Feasibility of Auto-Quantified Epicardial Adipose Tissue in Predicting Atrial Fibrillation Recurrence After Catheter Ablation (2024)

Abstract

Background: The aim of this study was to build an auto-segmented artificial intelligence model of the atria and epicardial adipose tissue (EAT) on computed tomography (CT) images, and examine the prognostic significance of auto-quantified left atrium (LA) and EAT volumes for AF.

Methods and Results: This retrospective study included 334 patients with AF who were referred for catheter ablation (CA) between 2015 and 2017. Atria and EAT volumes were auto-quantified using a pre-trained 3-dimensional (3D) U-Net model from pre-ablation CT images. After adjusting for factors associated with AF, Cox regression analysis was used to examine predictors of AF recurrence. The mean (±SD) age of patients was 56±11 years; 251 (75%) were men, and 79 (24%) had non-paroxysmal AF. Over 2 years of follow-up, 139 (42%) patients experienced recurrence. Diabetes, non-paroxysmal AF, non-pulmonary vein triggers, mitral line ablation, and larger LA, right atrium, and EAT volume indices were linked to increased hazards of AF recurrence. After multivariate adjustment, non-paroxysmal AF (hazard ratio [HR] 0.6; 95% confidence interval [CI] 0.4–0.8; P=0.003) and larger LA-EAT volume index (HR 1.1; 95% CI 1.0–1.2; P=0.009) remained independent predictors of AF recurrence.

Conclusions: LA-EAT volume measured using the auto-quantified 3D U-Net model is feasible for predicting AF recurrence after CA, regardless of AF type.

Atrial fibrillation (AF) is the most common sustained arrhythmia, significantly increasing the risk of adverse events such as ischemic stroke, coronary artery disease, heart failure, and mortality.1

Catheter ablation (CA) is a well-established treatment for symptomatic anti-arrhythmic drug-refractory AF and can significantly improve the quality of life of patients with AF.2^,3 Despite ongoing medical and technological advances in ablation therapy, many patients continue to experience complications and relapse of AF. Atrial structural remodeling is one of the main causes of AF occurrence and deterioration after CA. Previous studies have demonstrated distinct left atrial (LA) remodeling with increased LA volume, which is associated with a higher risk of AF recurrence after CA.4 In addition, the other well-known marker, epicardial adipose tissue (EAT), which is an active metabolic endocrine tissue, secretes a variety of cytokines and affects cardiac structure and function.5^,6 van Rosendael et al7 demonstrated that LA EAT mass was larger in patients with paroxysmal AF than in those with sinus rhythm, whereas large LA volumes were associated with a high prevalence of non-paroxysmal AF. LA EAT is intimately adjacent to the LA, and it has been hypothesized to be closely associated with the occurrence and relapse of AF after CA through multiple pathogeneses.8 Meta-analyses have shown that a larger volume of total EAT, LA EAT, and EAT thickness are associated with AF recurrence in patients undergoing CA.9^,10 However, the methodology used to quantify EAT thickness or volume varied between studies, and laborious work was performed manually using multimodal cardiovascular imaging.11^–13 An efficient autosegmented model for the quantification of atrial and EAT volumes has not been developed. Moreover, although previous studies have demonstrated that LA enlargement and abundant EAT are associated with AF recurrence after CA, the collaborative effect has not yet been examined. Hence, the aim of this study was to develop an auto-segmentation model of both atria and EAT on computed tomography (CT) using machine learning and to examine the combined prognostic significance of LA size and EAT volume for AF after CA.

Methods

This study was approved by the Institutional Review Board of Taipei Veterans General Hospital, Taipei, Taiwan (VGH-IRB no. 2022-06-016 AC). The procedures were performed in accordance with the Declaration of Helsinki and the ethical standards of the institutional committee on human experimentation. Patient records and information were anonymized and deidentified before analysis.

Study Population

We conducted a single-center retrospective study of 334 patients with AF who were referred to Taipei Veterans General Hospital for the first time for CA between January 2015 and December 2017 with a preprocedural cardiac CT study. Written informed consent was obtained from all patients to include both CA and anonymized medical information in this study.

CT Acquisition

CT images were obtained using a 64-slice CT scanner (Aquilion 64 CFX; Toshiba Medical Systems, Tokyo, Japan) and a 256-slice multidetector CT scanners (Brilliance CT40; Philips Healthcare, Amsterdam, Netherlands). The Aquilion 64 CFX CT scanner used a peak tube voltage of 100 kVp (kilovoltage peak) and a tube current of 350 mA, whereas the Brilliance iCT scanner used a peak tube voltage of 100 or 120 kVp and a tube current of 596 mA. The Aquilion 64 and Brilliance iCT scanners generated 16- and 12-bit grayscale CT images, respectively. Cardiac CT images from each patient were reconstructed using a slice thickness of 1 mm and stored in the DICOM format with a matrix size of 512×512. The variation of pixel sizes for the reconstructed images ranged from 0.34×0.34 to 0.72×0.72 mm².

Image Preprocessing and Automatic Segmentation of the Atria, Atrial Appendages, and EAT Using the 3-Dimensional (3D) U-Net Model

Detailed information regarding the image preprocessing algorithm (Figure 1) and the automatic segmentation of the atria, atrial appendages, and EAT using our pre-trained 3D U-Net, including hyperparameters, the training dataset, and model performance, has been published elsewhere.14 Briefly, after segmentation of the LA, right atrium (RA; Figure 2B), and pericardium (Figure 2C) using 3D U-Net models, the LA-EAT and RA-EAT were segmented as follows. First, because the mean thickness of the left and right atrioventricular sulcus EAT has been reported to be 12.7 and 13.9 mm, respectively,15 the endocardial borders of the LA and RA were dilated by 15 mm to detect fat around the LA and RA (Figure 2D). The regions of detection for the LA-EAT and RA-EAT were located between the endocardium and dilated boundaries. Second, the pericardial regions were used as a constraint to exclude pericardial fat (Figure 2C,E). At the intersection of the dilated areas and pericardium, the EAT could be identified as pixels between −190 HU and −30 HU (Figure 2G). Third, we measured the minimum Euclidean distance between each EAT pixel and both atria to reassign all EAT pixels as LA-EAT or RA-EAT (Figure 2F). Pixels are attributed to LA-EAT when the minimum Euclidean distance of an EAT pixel to the LA is shorter than the distance to the RA, and vice versa. When the minimum distance of an EAT pixel from the LA is equivalent to the distance from the RA, this pixel is commonly shared by LA-EAT and RA-EAT (Supplementary Movie).

Electroanatomical Mapping and Catheter Ablation

Standard mapping and AF ablation techniques at the Taipei Veterans General Hospital have been described elsewhere.16^–19 Briefly, all anti-arrhythmic drugs except amiodarone were discontinued for ≥5 half-lives before the start of the procedure. The patient was presented to the cardiac electrophysiology laboratory in a fasting state. Conscious sedation was typically performed. General anesthesia was administered when necessary for ventilation, oxygenation, or patient comfort. Intravenous heparin was administered to achieve an activated clotting time >250 s for a conventional mapping catheter and >300 s for a high-density mapping catheter after transseptal access. Under fluoroscopic guidance, 2 long 8.5-Fr transseptal sheaths (SL-0; St. Jude Medical, St. Paul, MN, USA) were advanced over a long guidewire to the superior vena cava, and a flushed BRK transseptal needle (St. Jude Medical) was introduced into the SL-0 sheaths. The BRK needle was then used to puncture the atrial septum, and the entire system was carefully advanced into the LA under fluoroscopic guidance. An electroanatomical voltage map was created during sinus rhythm using a PentaRay catheter (20 electrodes with 2-6-2 mm spacing), a Lasso catheter (circular 10-electrode catheter) using the Carto3 system (Biosense Webster Inc., Diamond Bar, CA, USA), or by an AFocus catheter (circular, 10 electrodes) using the Ensite Precision system. Mapping was performed with an equal distribution of points, using a fill threshold of 15 mm.

After LA voltage mapping, radiofrequency ablation was performed using a 3.5-mm open-irrigated catheter (CoolFlex [St. Jude Medical] or ThermoCool [Biosense Webster Inc.]) with wide antral pulmonary vein (PV) isolation. After sinus rhythm was restored from AF by procedural AF termination or electrical cardioversion, mapping and ablation were applied only to spontaneously initiated focal atrial tachycardia, atrial flutters, and non-PV triggers that initiated AF.20 If any non-PV trigger initiating AF from the superior vena cava was identified, isolation of the superior vena cava was guided by catheter recordings from the superior vena cava-atrial junction. PV entrance block was confirmed, and an inducibility test was performed using isoproterenol infusion until the heart rate was >100 beats/min or increased by >10%.

Clinical Follow-up

Patients were routinely evaluated 2 weeks after ablation and at 1- to 3-month intervals in our cardiology outpatient clinic. For patients not followed at the Taipei Veterans General Hospital, referring cardiologists were contacted, and recurrence was defined as ≥30 s of sustained atrial arrhythmias. Anti-arrhythmic drugs were prescribed continuously for 4–8 weeks after ablation to prevent the early recurrence of AF. The blanking period was defined as within 3 months of the procedure. A 24-hour Holter study or 7-day cardiac event monitoring was performed regularly at 3 months, within a year after the ablation procedure, and when patients experienced symptoms suggestive of tachyarrhythmias. Long-term efficacy was assessed using resting surface 12-lead electrocardiograms, a 24-hour Holter study, and 7-day cardiac event monitoring records.

Statistical Analysis

Continuous variables are expressed as the mean±SD and categorical variables are expressed as counts with percentages. Independent samples t-test, analysis of variance (ANOVA), and Fisher’s exact test were used to compare normally distributed continuous and categorical variables. Univariate Cox proportional hazard regression was used to identify baseline clinical variables predictive of atrial arrhythmias during follow-up. Relative risk estimates for atrial arrhythmias from the univariate Cox regression analyses were reported as hazard ratios (HRs) and 95% confidence intervals (CIs). Atrial arrhythmias at 2 years were studied using Cox time-dependent regression analysis. Atrial arrhythmias at follow-up were reported with a time-to-event analysis, survival curves were created using the Kaplan–Meier method, and differences between groups were compared using the log-rank test. The optimal cut-off value of LA-EAT for predicting atrial arrhythmia recurrence after CA was identified using receiver operating characteristic (ROC) curves. All statistical analyses were performed using SPSS version 24 (SPSS Inc., Chicago, IL, USA). Statistical significance was set at two-tailed P<0.05.

Results

Study Population

Between 2015 and 2017, 457 patients with AF underwent CA at the Taipei Veterans General Hospital. We excluded 109 patients who underwent repeated ablation procedures or underwent CT studies outside the hospital. Of the remaining 348 patients, a further 14 were excluded because of unsatisfactory CT image quality. Thus, 334 patients with AF who underwent first-time ablation and for whom adequate preprocedural cardiac CT images were acquired were enrolled in the study. A flowchart of study enrollment is shown in Figure 3.

Clinical Characteristics, Imaging Features, and Procedural Data

Baseline characteristics, cardiovascular imaging features, and procedural data are summarized in Tables 1 and 2. In all, 334 AF patients (age 56.4±10.7 years, 251 [75%] men, 79 [24%] non-paroxysmal AF) were included in the study. The mean CHA₂DS₂-VASc score was 1.4±1.2, and the mean left ventricular (LV) ejection fraction was 58.3±6.5%. Based on CT imaging features, mean volume indices were determined as follows: LA, 67.8±20.3 mL/m²; LA appendage, 6.2±2.9 mL/m²; LA-EAT, 8.3±3.6 mL/m²; RA, 65.4±19.6 mL/m²; RA appendage (RAA), 4.6±2.0 mL/m²; RA-EAT, 11.5±4.2 mL/m²; and total EAT, 47.0±18.7 mL/m². The mean total EAT density showed −76.0±7.0 HU. Baseline medications included Class Ic antiarrhythmic drugs (propafenone or flecainide) in 142 (43%) patients, amiodarone in 167 (50%) patients, dronedarone in 54 (16%) patients, β-blockers in 111 (33%) patients, and digoxin in 6 (2%) patients; anticoagulation included warfarin in 26 (8%) patients or non-vitamin K antagonist oral anticoagulants in 164 (49%) patients. Most patients who were not indicated for oral anticoagulants according to European and Asia Pacific guidelines21^,22 received aspirin (114 [34%] patients). During the ablation procedure, 321 (96%) patients received 4 PV isolations; 61 patients (18%) had non-PV triggers, and the triggers were eliminated. Further ablation of the mitral line (18 [5%] patients), septal line (1 [0.3%] patients), or roof line (22 [7%] patients) was performed to induce atypical atrial flutter.

Impact of Clinical and Imaging Features and Procedural Variables on Recurrence of Atrial Arrhythmia

After a median follow-up of 12 months (interquartile range [IQR] 4–38 months), 186 (56%) patients experienced atrial arrhythmia recurrence; 139 (42%) patients experienced atrial arrhythmia recurrence within 2 years. The univariate predictors of atrial arrhythmia recurrence are presented in Table 3. Patients with diabetes (HR 1.57; 95% CI 1.01–2.43; P=0.046), non-paroxysmal AF (HR 1.89; 95% CI 1.33–2.68; P<0.001), and larger volume indices of the LA (HR 1.01; 95% CI 1.00–1.02; P=0.006), LA-EAT (HR 1.07; 95% CI 1.02–1.12; P=0.003), RA (HR 1.01; 95% CI 1.01–1.02; P=0.001), RAA (HR 1.14; 95% CI 1.05–1.23; P=0.001), RA-EAT (HR 1.05; 95% CI 1.01–1.10; P=0.012), and total EAT (HR 1.01; 95% CI 1.00–1.02; P=0.015) had an increased risk of atrial arrhythmia recurrence.

With regard to procedural variables, the presence of non-PV triggers (HR 1.59; 95% CI 1.09–2.33; P=0.02) and the creation of a mitral line (which implied inducible mitral flutter during the procedure; HR 1.95; 95% CI 1.05–3.61; P=0.03) were predictors of recurrence.

After multivariable adjustment, non-paroxysmal AF (adjusted HR 0.58; 95% CI 0.41–0.83; P=0.003) and larger LA-EAT volume index (HR 1.13; 95% CI 1.03–1.24; P=0.009) remained independent predictors of atrial arrhythmia recurrence. The best cut-off value of the LA-EAT volume index for predicting atrial arrhythmia recurrence after CA was ≥8 mL/m². The CT imaging features of the RAA and total EAT volume index were not selected in the multivariate analysis to avoid the co-effects of total RA volume index and atrial EAT volume index in the multivariate model. Hence, AF patients with an LA-EAT volume index ≥8 mL/m²are defined as having abundant LA-EAT, and had a lower recurrence-free survival than patients with smaller amounts of LA-EAT (log-rank P=0.005) in the Kaplan-Meier survival analysis (Figure 4A). Moreover, in the subgroup analysis categorized by AF type, patients with abundant LA-EAT still had worse survival outcomes (log-rank P=0.008; Figure 4B).

Discussion

To the best of our knowledge, this is the first study to use the 3D U-Net model to simultaneously quantify atrial and EAT volumes on CT images and assess their combined impact on atrial arrhythmia recurrence after AF ablation. The established 3D U-Net model required approximately 7 s for auto-segmentation of the atria, atrial appendages, and pericardium. Finally, the volume of the atria, LA-EAT, RA-EAT, and total EAT could be obtained within 30 s, which greatly reduced the remaining burden of laborious work. Furthermore, we discovered that an abundant LA-EAT volume was independently associated with an increased risk of atrial arrhythmia recurrence, regardless of AF type. We provide the Multimodal Radiomics Platform (http://cflu.lab.nycu.edu.tw/MRP_MLinglioma.html) and a comprehensive flowchart for auto-segmentation of both atria and the pericardium on CT images. Moreover, all possible risk factors of recurrence after CA in AF patients were examined using a multivariate regression model, which is unique compared with prior studies.8^,23 In the present study, we trained the machine learning model for auto-segmentation of the LA, RA, and pericardium by manually contouring CT images from 30 patients, spending 1.5 h outlining each atrium and 2 h for the pericardium. The establishment of the 2-dimensional (2D) U-Net auto-segmentation model required a total of 150 h of work for contouring and training. Under 2D U-Net assistance, it took 30 min for auto-segmentation of the atria, and 1 h for the pericardium, which included efforts to manually correct the contours segmented by the 2D U-Net model. These contouring procedures were consistent with previous studies’ methods to quantify EAT and LA-EAT.8^,23^–25 Of note, we further developed a 3D U-Net model using CT images from 125 patients, and the Dice coefficients between manually contouring and auto-segmentation of the LA, RA, pericardium, EAT, LA-EAT, and RA-EAT were 0.960±0.010, 0.945±0.013, 0.967±0.006, 0.870±0.027, 0.846±0.057, and 0.841±0.071, respectively. Our study used a reliable machine learning model that could quantify EAT volume consistently and reproducibly compared with previous studies’ methodology using semiquantification for EAT volume measurement.8^,23^–25

A previous meta-analysis identified EAT as a predictor of AF recurrence after CA;10 however, the results of individual studies are conflicting.26^–28 Studies have varied in their methods of evaluating EAT and LA-EAT, including quantification of total EAT and LA-EAT volumes, with or without indexing by body surface area on CT images, or measuring only the wall thickness of the EAT and posterior wall thickness of the LA-EAT, either by CT or transthoracic echocardiography. Furthermore, the measurements were performed manually, and thus may have been inconsistent or laborious. The discrepancies in quantification approaches and different physicians evaluating the EAT may be critical reasons for the conflicting results regarding the impact of the EAT on the prognosis of AF ablation. In our study, we used a pre-trained 3D U-Net model for auto-segmentation, which could precisely, consistently, and efficiently quantify the volume of the atria, total EAT, and atrial EAT. This pre-trained machine learning model assessed the amount of EAT, especially LA-EAT, and associated its adverse impact with the recurrence of AF ablation.

Regarding the combined impact of LA size and the amount of LA-EAT, the LA-EAT volume index was positively correlated with the LA volume index, which has been confirmed in previous studies quantified using CT or cardiovascular magnetic resonance (CMR).28^,29 Our study observed a similar positive correlation (R=0.37, P<0.001) between LA-EAT and LA volume index. Figure 5 illustrates the LA-EAT volume index stratified by LA volume index tertiles in our cohort. Interestingly, abundant LA-EAT remained an independent predictor of atrial arrhythmia recurrence after adjusting for LA size. Our findings suggest that LA-EAT may be a better predictor than LA volume of arrhythmia recurrence after CA. One interesting issue is the relationship between EAT and obesity. Because obesity is a well-known indicator of AF perpetuation and a worse outcome of AF ablation,30^,31 it raises the assumption that abundant EAT is only a reflection of increased body size. In addition, whether EAT provides more information regarding prognostic prediction for AF ablation in patients with an average body size has not been clarified in the literature. A previous study reported that LA-EAT was associated with body mass index (BMI; R=0.56, P<0.001) measured on CMR images.13 Sanghai et al28 demonstrated that the LA-EAT volume index is associated with higher odds of AF recurrence, but only in obese patients with a BMI ≥30 kg/m². In our study, we observed a similar positive correlation between BMI and the LA-EAT volume index (R=0.44, P<0.001); however, abundant LA-EAT was associated with an increased risk of AF recurrence even in patients without morbid obesity. Because the patients in our study were all Asian, the average BMI was significantly lower than that of Western patients. In addition, even though the LA-EAT did not colocalize with the fibrotic tissue explored in a previous CMR study,13 local infiltration effects other than fibrosis can still affect LA remodeling and worsen LA function. Previous studies have explored the infiltration of LA-EAT into LA myocardium.32^,33 The effects of proinflammatory cytokines and chemokines secreted by infiltrating EAT stimulate atrial remodeling. In summary, abundant LA-EAT cannot be fully explained as coexisting with obesity, and can predict AF ablation outcomes in non-obese patients.

Study Limitations

This study has several limitations. First, it is a retrospective study that is based on a single-center experience, and it is unknown whether the associations explored in this study can be extrapolated to other populations. Second, patients with previously implanted cardiac implantable electronic devices were excluded from the study, which may have resulted in selection bias. Third, the reproducibility of the artificial intelligence (AI) auto-segmentation 3D U-Net model for analyzing CT images from other institutions has not been validated. However, because the AI model was trained using images acquired by the 2 most popular brands of CT scanners, it holds promise for successful auto-segmentation of CT images acquired from other hospitals.

Conclusions

The measurement of LA-EAT volume using an auto-quantified 3D U-Net model on CT images is feasible for patients with AF and can predict AF recurrence regardless of AF type. The integration of AI technology to assess LA-EAT on CT images enables efficient segmentation and serves as a valuable phenotype for predicting atrial arrhythmia recurrence after AF ablation.

Sources of Funding

This research was supported by grants from Taipei Veterans General Hospital (V112B-002), Szu-Yuan Research Foundation (113007), the National Science and Technology Council (NSTC 112-2314-B-A49-060), and the Veterans General Hospitals and University System of Taiwan Joint Research Program (VGHUST113-G1-2-3).

Disclosures

None.

IRB Information

This study was approved by the Taipei Veterans General Hospital, Taipei, Taiwan (VGH-IRB no. 2022-06-016 AC).

Data Availability

The data underlying this article will be shared upon reasonable request to the corresponding authors.

Supplementary Files

Supplementary Movie. Display of Pre-trained 3D U-Net model for Autosegmentation of Atria and EAT on computed tomography.

Please find supplementary file(s);

https://doi.org/10.1253/circj.CJ-23-0808

References

• 1.Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an independent risk factor for stroke: The Framingham Study. Stroke 1991; 22: 983–988, doi:10.1161/01.str.22.8.983.
• 2.Mao YJ, Wang H, Chen JX, Huang PF. Meta-analysis of medical management versus catheter ablation for atrial fibrillation. Rev Cardiovasc Med 2020; 21: 419–432, doi:10.31083/j.rcm.2020.03.60.
• 3.Kim YG, Shim J, Choi JI, Kim YH. Radiofrequency catheter ablation improves the quality of life measured with a Short Form-36 questionnaire in atrial fibrillation patients: A systematic review and meta-analysis. PLoS One 2016; 11: e0163755, doi:10.1371/journal.pone.0163755.
• 4.Njoku A, Kannabhiran M, Arora R, Reddy P, Gopinathannair R, Lakkireddy D, et al. Left atrial volume predicts atrial fibrillation recurrence after radiofrequency ablation: A meta-analysis. Europace 2018; 20: 33–42, doi:10.1093/europace/eux013.
• 5.Hatem SN, Sanders P. Epicardial adipose tissue and atrial fibrillation. Cardiovasc Res 2014; 102: 205–213, doi:10.1093/cvr/cvu045.
• 6.Poggi AL, Gaborit B, Schindler TH, Liberale L, Montecucco F, Carbone F. Epicardial fat and atrial fibrillation: The perils of atrial failure. Europace 2022; 24: 1201–1212, doi:10.1093/europace/euac015.
• 7.van Rosendael AR, Smit JM, El’Mahdiui M, van Rosendael PJ, Leung M, Delgado V, et al. Association between left atrial epicardial fat, left atrial volume, and the severity of atrial fibrillation. Europace 2022; 24: 1223–1228, doi:10.1093/europace/euac031.
• 8.Tsao HM, Hu WC, Wu MH, Tai CT, Lin YJ, Chang SL, et al. Quantitative analysis of quantity and distribution of epicardial adipose tissue surrounding the left atrium in patients with atrial fibrillation and effect of recurrence after ablation. Am J Cardiol 2011; 107: 1498–1503, doi:10.1016/j.amjcard.2011.01.027.
• 9.Sepehri Shamloo A, Dagres N, Dinov B, Sommer P, Husser-Bollmann D, Bollmann A, et al. Is epicardial fat tissue associated with atrial fibrillation recurrence after ablation? A systematic review and meta-analysis. Int J Cardiol Heart Vasc 2019; 22: 132–138, doi:10.1016/j.ijcha.2019.01.003.
• 10.Chen J, Mei Z, Yang Y, Dai C, Wang Y, Zeng R, et al. Epicardial adipose tissue is associated with higher recurrence risk after catheter ablation in atrial fibrillation patients: A systematic review and meta-analysis. BMC Cardiovasc Disord 2022; 22: 264, doi:10.1186/s12872-022-02703-9.
• 11.Chao TF, Hung CL, Tsao HM, Lin YJ, Yun CH, Lai YH, et al. Epicardial adipose tissue thickness and ablation outcome of atrial fibrillation. PLoS One 2013; 8: e74926, doi:10.1371/journal.pone.0074926.
• 12.Canpolat U, Aytemir K, Yorgun H, Asil S, Dural M, Ozer N. The impact of echocardiographic epicardial fat thickness on outcomes of cryoballoon-based atrial fibrillation ablation. Echocardiography 2016; 33: 821–829, doi:10.1111/echo.13193.
• 13.Chahine Y, Askari-Atapour B, Kwan KT, Anderson CA, Macheret F, Afroze T, et al. Epicardial adipose tissue is associated with left atrial volume and fibrosis in patients with atrial fibrillation. Front Cardiovasc Med 2022; 9: 1045730, doi:10.3389/fcvm.2022.1045730.
• 14.Wang GJ, Kuo L, Chang SL, Lin YJ, Chung FP, Lo LW, et al. Deep learning-based workflow for automatic extraction of atria and epicardial adipose tissue on cardiac computed tomography in atrial fibrillation. medRxiv 2023, doi:10.1101/2023.05.03.23289448.
• 15.Wang TD, Lee WJ, Shih FY, Huang CH, Chang YC, Chen WJ, et al. Relations of epicardial adipose tissue measured by multidetector computed tomography to components of the metabolic syndrome are region-specific and independent of anthropometric indexes and intraabdominal visceral fat. J Clin Endocrinol Metab 2009; 94: 662–669, doi:10.1210/jc.2008-0834.
• 16.Lo LW, Lin YJ, Chang SL, Hu YF, Chao TF, Chung FP, et al. Predictors and characteristics of multiple (more than 2) catheter ablation procedures for atrial fibrillation. J Cardiovasc Electrophysiol 2015; 26: 1048–1056, doi:10.1111/jce.12748.
• 17.Verma A, Jiang CY, Betts TR, Chen J, Deisenhofer I, Mantovan R, et al. Approaches to catheter ablation for persistent atrial fibrillation. N Engl J Med 2015; 372: 1812–1822, doi:10.1056/NEJMoa1408288.
• 18.Cheng WH, Lo LW, Lin YJ, Chang SL, Hu YF, Hung Y, et al. Ten-year ablation outcomes of patients with paroxysmal atrial fibrillation undergoing pulmonary vein isolation. Heart Rhythm 2019; 16: 1327–1333, doi:10.1016/j.hrthm.2019.03.028.
• 19.Liu CM, Chang SL, Chen HH, Chen WS, Lin YJ, Lo LW, et al. The clinical application of the deep learning technique for predicting trigger origins in patients with paroxysmal atrial fibrillation with catheter ablation. Circ Arrhythm Electrophysiol 2020; 13: e008518, doi:10.1161/CIRCEP.120.008518.
• 20.Lin WS, Tai CT, Hsieh MH, Tsai CF, Lin YK, Tsao HM, et al. Catheter ablation of paroxysmal atrial fibrillation initiated by non-pulmonary vein ectopy. Circulation 2003; 107: 3176–3183, doi:10.1161/01.CIR.0000074206.52056.2D.
• 21.Chao TF, Joung B, Takahashi Y, Lim TW, Choi EK, Chan YH, et al. 2021 Focused update of the 2017 consensus guidelines of the Asia Pacific Heart Rhythm Society (APHRS) on stroke prevention in atrial fibrillation. J Arrhythm 2021; 37: 1389–1426, doi:10.1002/joa3.12652.
• 22.Hindricks G, Potpara T, Dagres N, Arbelo E, Bax JJ, Blomstrom-Lundqvist C, et al. 2020 ESC guidelines for the diagnosis and management of atrial fibrillation developed in collaboration with the European Association for Cardio-Thoracic Surgery (EACTS): The Task Force for the diagnosis and management of atrial fibrillation of the European Society of Cardiology (ESC) developed with the special contribution of the European Heart Rhythm Association (EHRA) of the ESC. Eur Heart J 2021; 42: 373–498, doi:10.1093/eurheartj/ehaa612.
• 23.Huber AT, Fankhauser S, Chollet L, Wittmer S, Lam A, Baldinger S, et al. The relationship between enhancing left atrial adipose tissue at CT and recurrent atrial fibrillation. Radiology 2022; 305: 56–65, doi:10.1148/radiol.212644.
• 24.Wang X, Butcher SC, Kuneman JH, Lustosa RP, Fortuni F, Ajmone Marsan N, et al. The quantity of epicardial adipose tissue in patients having ablation for atrial fibrillation with and without heart failure. Am J Cardiol 2022; 172: 54–61, doi:10.1016/j.amjcard.2022.02.021.
• 25.Vroomen M, Olsthoorn JR, Maesen B, L’Espoir V, La Meir M, Das M, et al. Quantification of epicardial adipose tissue in patients undergoing hybrid ablation for atrial fibrillation. Eur J Cardiothorac Surg 2019; 56: 79–86, doi:10.1093/ejcts/ezy472.
• 26.Cruz I, Lopes Fernandes S, Diaz SO, Saraiva F, Barros AS, Primo J, et al. Epicardial adipose tissue volume is not an independent predictor of atrial fibrillation recurrence after catheter ablation. Rev Esp Cardiol (Engl Ed) 2023; 76: 539–547, doi:10.1016/j.rec.2022.11.006.
• 27.Masuda M, Mizuno H, Enchi Y, Minamiguchi H, Konishi S, Ohtani T, et al. Abundant epicardial adipose tissue surrounding the left atrium predicts early rather than late recurrence of atrial fibrillation after catheter ablation. J Interv Card Electrophysiol 2015; 44: 31–37, doi:10.1007/s10840-015-0031-3.
• 28.Sanghai SR, Sardana M, Hansra B, Lessard DM, Dahlberg ST, Aurigemma GP, et al. Indexed left atrial adipose tissue area is associated with severity of atrial fibrillation and atrial fibrillation recurrence among patients undergoing catheter ablation. Front Cardiovasc Med 2018; 5: 76, doi:10.3389/fcvm.2018.00076.
• 29.Kinjo N, Ikeda Y, Taguchi K, Sugimoto R, Maehara S, Tsujita E, et al. Hepatic resection of hepatocellular carcinoma after proton beam therapy: A case report. Hepatol Res 2016; 46: 483–488, doi:10.1111/hepr.12576.
• 30.Winkle RA, Mead RH, Engel G, Kong MH, Fleming W, Salcedo J, et al. Impact of obesity on atrial fibrillation ablation: Patient characteristics, long-term outcomes, and complications. Heart Rhythm 2017; 14: 819–827, doi:10.1016/j.hrthm.2017.02.023.
• 31.Donnellan E, Wazni O, Kanj M, Hussein A, Baranowski B, Lindsay B, et al. Outcomes of atrial fibrillation ablation in morbidly obese patients following bariatric surgery compared with a nonobese cohort. Circ Arrhythm Electrophysiol 2019; 12: e007598, doi:10.1161/CIRCEP.119.007598.
• 32.Takahashi N, Abe I, Kira S, Ishii Y. Role of epicardial adipose tissue in human atrial fibrillation. J Arrhythm 2023; 39: 93–110, doi:10.1002/joa3.12825.
• 33.Abe I, Teshima Y, Kondo H, Kaku H, Kira S, Ikebe Y, et al. Association of fibrotic remodeling and cytokines/chemokines content in epicardial adipose tissue with atrial myocardial fibrosis in patients with atrial fibrillation. Heart Rhythm 2018; 15: 1717–1727, doi:10.1016/j.hrthm.2018.06.025.