cardiac C-arm CT have the potential to further improve the workflow and integration accuracy of 3D-augmented fluoroscopy.14,15 By acquiring CT images in the electrophysiology room with the patient in the same position on the same table, more accurate registration approaches can further enhance 3D fluoroscopy image integration. Moreover, this approach eliminates possible differences in cardiac loading conditions between imaging and ablation. Compensation of catheter movement due to cardiac and respiratory motion during 3D-augmented fluoroscopy are other important fields currently under investigation.16 Similar to electroanatomic mapping systems, ECG-gated fluoroscopic image acquisition can be used to eliminate catheter movement due to cardiac motion, while at the same time reducing the procedural radiation dose. Correction of fluoroscopic catheter positions for respiratory movements is currently not possible during integration of 3D models with live fluoroscopy. The resulting drawback for 3D fluoroscopy integration is a mismatch between catheter position and 3D overlay during a part of the respiratory cycle. As the 3D LA models were registered to an angiographic image of the LA acquired during apnea (i.e., expiration), the operator had to take into account a periodic downward movement of the ablation catheter during inspiration. The 3D position of the ablation catheter tip relative to the integrated 3D model was therefore preferentially
964 assessed during expiration. Instead of such an intuitive approach, adapting the position and/or configuration of an integrated 3D LA PV model to patient respiration could improve the accuracy of fluoroscopic image integration and potentially allows more accurate lesion placement. In patients under general anesthesia, special high-frequency, low-volume ventilation modes could also be used to minimize the detrimental effects of respiratory motion on image integration.17
Study limitations The quantitative evaluation of 3D image integration accuracy is difficult owing to the lack of a gold standard, that is, knowledge of the true 3D position of LA and PVs in the patient during the procedure. However, measurement of alignment errors with angiographic images of the PVs reflects the integration accuracy in the area of interest and is in our opinion the most reliable method to assess integration accuracy relevant for ablation. Since small rotational errors are not reflected in alignment errors with the angiographic PV ostia, we opted to also score alignment based on the PVs in their more distal course. We did not test the integration accuracy of other cardiac chambers or vascular structures as this would require separate angiographic injections to verify integration accuracy. Ablation procedures in our study were associated with long fluoroscopy and procedure times. This can be explained by the use of relatively low-power ablation settings to increase procedural safety, the systematic ablation of ATP-induced dormant conduction, and the additional cavotricuspid isthmus ablation in 70% of patients. Low frame rate pulsed fluoroscopy was used to reduce the patient radiation dose associated with fluoroscopy. We consider the patient radiation doses with our approach (fluoroscopy 3 frames/second, nongated CT) certainly within acceptable limits.10 As mentioned, there are still many opportunities to further reduce it (e.g., preprocedural MRI, ECG-gated fluoroscopy). Freedom of AF was not evaluated as a clinical endpoint in our study. However, procedural endpoints were the same for all patients in the 2D and 3D⫹ group, and large differences in clinical outcome are therefore not expected.
Conclusion Biplane 3D-augmented fluoroscopy can be used as a safe and accurate stand-alone method to guide AF ablation procedures. Preprocedural imaging with non-ECG-gated cardiac CT results in an important reduction in patient radiation dose without substantially reducing 3D image integration accuracy.
Heart Rhythm, Vol 5, No 7, July 2008
Appendix Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.hrthm.2008.03.024.
References 1. Calkins H, Brugada J, Packer DL, et al. HRS/EHRA/ECAS expert consensus statement on catheter and surgical ablation of atrial fibrillation: recommendations for personnel, policy, procedures and follow-up. A report of the Heart Rhythm Society (HRS) Task Force on catheter and surgical ablation of atrial fibrillation. Heart Rhythm 2007;4:816 – 861. 2. Ector J, De Buck S, Adams J, et al. Cardiac three-dimensional magnetic resonance imaging and fluoroscopy merging: a new approach for electroanatomic mapping to assist catheter ablation. Circulation 2005;112:3769 –3776. 3. Sra J, Krum D, Malloy A, et al. Registration of three-dimensional left atrial computed tomographic images with projection images obtained using fluoroscopy. Circulation 2005;112:3763–3768. 4. Sra J, Narayan G, Krum D, et al. Computed tomography-fluoroscopy image integration-guided catheter ablation of atrial fibrillation. J Cardiovasc Electrophysiol 2007;18:409 – 414. 5. Maksimoviç R, Cademartiri F, Scholten M, et al. Sixteen-row multislice computed tomography in the assessment of pulmonary veins prior to ablative treatment: validation vs conventional pulmonary venography and study of reproducibility. Eur Radiol 2004;14:369 –374. 6. Noseworthy PA, Malchano ZJ, Ahmed J, et al. The impact of respiration on left atrial and pulmonary venous anatomy: implications for image-guided intervention. Heart Rhythm 2005;2:1173–1178. 7. Heidbuchel H, Ector H, Adams J, et al. Use of only a regular diagnostic His-bundle catheter for both fast and reproducible “para-Hisian pacing” and stable right ventricular pacing. J Cardiovasc Electrophysiol 1997;8:1121–1132. 8. De Buck S, Maes F, Ector J, et al. An augmented reality system for patientspecific guidance of cardiac catheter ablation procedures. IEEE Trans Med Imaging 2005;24:1512–1524. 9. De Buck S, Ector J, Maes F, et al. Interactive augmented reality for cardiac ablation therapy in the left and right atrium. Workshop proceedings of the 10th international conference on medical image computing and computer assisted intervention—MICCAI 2007, workshop on interaction in medical image analysis and visualization 2007;65–72. 10. Ector J, Dragusin O, Adriaenssens B, et al. Obesity is a major determinant of radiation dose in patients undergoing pulmonary vein isolation for atrial fibrillation. J Am Coll Cardiol 2007;50:234 –242. 11. European Commission. European guidelines on quality criteria for computed tomography, EUR 16262EN. Luxembourg: Office for Official Publications of the European Communities 2000. 12. Hsieh J, Londt J, Vass M, et al. Step-and-shoot data acquisition and reconstruction for cardiac x-ray computed tomography. Med Phys 2006;33:4236 – 4248. 13. Dong J, Dalal D, Scherr D, et al. Impact of heart rhythm status on registration accuracy of the left atrium for catheter ablation of atrial fibrillation. J Cardiovasc Electrophysiol 2007;18:1269 –1276. 14. Lauritsch G, Boese J, Wigstrom L, et al. Towards cardiac C-arm computed tomography. IEEE Trans Med Imaging 2006;25:922–934. 15. Thiagalingam A, Manzke R, D’Avila A, et al. Intraprocedural volume imaging of the left atrium and pulmonary veins with rotational X-ray angiography: implications for catheter ablation of atrial fibrillation. J Cardiovasc Electrophysiol 2008;19:293–300. 16. Ector J, De Buck S, Loeckx D, et al. Absolute and relative changes in left atrial and pulmonary venous anatomy due to respiration: relevance for 3D image integration during ablation procedures for atrial fibrillation. J Cardiovasc Electrophysiol 2008; March 26 [Epub ahead of print]. 17. Goode J, Taylor RL, Buffington CW, et al. High-frequency jet ventilation: utility in posterior left atrial catheter ablation. Heart Rhythm 2006;3: 13–19.
Introduction Preprocedural imaging and three-dimensional (3D) reconstruction of the left atrium (LA) and pulmonary veins (PVs) is performed in the majority of centers before atrial fibrillation (AF) ablation procedures.1 The detailed patient-specific anatomical information can help achieve a more efficient and successful ablation and may prevent procedure-related complications. In addition, nonfluoroscopic 3D mapping
Hein Heidbüchel is a member of the scientific advisory board of Biosense Webster, Inc., and St. Jude Medical, Inc. The other authors have no conflicts of interest to declare. Stijn De Buck is funded through the Institute for the Promotion of Innovation by Science and Technology in Flanders of the Flemish government (project OZM 080511). Address reprint requests and correspondence: Joris Ector, M.D., Cardiology, University Hospital Gasthuisberg, Herestraat 49, B-3000 Leuven, Belgium, Europe. E-mail address: [email protected]. (Received January 24, 2008; accepted March 18, 2008.)
effects of ECG gating and respiratory phase during CT acquisition on integration accuracy were small and clinically irrelevant. The use of 3D-augmented fluoroscopy in the 3D⫹ group was gauged as extremely helpful by the operator. It resulted in a significant reduction of fluoroscopy time (61 ⫾ 18 minutes vs. 77 ⫾ 26 minutes; P ⫽ .009) and a trend toward shorter procedure duration (230 ⫾ 67 minutes vs. 257 ⫾ 58 minutes; P ⫽ .06) versus conventional procedures. The systematic use of nongated cardiac CT in the 3D⫹ group resulted in an important reduction in total effective patient radiation dose due to CT⫹fluoroscopy (4 ⫹ 14 ⫽ 18 ⫾ 8 mSv vs.17 ⫹ 16 ⫽ 33 ⫾ 13 mSv; P ⬍.001). CONCLUSIONS Biplane 3D-augmented fluoroscopy can be used as a safe and accurate stand-alone method to guide AF ablation procedures. The use of nongated cardiac CT substantially reduces total patient radiation dose without a relevant reduction in integration accuracy. KEYWORDS Atrial fibrillation; Ablation; Image integration; Fluoroscopy; Computed tomography (Heart Rhythm 2008;5:957–964) © 2008 Heart Rhythm Society. All rights reserved.
systems allow the integration of this 3D anatomical information during the ablation procedure to directly guide catheter navigation and ablation. Recently, integration of 3D models of the LA and PVs has also been applied to fluoroscopic imaging. In 3D-augmented fluoroscopy, computed tomography (CT)/magnetic resonance imaging (MRI)-based 3D models are shown as a semitransparent overlay on fluoroscopic images, using specific calibration and registration procedures.2,3 When applied to fluoroscopy in a single plane, this method provides a two-dimensional (2D) projection of fluoroscopic catheter positions on the integrated 3D overlay but does not allow determination of 3D catheter position unless continuosly changing the fluoroscopic view angles. Monoplane 3Daugmented fluoroscopy has therefore only been reported as an additional tool used in combination with electroanatomic mapping to guide AF ablation procedures.4
1547-5271/$ -see front matter © 2008 Heart Rhythm Society. All rights reserved.
doi:10.1016/j.hrthm.2008.03.024
958 Given the limitations of monoplane 3D-augmented fluoroscopy, our research group developed new methods for real time biplane integration of 3D LA models with fluoroscopic images. In this work, the integration accuracy of our approach is quantitatively assessed, and the influence of different registration methods and preprocedural imaging parameters is evaluated. Moreover, biplane 3D-augmented fluoroscopy was used as a single navigation tool during AF ablation to evaluate its clinical value and possible drawbacks, such as the associated patient radiation dose.
Methods Study population Our study included 60 patients (54 men and 6 women, age 53 ⫾ 7 years) undergoing a first ablation procedure because of symptomatic, drug-refractory AF. AF was paroxysmal in 41 (68%) and persistent in 19 (32%) patients. Structural heart disease was present in only three patients (hypertrophic cardiomyopathy in one, myocardial infarction in two). Mean patient weight, height, and body mass index were 83 ⫾ 11 kg, 179 ⫾ 9 cm, and 26.1 ⫾ 2.7 kg/m2 respectively.
Preprocedural imaging and 3D reconstruction All patients underwent cardiac CT on the day before the ablation procedure with a 64-slice multidetector scanner. The first 31 patients underwent cardiac CT with retrospective electrocardiogram (ECG) gating on a Siemens Somatom Sensation 64 scanner Siemens AG, Medical Solutions, Forchheim, Germany. The 29 following patients underwent non-ECG-gated cardiac CT on a Philips Brilliance 64 scanner (Philips Medical Systems, Eindhoven, The Netherlands), to examine the influence of ECG gating on 3D integration accuracy and patient radiation dose.5 Sixty to 80 mL of iodine contrast agent (Iomeron 400, Bracco, Milan, Italy) were administered intravenously in all patients at a rate of 5 mL/s to obtain contrast opacification of the LA and PVs. To examine the influence of respiration during preprocedural imaging on 3D fluoroscopy integration accuracy,6 CT acquisition was performed during inspiration in 11/31 ECG-gated acquisitions and in both inspiration and expiration in 13/29 nongated CT acquisitions. All other patients underwent CT imaging during expiration only. The resulting axial images were imported into a 3D reconstruction software package (Amira 4.1, Mercury Computer Systems SAS). Intensity-based segmentation of the LA and PVs was performed to reconstruct a detailed 3D LA-PV surface model.
Ablation procedures and 3D model integration General procedural aspects Ablation procedures consisted of Lasso-guided (Lasso, Biosense Webster Diamond Bar, CA) electrical isolation of the four PVs. All procedures were performed under fluoroscopic guidance with a Siemens Coroskop C biplane image intensifier system. The system was routinely set to pulsed
Heart Rhythm, Vol 5, No 7, July 2008 fluoroscopy at 3 frames per second and a cine frame rate of 12.5 frames per second to minimize radiation exposure. Standard 30° right anterior oblique (RAO)/60° left anterior oblique (LAO) fluoroscopic view angles were adjusted to the intracardiac catheter positions as outlined before,7 resulting in angles of 43° ⫾ 9° in the RAO view and 50° ⫾ 8° in the LAO view. All procedures were performed under general anesthesia with propofol and mechanical ventilation. Transseptal catheterization was performed using two transseptal sheaths (SR 0, Daig Corp, Minnetonka, MN). Intravenous heparin was given to maintain an activated clotting time of 250 –350 seconds during the procedure. Selective angiography of all four PVs was performed before ablation by manual injection of 10 mL of contrast agent during apnea. Ablation was performed using temperature feedback with a target temperature of 50°C and a maximum power output of 30 W. The endpoint for PV isolation was the creation of bidirectional conduction block from atrium to PV and vice versa. The presence of adenosine triphosphate (ATP)-induced dormant conduction was systematically evaluated and ablated using additional radiofrequency applications when present. Linear ablation lesions from the left inferior PV toward the mitral valve and in the roof of the LA were applied in 22 patients (37%) if AF was persistent and/or still inducible after PV isolation. The inferior flutter isthmus between the tricuspid ring and inferior caval vein orifice was ablated in 42 (70%) of 60 patients during the same procedure. Integration of 3D model with biplane fluoroscopic images Integration of the CT-based 3D models of the LA and PVs with biplane fluoroscopic images was achieved on a personal computer with custom in-house developed software (LARCA, Leuven Augmented Reality Catheter Ablation).2,8 Fluoroscopic images were captured from the Siemens Coroskop system to the personal computer as a standard composite video signal and A/D converted. After removal of nonlinear distortion, angiographic images of the four PVs were combined in each plane into a single angiographic image by taking the minimum of corresponding pixels of the four images (Figure 1). This combined angiographic image served as reference for the registration of the 3D LA model into the fluoroscopic framework. Two different registration strategies were evaluated (Figure 2): Visual matching In a first step, a calibration of the fluoroscopic images is computed after manually marking seven electrodes with known interelectrode distances on the coronary sinus catheter in the RAO and LAO planes. This allows the 3D model to be introduced into the fluoroscopic images in its correct dimensions. In a second step, a rigid registration (translation and rotation) of the 3D model is used to visually align it with the angiographic reference image of the PVs in each plane. The initial rotation of the 3D model is determined by the fluoroscopic view angles. Additional manual translation and rotation is achieved by
Ector et al
Biplane 3D-Augmented Fluoroscopy for Ablation of AF
959
Figure 1 Construction of a combined angiographic reference image for biplane 3D fluoroscopy image integration. Selective angiographic images of the four PVs are acquired during apnea and combined by taking the minimum of corresponding pixels of the four images in each imaging plane. The position of the coronary sinus catheter is used to select images acquired immediately before onset of atrial contraction. RSPV: right superior PV, RIPV: right inferior PV, LSPV: left superior PV, LIPV: left inferior PV.
means of a 3D SpaceMouse (3Dconnexion Inc., Los Gatos, CA; Figure 2, upper panels). Landmark-based registration This approach is based on manual marking of corresponding landmarks on the combined angiographic image and the 3D model seen from a corresponding view angle. A variable number of correspondences is marked only at clearly identifiable landmarks such as PV
bifurcations and PV ostia (Figure 2, lower panels). Based on the marked correspondences, a 3D-2D registration function is automatically computed that minimizes the distances between the projection of landmarks on the 3D model and their correspondences on the angiographic image.9 A minimum of three corresponding landmarks is needed to calculate the 3D-2D registration function.
Figure 2 Different registration approaches used for 3D-augmented fluoroscopy. The upper panel is an illustration of the visual matching registration approach. After calibration of the fluoroscopic images by marking seven electrodes on the coronary sinus catheter, a rigid registration is performed by manually translating and rotating the integrated 3D model with a 3D spacemouse to align it with the angiographic reference image. The lower panel is an illustration of the landmark-based registration approach. A number of correspondences are marked at clearly identifiable landmarks on the angiographic image (left) and on the 3D model (right) shown from a corresponding view angle. A 3D-2D registration function is then automatically computed to register the 3D model with the angiographic reference image.
960
Heart Rhythm, Vol 5, No 7, July 2008
Figure 3 Quantitative assessment of 3D fluoroscopy integration accuracy. A: The alignment error is measured for each PV as the distance in millimeters between the PV ostium on the angiographic image (blue broken line) and its position on the 3D model after integration. In this case, the alignment error for the right superior PV is 5 mm in the RAO plane. B: Alignment scores are given to each PV to reflect the 3D angiographic integration accuracy in its more distal course. An alignment score of 0 reflects perfect alignment of the main PV and all side branches. An alignment score of 1 reflects angular deviation of one PV side branch but perfect alignment of main PV and all other side branches; an alignment score of 2 reflects angular deviation of two or more PV side branches; an alignment score of 3 reflects angular deviation of both the main PV and two or more side branches. The deviation of the distal PVs is illustrated for different scores by the yellow broken lines on an angiographic image of the right superior PV in the LAO plane. RSPV: right superior PV.
The accuracy of the 3D fluoroscopy integration process was quantitatively assessed by measuring for each PV the alignment error (in millimeters) between the PV ostium on the angiographic image and its position on the integrated 3D model (Figure 3A). Moreover, an alignment score (0 –3) was given to each PV to describe the 3D angiographic alignment of the PV in its more distal course. Different alignment scores are defined and illustrated in Figure 3B. The global in-plane alignment error and score were defined as the sum of alignment errors respectively scores of all four PVs (for each imaging plane). Integration accuracy in a plane was very stringently defined to be “optimal” if global in-plane alignment error and score were both ⱕ4.
Clinical use of biplane 3D-augmented fluoroscopy to guide AF ablation After performing the integration of the 3D model with the biplane fluoroscopic images, real time fluoroscopy could be shown together with a semitransparent 3D overlay in an
augmented reality view on the PC. In the last 30 patients (3D⫹ group), the 3D-augmented view was displayed on a dual monitor in the intervention room and was used to guide ablation. The intervention was facilitated by showing 3D ablation target lines on the integrated 3D models (Figure 4 and movie in online data supplement). Integration in the 3D⫹ group was based on the landmark-based registration approach. In the first 30 patients (2D group), ablation target sites were defined by angiography and Lasso position, and real time visualization of fluoroscopic catheter position relative to the 3D model was not available during the procedure. Procedures in the 2D and 3D⫹ group were compared regarding fluoroscopy time, procedure duration, and total radiofrequency ablation time (minutes) and radiofrequency ablation energy (Joules) needed to achieve complete PV isolation. All procedures were performed by the same experienced operator (⬎150 procedures), which excludes a potential learning curve bias.
Ector et al
Biplane 3D-Augmented Fluoroscopy for Ablation of AF
961
Figure 4 Screenshot from the software for biplane 3D-augmented fluoroscopy taken during ablation of the right inferior PV. The integrated 3D model of the LA is shown as an orange semitransparent surface in RAO and LAO views after landmark-based registration. The anterior part of the LA is clipped away to clearly show the PV ostia and posterior LA surface. Yellow dotted circles indicate the circumference of the PV ostia and are used as ablation target lines. The purplemarked point indicates the current position of the ablation catheter in the augmented fluoroscopy views. The corresponding 3D position of the ablation catheter is shown on the LA surface in the 3D viewer. This process is illustrated for different ablation catheter positions in the accompanying movie file.
Evaluation of patient radiation dose The effective patient radiation dose was calculated both for the ablation procedure and for the preprocedural CT imaging. During AF ablation, the radiation dose was quantified with dose-area product (DAP) meters incorporated in the fluoroscopy unit. These DAP meters use an air-ionization chamber mounted just beyond the X-ray collimators and integrate exposure over the entire image field. DAP was recorded as a total for both X-ray tubes. The conversion of these DAP values to effective patient radiation dose was done with conversion factors previously calculated and published for AF ablation procedures in our center.10 Different conversion factors were used for normal weight, overweight, and obese patients. The effective radiation dose due to cardiac CT examinations was calculated by multiplying the CT dose-length product with the conversion factor for chest examinations (0.017 mSv mGy⫺1 cm⫺1) according to the European guidelines on quality criteria for CT.11
Statistical analysis Summary values are given as mean ⫾ standard deviation or median [interquartile range (IQR)] for not normally distributed values. The Shapiro-Wilk W-test was used to test for normality. Comparisons of not normally distributed data were performed with a Wilcoxon matched-pairs test for dependent samples and a Mann-Whitney U-test for independent samples. Normally distributed data were compared with a paired or unpaired t-test for dependent or indepen-
dent samples, respectively. P ⱕ.05 was considered statistically significant. The local ethics committee approved this study, and informed consent was obtained from all patients. The authors had full access to the data and take responsibility for its integrity. All authors have read and agree to the manuscript as written.
Results 1. Integration accuracy of biplane 3D-augmented fluoroscopy Alignment errors and alignment scores were measured for all four PVs in both RAO and LAO planes in 57 of 60 patients. In the other three patients, an angiography of one of the PVs was of insufficient quality to calculate the integration accuracy for that PV. Registration and evaluation of integration accuracy was based on the three other veins in these patients. Total time needed to perform the visual matching registration was 130 ⫾ 45 seconds versus 161 ⫾ 31 seconds for landmark registration in both planes (P ⬍.001). When the visual matching registration approach was limited to calibration of the fluoroscopic images and translation of the 3D model without additional manual rotation, large alignment errors occurred in both planes (global in-plane alignment error RAO: 12 [IQR 6 –23] mm, score 7 [range 5– 8]; LAO error 16 mm [IQR 7–22] mm, score 6 [IQR 3– 8]; Table 1). Additional manual rotation using the 3D mouse reduced the global RAO alignment error and score to 7 (IQR 3–10) mm and 4
962
Heart Rhythm, Vol 5, No 7, July 2008
Table 1 Integration accuracy of 3D-augmented fluoroscopy according to registration approach and preprocedural image acquisition characteristics RAO
Visual matching: Translation only Translation ⫹ rotation Landmark registration: ECG gating: ⫹ ⫺ Respiration: Expiration Inspiration
LAO
Alignment error
Alignment score
n
Median
[IQR]
Median
[IQR]
% Optimal integration
Alignment error
Alignment score
Median
[IQR]
Median
[IQR]
% Optimal integration
60 60 60
12 6.5 2*
[6–23] [3–10] [0.5–5]
7 4 3*
[5–8] [3–6] [2–4]
11.7 35.0 63.3*
15.5 4 2*
[7–22] [2–8] [0–4]
6 3 1*
[3–8] [1–4] [0–3]
8.3 48.3 80.0*
31 29
2§ 2
[0–4] [2–6]
3 3
[2–3] [2–5]
77.4§§ 48.3
2 2
[0–3] [0–4]
2 1
[0–3] [0–3]
77.4 82.8
49 24
2 2
[1–5] [1–4.5]
3 3
[2–4] [2.5–5]
59.2 62.5
1 3
[0–3] [0–4.5]
1 2
[0–3] [0.5–3]
83.7 70.8
Global alignment error and alignment score in right-anterior oblique (RAO) and left-anterior oblique (LAO) views for different registration strategies. Landmark-based registration results in significantly better integration accuracy than a visual matching registration approach (* marks P-values ⱕ0.001 vs. visual matching approach using translation and rotation). Landmark-based registration results in an ‘optimal’ integration in 63% and 80% of patients in RAOand LAO-views respectively. CT-acquisition during inspiration vs. expiration and ECG-gated CT-acquisition vs. no gating had no clinically relevant effects on integration accuracy (§P-value ⫽ .05, §§P-value ⬍.001 vs. non-gated CT-acquisition).
(IQR 3– 6) mm, respectively, and the LAO alignment error and score to 4 mm (IQR 2– 8) mm and 3 (IQR 1– 4), respectively. This corresponded to an “optimal” registration accuracy in 35% of patients for the RAO plane and 52% in the LAO plane. The landmark registration approach was based on manual marking of 11 ⫾ 2 corresponding landmarks in both the RAO (range 5–16) and LAO (range 7–16) planes. The RAO global alignment error and score were 2 mm (range 0.5–5 mm) and 3 (range 2– 4), respectively, versus an alignment error of 2 mm (range 0 – 4 mm) and score of 1 (range 0 –3) in the LAO plane. An optimal registration accuracy was achieved with this registration approach in 63% of patients in the RAO plane versus in 80% in the LAO plane (P ⫽ .04). The landmark-based registration approach therefore significantly reduced alignment errors and scores in both imaging planes when compared with the visual matching registration approach (P ⬍.001 for all accuracy parameters). The effects of ECG gating and respiratory phase during CT acquisition on integration accuracy were small and clinically irrelevant when landmark-based registration is used (Table 1). Only the alignment error in the RAO plane showed a borderline significant reduction for ECG-gated versus non-gated CT acquisitions, leading to a higher percentage of optimal integration accuracy in the RAO plane (77% vs. 48%; P ⫽ .001). A change in cardiac rhythm status (from sinus rhythm to AF or vice versa) between preprocedural CT and per-procedural angiography occurred in eight patients and did not result in a decreased alignment error or alignment score in either imaging plane; neither was image integration accuracy reduced in patients in AF during CT acquisition (n ⫽ 10).
2. Clinical utility of 3D-augmented fluoroscopy The use of 3D-augmented fluoroscopy in the 3D⫹ group was gauged as extremely helpful by the operator during
catheter navigation and ablation. The optimal ablation trajectory, as shown by the 3D target lines on the integrated 3D models, could often not be inferred from PV angiography or Lasso position and therefore contributed to a safer ablation procedure. Procedures in the 3D⫹ group were associated with a significantly shorter fluoroscopy time than procedures in the 2D group (61 ⫾ 18 minutes vs. 77 ⫾ 26 minutes; P ⫽ .009), and there was a trend toward shorter total procedure duration (230 ⫾ 67 minutes vs. 257 ⫾ 58 minutes; P ⫽ .06). These reductions occurred despite a higher rate of additional linear lesions in the 3D⫹ group (13/30 vs. 9/30 patients) and an identical rate of cavotricuspid isthmus ablations in both groups (21/30 patients). Complete bidirectional PV isolation was achieved in 117 (98%) of 119 targeted veins in the 2D group and 118 (98%) of 120 targeted veins in the 3D⫹ group. The subjective feeling of superior lesion deployment based on 3D target lesions was not associated with a significant reduction in the total radiofrequency ablation time/energy needed to achieve complete PV isolation (ablation time 57.2 ⫾ 28 minutes vs. 57.5 ⫾ 23 minutes; P ⫽ .94; ablation energy 65,887 ⫾ 32,129 J vs. 69,592 ⫾ 26,132 J; P ⫽ .63 for 3D⫹ and 2D groups, respectively). There were no procedure-related complications in the 3D⫹ group. Complications in the 2D group consisted of a ⬎50% PV stenosis of the right-inferior PV in one patient and an arteriovenous fistula in the groin in another patient.
3. Radiation dose associated with 3D-augmented fluoroscopy There were no significant differences in the DAP values and patient effective doses associated with procedures in the 3D⫹ and 2D group (DAP 78.6 ⫾ 35.0 Gy cm2 vs. 88.5 ⫾ 56.8 Gy cm2, P ⫽ .41; effective dose 14.2 ⫾ 9.7 mSv vs. 16.2 ⫾ 9.7 mSv; P ⫽ .33, in 3D⫹ and 2D groups, respectively).
Ector et al
Biplane 3D-Augmented Fluoroscopy for Ablation of AF
There was a very important and significant difference in the patient effective dose associated with ECG-gated versus non-gated preprocedural cardiac CT (17.3 ⫾ 5.2 mSv vs. 4.4 ⫾ 3 mSv; P ⬍.001). The use of nongated cardiac CT for biplane 3D-augmented fluoroscopy in the 3D⫹ group therefore resulted in a large decrease in the combined patient radiation dose from procedure plus CT scan. Whereas the total patient radiation dose in the 2D group was 34.6 ⫾ 12.7 mSv, patients in the 3D⫹ group received a total effective dose of 18.8 ⫾ 8.4 mSv (P ⬍.001; 54% of the dose in the 2D group), which is comparable to the effective dose of the ECG-gated CT scan alone in the 2D group.
Discussion This is the first report describing the use of biplane 3Daugmented fluoroscopy as a single navigation tool for AF ablation. Moreover, our work contains the first quantitative evaluation of the integration accuracy of 3D-augmented fluoroscopy based on angiographic images of the four PVs for completeness. Our data show the importance of the registration approach itself for achieving accurate and useful 3D fluoroscopy image integration. A landmark-based registration approach significantly improved integration accuracy in both imaging planes when compared with a visual matching approach based on free manual translation and rotation of 3D LA models. For each registration approach, integration in the LAO plane was more accurate than in the RAO plane. This finding is probably caused by the higher superposition of angiographic structures in the RAO plane, complicating optimal 3D fluoroscopy registration. However, the observed global alignment error of 2 mm in the RAO plane still compares favorably to the other registration approaches in any imaging plane and is certainly acceptable for clinical use. In contrast to the important influence of registration strategy, the use of ECG gating and the respiratory phase during preprocedural CT acquisition had only a minor impact on the 3D fluoroscopy integration accuracy. ECGgated CT only resulted in a borderline significant (but clinically irrelevant) improvement of the integration alignment error in the RAO plane but not in the LAO plane. Given the high patient radiation doses associated with ECG-gated cardiac CT, the possibility of using nongated CT acquisitions without appreciably affecting integration accuracy is an important advantage for 3D-augmented fluoroscopy. The total patient effective radiation dose due to preprocedural nongated CT and 3D-augmented biplane fluoroscopy during the ablation procedure in our study was 18.8 ⫾ 8.4 mSv. This combined effective dose came close to the effective dose estimated for ECG-gated cardiac CT acquisition alone (17.3 ⫾ 5.2 mSv). The use of biplane 3D-augmented fluoroscopy as a single navigation tool for AF ablation therefore most likely does not increase the patient radiation dose when compared with ablation procedures of any kind that are preceded by ECG-gated CT acquisition. Rather than a true positive argument for using 3D-augmented fluoros-
963
copy, these results should encourage the use of nongated cardiac CT or other low-dose alternatives (such as cardiac MRI or CT with prospective ECG gating12) as preprocedural imaging tools before AF ablation. Although data on the accuracy of 3D image integration based on ECG-gated versus ECG-non-gated CT are currently not available for 3D electroanatomic mapping systems, some indirect evidence supports the use of nongated CT for accurate 3D image integration in this setting as well.13 The use of 3D-augmented fluoroscopy was felt to be very helpful and provided additional 3D information to guide catheter navigation and ablation during the procedures. The significant decrease in fluoroscopy time and trend toward shorter procedure times found in our study, however, need to be interpreted cautiously as they are based on nonrandomized data. Nevertheless, the same findings were noted in a study from Sra et al.,4 in which 50 consecutive patients were randomized to AF ablation procedures guided by electroanatomic mapping alone or in combination with monoplane CT fluoroscopy image integration. These investigators also reported a significant decrease in procedure duration and fluoroscopy times in the group with CT fluoroscopy guidance. A prospective randomized evaluation of the procedural characteristics and clinical outcome of biplane 3D-augmented fluoroscopy as a single navigation tool for AF ablation is currently ongoing in our center. An advantage of our implementation is that it has been developed independently in an academic environment and can be used with any type of fluoroscopy system and images from any type of CT or MRI unit. New techniques such as cardiac C-arm CT have the potential to further improve the workflow and integration accuracy of 3D-augmented fluoroscopy.14,15 By acquiring CT images in the electrophysiology room with the patient in the same position on the same table, more accurate registration approaches can further enhance 3D fluoroscopy image integration. Moreover, this approach eliminates possible differences in cardiac loading conditions between imaging and ablation. Compensation of catheter movement due to cardiac and respiratory motion during 3D-augmented fluoroscopy are other important fields currently under investigation.16 Similar to electroanatomic mapping systems, ECG-gated fluoroscopic image acquisition can be used to eliminate catheter movement due to cardiac motion, while at the same time reducing the procedural radiation dose. Correction of fluoroscopic catheter positions for respiratory movements is currently not possible during integration of 3D models with live fluoroscopy. The resulting drawback for 3D fluoroscopy integration is a mismatch between catheter position and 3D overlay during a part of the respiratory cycle. As the 3D LA models were registered to an angiographic image of the LA acquired during apnea (i.e., expiration), the operator had to take into account a periodic downward movement of the ablation catheter during inspiration. The 3D position of the ablation catheter tip relative to the integrated 3D model was therefore preferentially
964 assessed during expiration. Instead of such an intuitive approach, adapting the position and/or configuration of an integrated 3D LA PV model to patient respiration could improve the accuracy of fluoroscopic image integration and potentially allows more accurate lesion placement. In patients under general anesthesia, special high-frequency, low-volume ventilation modes could also be used to minimize the detrimental effects of respiratory motion on image integration.17
Study limitations The quantitative evaluation of 3D image integration accuracy is difficult owing to the lack of a gold standard, that is, knowledge of the true 3D position of LA and PVs in the patient during the procedure. However, measurement of alignment errors with angiographic images of the PVs reflects the integration accuracy in the area of interest and is in our opinion the most reliable method to assess integration accuracy relevant for ablation. Since small rotational errors are not reflected in alignment errors with the angiographic PV ostia, we opted to also score alignment based on the PVs in their more distal course. We did not test the integration accuracy of other cardiac chambers or vascular structures as this would require separate angiographic injections to verify integration accuracy. Ablation procedures in our study were associated with long fluoroscopy and procedure times. This can be explained by the use of relatively low-power ablation settings to increase procedural safety, the systematic ablation of ATP-induced dormant conduction, and the additional cavotricuspid isthmus ablation in 70% of patients. Low frame rate pulsed fluoroscopy was used to reduce the patient radiation dose associated with fluoroscopy. We consider the patient radiation doses with our approach (fluoroscopy 3 frames/second, nongated CT) certainly within acceptable limits.10 As mentioned, there are still many opportunities to further reduce it (e.g., preprocedural MRI, ECG-gated fluoroscopy). Freedom of AF was not evaluated as a clinical endpoint in our study. However, procedural endpoints were the same for all patients in the 2D and 3D⫹ group, and large differences in clinical outcome are therefore not expected.
Conclusion Biplane 3D-augmented fluoroscopy can be used as a safe and accurate stand-alone method to guide AF ablation procedures. Preprocedural imaging with non-ECG-gated cardiac CT results in an important reduction in patient radiation dose without substantially reducing 3D image integration accuracy.
Heart Rhythm, Vol 5, No 7, July 2008
Appendix Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.hrthm.2008.03.024.
References 1. Calkins H, Brugada J, Packer DL, et al. HRS/EHRA/ECAS expert consensus statement on catheter and surgical ablation of atrial fibrillation: recommendations for personnel, policy, procedures and follow-up. A report of the Heart Rhythm Society (HRS) Task Force on catheter and surgical ablation of atrial fibrillation. Heart Rhythm 2007;4:816 – 861. 2. Ector J, De Buck S, Adams J, et al. Cardiac three-dimensional magnetic resonance imaging and fluoroscopy merging: a new approach for electroanatomic mapping to assist catheter ablation. Circulation 2005;112:3769 –3776. 3. Sra J, Krum D, Malloy A, et al. Registration of three-dimensional left atrial computed tomographic images with projection images obtained using fluoroscopy. Circulation 2005;112:3763–3768. 4. Sra J, Narayan G, Krum D, et al. Computed tomography-fluoroscopy image integration-guided catheter ablation of atrial fibrillation. J Cardiovasc Electrophysiol 2007;18:409 – 414. 5. Maksimoviç R, Cademartiri F, Scholten M, et al. Sixteen-row multislice computed tomography in the assessment of pulmonary veins prior to ablative treatment: validation vs conventional pulmonary venography and study of reproducibility. Eur Radiol 2004;14:369 –374. 6. Noseworthy PA, Malchano ZJ, Ahmed J, et al. The impact of respiration on left atrial and pulmonary venous anatomy: implications for image-guided intervention. Heart Rhythm 2005;2:1173–1178. 7. Heidbuchel H, Ector H, Adams J, et al. Use of only a regular diagnostic His-bundle catheter for both fast and reproducible “para-Hisian pacing” and stable right ventricular pacing. J Cardiovasc Electrophysiol 1997;8:1121–1132. 8. De Buck S, Maes F, Ector J, et al. An augmented reality system for patientspecific guidance of cardiac catheter ablation procedures. IEEE Trans Med Imaging 2005;24:1512–1524. 9. De Buck S, Ector J, Maes F, et al. Interactive augmented reality for cardiac ablation therapy in the left and right atrium. Workshop proceedings of the 10th international conference on medical image computing and computer assisted intervention—MICCAI 2007, workshop on interaction in medical image analysis and visualization 2007;65–72. 10. Ector J, Dragusin O, Adriaenssens B, et al. Obesity is a major determinant of radiation dose in patients undergoing pulmonary vein isolation for atrial fibrillation. J Am Coll Cardiol 2007;50:234 –242. 11. European Commission. European guidelines on quality criteria for computed tomography, EUR 16262EN. Luxembourg: Office for Official Publications of the European Communities 2000. 12. Hsieh J, Londt J, Vass M, et al. Step-and-shoot data acquisition and reconstruction for cardiac x-ray computed tomography. Med Phys 2006;33:4236 – 4248. 13. Dong J, Dalal D, Scherr D, et al. Impact of heart rhythm status on registration accuracy of the left atrium for catheter ablation of atrial fibrillation. J Cardiovasc Electrophysiol 2007;18:1269 –1276. 14. Lauritsch G, Boese J, Wigstrom L, et al. Towards cardiac C-arm computed tomography. IEEE Trans Med Imaging 2006;25:922–934. 15. Thiagalingam A, Manzke R, D’Avila A, et al. Intraprocedural volume imaging of the left atrium and pulmonary veins with rotational X-ray angiography: implications for catheter ablation of atrial fibrillation. J Cardiovasc Electrophysiol 2008;19:293–300. 16. Ector J, De Buck S, Loeckx D, et al. Absolute and relative changes in left atrial and pulmonary venous anatomy due to respiration: relevance for 3D image integration during ablation procedures for atrial fibrillation. J Cardiovasc Electrophysiol 2008; March 26 [Epub ahead of print]. 17. Goode J, Taylor RL, Buffington CW, et al. High-frequency jet ventilation: utility in posterior left atrial catheter ablation. Heart Rhythm 2006;3: 13–19.