Detection Of Atrial Fibrillation Research Articles

Mitigating Data Variability and Overfitting in Deep Learning Models for Atrial Fibrillation Detection Using Single-Lead ECGs

Mitigating Data Variability and Overfitting in Deep Learning Models for Atrial Fibrillation Detection Using Single-Lead ECGs

Predicting atrial fibrillation risk following embolic stroke of unknown source using cardiac magnetic resonance imaging

Abstract Background Approximately 25% of embolic strokes of unknown source (ESUS) are attributed to undiagnosed atrial fibrillation (AF).1 The European Stroke Organisation (ESO) recommends implantable loop recorder (ILR) monitoring to identify ESUS patients with AF, permit the initiation of anticoagulation and reduce risk of recurrent stroke.2 However, this approach delays treatment and carries significant cost and resource implications. An alternative approach, where early anticoagulation is initiated in the absence of AF has failed to reduce recurrent stroke in four randomised controlled trials (NAVIGATE ESUS, RE-SPECT ESUS, ATTICUS and ARCADIA).3-6 An urgent need exists for new strategies to identify ESUS patients with undiagnosed AF that are at high risk of future thromboembolism. Purpose The primary objective of this study was to produce a substrate-based predictive model for AF in ESUS patients using atrial cardiac magnetic resonance imaging (CMRi). Method The Atrial CARdiac Magnetic resonance imaging in patients with embolic stroke of unknown source without documented AF (CARM-AF) Study is a prospective, multi-centre, observational study. All patients underwent CMRi and ILR insertion for AF detection within 3 months of ESUS. The main inclusion criteria were expected survival >12 months, CHA2DS2VASc≥3 and eGFR >30ml/min. Patients were allocated to the study group if AF >30 seconds was detected by ILR during the first year of follow-up and control group if no AF detected (Figure 1). Univariable analysis was used to identify clinical, echocardiographic and CMR parameters associated with AF diagnosis. Parameters with p<0.2 and no co-linearity were used to develop multivariable logistic regression models for AF prediction. Results From September 2020 to September 2022, 102 patients were enrolled and 91 were included in the final analysis. AF was detected in 17 patients during the first year of follow-up (18.6%). In univariable analysis, patients with AF were more likely to be female (p=0.02), but there were no significant differences in age, race, body mass index (BMI) or CHA2DS2VASc score (p=0.06, 0.272, 0.195, 0.211) between groups. Increased left atrial (LA) volume, surface area and reduced LA ejection fraction were associated with increased risk of AF detection at 1 year (p=0.006, 0.044, 0.008). LA fibrosis and sphericity were not associated with AF (p=0.84, 0.98). Logistic regression models for AF risk prediction were developed using patient, echocardiographic and CMR parameters. A combined model of patient and CMR parameters outperformed assessment of patient characteristics alone, achieving an AUC of 0.85 for predicting AF occurrence (p<0.005). Pseudo R2 of 0.30 (McFadden) indicated excellent model fit (Figure 2). Conclusion CARM-AF is the first study to demonstrate feasibility and utility of CMR imaging to predict AF risk following ESUS. A randomised controlled trial initiating anticoagulation using the developed model is recommended.Study DesignModel Comparison (ROC Curves)

Effectiveness of a stepwise approach for screening of atrial fibrillation after stroke: insights from the SAFAS study

Abstract Background Detection of atrial fibrillation (AF) is critical after ischemic stroke, providing information regarding the mechanism of the event and leading to modification in the antithrombotic strategy. While most guidelines recommend screening patients for AF with 12-lead ECG, telemetry, long-duration Holter monitoring and implantable cardiac monitor (ICM), the optimal timing and combination of such screening tools remain unclear. Objective This study aimed at investigating the suitability of a sequential combination of screening techniques (12-lead ECG, telemetry, in hospital long-lasting Holter monitoring, and ICM in the detection of AF after stroke. Methods Patients without previously known AF admitted to the Dijon University Hospital stroke unit for acute ischemic stroke were prospectively enrolled. After a stepwise screening approach for AF based on admission ECG, telemetry monitoring during the stroke unit stay and long-duration Holter monitoring during hospital stay, cryptogenic stroke patients were implanted of an ICM. Primary endpoint was the presence of AF detected during the 3-year period after stroke based on this sequential screening approach. Results A total of 240 patients were included. Among them, 104 (43.3%) patients had a documented cause of stroke non-related to AF. Among the remaining 136 patients (53.7% male, 70.8±13.7 yo), AF was detected in 82 (60%) patients over the acute screening phase or the 3-year follow-up with ICM. AF was diagnosed using 12-lead ECG, in-hospital telemetry, and in hospital long-lasting Holter monitoring in 17 (13%), 25 (18%), and 18 (13%) patients, respectively. AF was detected after the first 24 hours on the long-lasting Holter monitoring in 66% of patients. Among the 76 (56%) patients classified as cryptogenic after the complete stroke work-up and implanted from an ICM, AF was detected in 22 (29%) patients. AF occurred during the first, second, and third years of implantable monitoring in 14 (18.4%), 5 (6.6%), and 3 (3.9%) patients, respectively (Figure 1). Mean time from ICM implantation to AF diagnosis was 308+/-279 days. Finally, among all AF detected, 72% (60/83) were found during the initial intensive in-hospital screening. Conclusion A stepwise approach for AF screening after ischemic stroke allows the early detection of AF in a substantial number of patients during hospital stay. Even with such proactive initial monitoring strategy, invasive monitoring remains complementary to non-invasive screening tools not to overlook more distant AF episodes. Studies focusing on the relative risk of ischemic stroke recurrence according to AF timing and burden are needed.Figure 1

Cost-effectiveness of atrial fibrillation detection with insertable cardiac monitors in large-artery or small-vessel disease stroke in the United Kingdom

Abstract Background The randomised STROKE-AF trial demonstrated clinical effectiveness of continuous cardiac monitoring to detect atrial fibrillation (AF) with insertable cardiac monitors (ICMs) for patients with ischemic stroke attributed to large-artery or small-vessel disease (LAD/SVD). ICMs were ten-fold more effective at detecting AF compared to the standard of care (SoC). However, cost-effectiveness of this monitoring strategy is unknown. Purpose The aim of this economic evaluation was to assess the cost-effectiveness of long-term continuous cardiac monitoring via ICMs, for detection of AF (and, in event of AF diagnosis, subsequent anticoagulation), in patients with ischemic stroke attributed to large-or-small vessel disease compared to the SoC of conventional follow up in the United Kingdom. Methods We constructed a Markov model with a cycle length of three months and a lifetime horizon to conduct the cost-effectiveness analysis, from the National Health Service (NHS) perspective. Costs and health-related benefits were discounted at an annual rate of 3.5%. We expressed outcomes in terms of quality-adjusted life-years (QALYs). Patient characteristics in the analysis and detection probabilities were based on the STROKE-AF trial. Detection of AF resulted in a switch in treatment from antiplatelet to direct oral anticoagulant (DOAC). Risk of cerebrovascular events were dependent on a patient’s AF detection status and subsequent treatment. Scenario, deterministic, and probabilistic sensitivity analyses were conducted to test the robustness of the base-case incremental cost-effectiveness ratio (ICER) and account for uncertainties in the model. We stratified patients into risk groups based on CHA2DS2-VASc scores and risk of AF detection. Results The base case incremental cost-effectiveness ratio of the implantation of an ICM was £15,485 per QALY, substantially under the NICE willingness-to-pay threshold of £20,000-£30,000 per QALY. ICM implantations resulted in 0.139 QALYs gained per patient at an incremental cost of £2,147. There were 0.054 fewer stroke events per patient in the ICM arm versus SoC. Results were robust under sensitivity analysis. Conclusion ICMs were cost-effective compared to SoC for detection of AF in patients with LAD/SVD stroke in the United Kingdom driven largely by more stroke events avoided. Subgroup analysis showed that ICMs were more cost-effective in patients with higher CHA2DS2-VASc scores and those with a higher individual risk of AF.Markov Model SchematicSummary ICER Results and Events avoided

A novel method to identify critical isthmus of right atrial reentrant tachycardia; Functional substrate mapping characteristics during sinus rhythm

Abstract Background Right atrial tachycardia (AT) is a frequent rhythm disorder in patients with atrial scar mainly due to surgical incisions or congenital heart diseases. Despite the mounting evidence about AT mechanisms and types, data is scarce regarding the conduction properties as well as functional characteristics of atrial substrate during sinus rhythm that has role in the maintenance of the tachycardia. Purpose We sought to evaluate the relationship between the functional substrate mapping (FSM) characteristics of right atrium and the critical isthmus (CI) of re-entrant ATs in patients with underlying atrial scar. Methods Among patients with the history of right AT who underwent catheter ablation with 3-D mapping were retrospectively enrolled. Voltage map and isochronal late activation mapping (ILAM) were created during sinus / paced rhythm using multielectrode catheters to detect deceleration zones (DZ). Subsequently AT was induced with programmed stimulation and activation mapping was performed to detect CI of the tachycardia. Atrial tachyarrhythmia (ATa) recurrence was defined as detection of atrial fibrillation or AT (≥ 30 s) during the follow-up. Results A total of 24 patients [mean age: 46 ±15, gender: 13 (54 %) female] with right AT were included. A total of 37 ATs were mapped [16 (43.2%); localized reentry, 21 (56.7%), macroreentry]. Atrial low voltage zones comprised 23.3 ± 13.0 % of total right atrium. The mean value of bipolar voltage, EGM duration and conduction velocity during sinus rhythm corresponding to CI of ATs were 0.18 ± 0.10 mV, 121.7 ± 29.4 ms, 0.06 ± 0.04 m/s and, respectively. Total number of DZs per chamber was 1.1 ± 0.3, which were all located in the low-voltage zone (<0.5 mV) detected by high density mapping. All CIs of non- cavo-tricuspid-isthmus (CTI) dependent reentry were colocalized with DZs detected during FSM. The positive predictive value of DZs to detect CI of inducible ATs is 80.8%. During the mean follow-up of 11.7 ± 8.1 months, freedom from ATa was 87.5%. Conclusion Although, CTI-dependent macroreentry is the most common mechanism in patients with RA scar, our findings demonstrated the relevance of FSM to predict non-CTI dependent ATs. Conduction slowing manifested as DZs with continuous-fragmented signal morphology may guide to tailor ablation strategy in case of underlying RA scar.

Feasibility and diagnostic yield of a photoplethysmography-based arrhythmia screening and integrated management pathway in a chronic obstructive pulmonary disease outpatient clinic

Abstract Background Guidelines recommend opportunistic screening for atrial fibrillation (AF) in high-risk populations, including chronic obstructive pulmonary disease (COPD) patients. Purpose We aimed to assess the feasibility and yield of a photoplethysmography (PPG) smartphone app-based pathway for AF detection within a COPD outpatient clinic. Methods Ninety-nine eligible patients from the COPD outpatient clinic at the Maastricht University Medical Center+ underwent a 14-days PPG app-based AF screening. PPG-based AF detection triggered an electrocardiogram (ECG) for AF diagnosis and initiation of a comprehensive AF management. Patients’ adherence and system usability were assessed. Results Patient onboarding was easy, and patients demonstrated good adherence to the protocol (101% [82%-123%]). The usability of the PPG app was satisfactory with a high score in the System Usability Score questionnaire (80.0 [55.0-92.5]). Among 99 patients (mean age 66±9 years; 54% female), 8% had a prior AF diagnosis. Premature beats were common, with 96% of patients showing isolated premature beats (<6 in 960 PPG recordings) and 32% frequent premature beats (≥6 in 210 PPG recordings). The PPG algorithm suspected paroxysmal AF in 6 new cases. AF could be confirmed in 2 patients through resting 12-lead ECG and 1-lead handheld ECG device. One patient who was lost to follow-up. In the remaining 3 patients with suspected AF PPG recordings, analysis of 24-hour, 48-hour and 7-day Holter recordings revealed high burden premature (supra)ventricular beats. Two of these patients manifested self-terminating short episodes of supraventricular tachycardia. Conclusion 14-days opportunistic PPG app-based screening is feasible in COPD outpatient care and results in the detection of new AF cases and premature beats. This novel digital approach offers a practical method to implement recommended AF screening in COPD patients.

Effects of long-term continuous screening for atrial fibrillation on heart failure events - results from the randomized LOOP Study

Abstract Background Previous research established a solid link between atrial fibrillation (AF) and heart failure (HF). However, while two landmark trials have assessed AF screening for stroke prevention, the potential impact of AF screening on HF remains unclear. Purpose This secondary analysis of the LOOP Study aimed to investigate the effects of AF screening on HF events. Methods The LOOP Study was a randomized clinical trial evaluating long-term continuous AF screening with implantable loop recorder (ILR) in individuals aged 70-90 years and with ≥1 of arterial hypertension, diabetes mellitus, HF, and/or prior stroke. At inclusion, study participants were randomized 1:3 to either ILR screening or usual care. Anticoagulation was recommended upon AF detection, whereas other treatment was left to the discretion of the treating physician. In this secondary analysis, a cause-specific Cox regression model was applied to explore the effects of AF screening vs usual care on 1) time-to-first HF event and 2) total HF events, with the latter performed using the Andersen-and-Gill extension. HF event was defined as newly diagnosed HF with left ventricular ejection fraction ≤40% or any HF hospitalization in participants with prior diagnosis of HF with reduced ejection fraction. Results In total, 6004 participants were enrolled in the LOOP Study: 1501 in the ILR group vs 4503 in the Control group. The study population had a mean age of 75 years and slightly more males than females (53% vs 47%), whereas 266 (4.4%) were previously diagnosed with HF at baseline. During a median follow-up of 5.4 years, a total of 208 HF events were reported in 175 (2.9%) participants: 37 events among 34 (2.3%) participants in the ILR group vs 171 events among 141 (3.1%) participants in the Control group. For time-to-first HF event, the annualized event rate was 0.44%/year [95% confidence interval (CI): 0.30-0.62] vs 0.61%/year [95% CI: 0.51-0.72] for ILR vs Control, corresponding to a hazard ratio (HR) of 0.72 [95% CI: 0.50-1.05]. When assessing total HF events, a significantly lower event rate was observed in the ILR group (0.47%/year [95% CI: 0.33-0.65]) compared with the Control group (0.73%/year [95% CI: 0.62-0.84]), resulting in a HR of 0.65 [95% CI: 0.44-0.97]. Conclusions Among elderly individuals with additional stroke risk factors, long-term continuous AF screening led to a significantly reduced risk of total HF events, compared with usual care. This should be considered as hypothesis-generating and warrants further investigation.Cumulative incidence of HF event

Association of the pattern of new-onset post-operative atrial fibrillation with adverse clinical events over 1 year follow-up: insights from the VISION Cardiac Surgery Cohort

Abstract Background New-onset post-operative atrial fibrillation (AF) affects approximately 1 in 3 patients after cardiac surgery and is associated with adverse outcomes. Whether the amount of time spent in AF is associated with outcomes is unknown. Methods VISION Cardiac Surgery was an international, prospective cohort of patients who underwent cardiac surgery in 12 countries. We divided participants according to their pattern of AF: no new-onset AF, duration <24 h, duration 24-48 h, duration 48-72 h, duration >72 h and discharged in AF. We created a separate category for those who received electrical cardioversion. We excluded participants with a history of AF before surgery and participants with unknown duration of AF. We assessed the association of different patterns of AF, as compared to no AF, with clinical events occurring between 30 days post-operatively and 1 year of follow up. For each outcome, we created a Cox proportional hazards model adjusted for CHA2DS2-VASc score, smoking, hemodialysis, prior cardiac surgery, surgery type and antithrombotic use at hospital discharge (oral anticoagulation alone, oral anticoagulation plus antiplatelet, antiplatelet alone, or none). Results Among 12,234 eligible participants, 3,887 (30.3%) had new-onset post-operative AF and 3,711 had a known duration of AF. The proportion of patients with new-onset AF who were discharged from hospital on oral anticoagulation was 39%. Table 1 displays event rates by AF pattern. Compared to participants without AF, we observed a higher risk of a composite of stroke or vascular death at one year in participants with estimated AF duration >72 h (3.9%, adjusted hazard ratio (aHR) 1.72; 95% CI 1.03-2.87) and in those who underwent electrical cardioversion (3.7%, aHR 2.18; 95%CI 1.16-4.11). Only AF duration >72 h was associated with an increased risk of stroke (2.1%; aHR 2.42; 95%CI 1.15-5.09). We observed a higher risk of all-cause mortality in participants who were in AF at hospital discharge (4.1%, aHR 1.83; 95%CI 1.10-3.05) and in participants who received electrical cardioversion (5.8%, aHR 2.81; 95%CI 1.67-4.73). Only being in AF at discharge was associated with an increased risk of heart failure (3.8%; aHR 2.63; 95% CI 1.49-4.63). All patterns of AF were associated with AF detection in follow-up, with the strongest associations observed in participants who were in AF at hospital discharge (31.8%, aHR 73.4; 95%CI 51.4- 105.0) and in those who received electrical cardioversion (12.5%, aHR 17.9; 95%CI 11.1-28.8). Conclusions Among patients with new-onset post-operative AF following cardiac surgery, duration of AF >72 h, electrical cardioversion and being in AF at discharge are all markers of adverse events in the year following surgery. All patterns of AF confer a higher risk of AF detection in follow-up. Duration and treatment of early post-operative AF may have implications regarding the longer-term risk of adverse outcomes and recurrent AF.

Impact of photoplethysmography-based smartphone rhythm monitoring on atrial fibrillation detection and management after cardiac surgery: a pragmatic randomized trial

Abstract Introduction Postoperative atrial fibrillation (POAF) occurring after cardiac surgery is common and associated with adverse outcomes. Systematic monitoring of POAF beyond discharge is cumbersome. The emergence of photoplethysmography (PPG)-based rhythm monitoring with digital consumer devices could potentially mitigate these hurdles. Selected smartphone applications leveraging this technology have demonstrated excellent usability and accuracy for the detection of atrial fibrillation (AF). However, the impact of PPG-detected POAF on real-life clinical practice remains uncertain. Purpose To determine whether intermittent PPG-based smartphone rhythm monitoring, after being discharged home following cardiac surgery, impacts AF management. Methods The SURGICAL-AF 2 study is a pragmatic, investigator-initiated, open-label, multicenter, randomized clinical trial, conducted in three Belgian centers. The intervention group performed one-minute rhythm checks three times daily with a smartphone-based PPG application after hospitalization for cardiac surgery until the first follow-up visit with a cardiologist, scheduled at 21 – 91 days. Rhythm monitoring was not mandated in the usual care group. The primary endpoint was a composite of initiation of oral anticoagulation (OAC), cardioversion, up-titration or initiation of antiarrhythmic drugs (Vaughan-Williams class I or III) or implantation of a cardiac implantable electronic device (CIED). Secondary endpoints were incidence of POAF and actionable POAF, defined as a detection in patients with CHA2DS2-VASc score ≥2 for women or ≥1 for men who are not treated with OAC. Results Of the 450 patients randomized (238 patients in the intervention group and 212 patients in the usual care group; mean [SD] age, 64.1 [9.2] years; 96 women [21.3%]; 130 patients with AF before inclusion [28.9%]; 103 patients on OAC [22.9%], median [interquartile range] CHA2DS2-VASc score, 2 [1-3]), 98.7% completed the trial. In the intent-to-treat analysis, the primary end point occurred in 24 patients (10.3%) in the intervention group versus 5 patients (2.4%) in the usual care group (odds ratio (OR) 4.7, 95% CI, 1.8 - 12.6; P =.002). POAF was detected in 44 patients (18.8%) in the intervention group and was actionable in 25 patients (10.7%) versus 4 patients (1.9%) in the usual care group and actionable in 2 patients (0.9%). (POAF detection, OR 12.0, 95% CI, 4.2 - 34.5; P < .001; actionable POAF, OR 12.5, 95% CI, 2.9 - 52.6; P < .001) Conclusions In unselected patients discharged home following cardiac surgery, PPG-based smartphone monitoring revealed significantly more POAF which led to changes in AF management (OAC initiation, rhythm control therapy or CIED insertion). Longer follow-up is needed to determine whether these changes will lead to improved outcomes.

Diagnostic accuracy of smartwatches for automated atrial fibrillation detection over time: insights from the basel wearable study

Abstract Background The first smart-watch (SW) able to record a single lead ECG (SL-ECG) with automated atrial fibrillation (AF) detection algorithm was introduced 2018. Multiple other manufacturers followed. While manufacturers claim to continuously improve their devices’ algorithm, little is known about the impact on diagnostic accuracy on AF detection over time in a real-world cohort of patients. Purpose To assess if the diagnostic accuracy of algorithm-based AF detection of 5 smart devices has improved over time. Methods In this prospective study, consecutive patients undergoing electrophysiological procedures in a tertiary referral center were included between April 2021 to May 2023. Each participant received a 12-lead ECG and 5 SL-ECGs with different smart devices (AliveCor KardiaMobile, Apple Watch 6, Fitbit Sense, Samsung Galaxy Watch 3 and Withings ScanWatch). All SW were continuously updated for newest software or firmware as recommend by the manufacturers. The SW rhythm classification was compared to manual cardiologist-based interpretation of SL-ECGs. The change in number of inconclusive SL-ECGs by manufacturer’s algorithm were investigated over time via Cochran-Armitage test for trend and illustrated over time via logistic regression. Results 247 participants were included (30% female, mean age 66.3 [IQR: 57-74] years) resulting in a total of 1’235 SL-ECGs. Among these, 202 SL-ECGs (16%) were labeled as inconclusive by at least one smart-device. Inconclusive rates for individual SW ranged from 15% to 19%. When comparing the first and last quintile of patients in the study period, the inconclusive rates for the KardiaMobile were 14% vs. 6%, for the Apple Watch 14% vs. 6%, for the Fitbit Sense 20% vs. 14%, for the Samsung Galaxy 18% vs. 10% and for the Withings ScanWatch 22% vs. 14%, respectively. When assessed continuously over time, we observed a trend towards fewer inconclusive tracings, without reaching statistical significance (Figure 1). Conclusion Over 26 months, inconclusive tracings of 5 SW didn’t significantly decrease despite frequent software updates in a real-world patient cohort. Larger, long-term cohorts are needed to assess if automated AF detection has significantly improved over time.Figure 1

Effectiveness of photoplethysmography-based rhythm monitoring after atrial fibrillation ablation using a smartphone application

Abstract Introduction Recurrence of atrial fibrillation (AF) after AF ablation is frequent. Monitoring recurrence is essential for attributing patient symptoms and for assessing both AF burden and treatment effectiveness. The conventional approach relies on periodic (long-term) electrocardiograms (ECG) and is constrained by limited monitoring time, measurement dispersion, and cost. Furthermore, patients are unable to prioritize measurements based on symptoms. Photoplethysmography (PPG) emerges as a technology that could potentially transform any smartphone into a rhythm monitoring device, thereby mitigating the limitations associated with current practices. Selected applications leveraging this technology have demonstrated excellent usability and accuracy for the detection of AF. However, the effectiveness of PPG-based follow-up remains uncertain. Purpose To compare the effectiveness of digital follow-up utilizing a PPG-based smartphone application with conventional ECG-based follow-up for the detection of atrial arrhythmia recurrence after ablation. Methods Patients undergoing AF ablation were consecutively enrolled and monitored using a 24h-ECG at 3-6-12 months on top of ECGs conducted for clinical indications (conventional follow-up). Additionally, patients were instructed to perform PPG measurements twice daily or whenever symptoms were perceived, over the course of one year (digital follow-up). The primary outcome was recurrence of atrial arrhythmia. Results In total, 96 patients (69% male, mean age 64 [58 – 68] years) performed 39,895 PPG measurements. The compliance rate (n performed/prescribed) was 92.6% for ECG and 78.2% for PPG recordings. After one year follow-up, atrial arrhythmia recurrence was detected by conventional follow-up in 17.9% and in 38.5% of patients by digital follow-up (odds ratio, 3.4; 95% CI: 1.7-7.1). The confidence interval lower limit exceeded the predefined non-inferiority margin (p for non-inferiority > 0.001). Hence, superiority was tested (p for superiority = 0.001). The negative predictive value of digital follow-up for atrial arrhythmia detected with conventional follow-up was 98.3% (95% CI: 90.1-99.9%). Symptoms were reported in 81.1% of PPG-detected atrial arrhythmias and episodes sustained for 3.9 [1.1 – 22.7] hours. Conclusions Among patients with AF ablation, digital PPG-based rhythm monitoring with a smartphone, was non-inferior in the detection of atrial arrhythmia recurrence after ablation. Moreover, the detection of atrial arrhythmia was significantly increased compared to conventional ECG-based monitoring.

Comparative analysis of classical artificial intelligence and deep learning models for atrial fibrillation detection in electrocardiograms

Abstract Background/Introduction Early diagnosis of atrial fibrillation (AF) presents a challenging yet critical task for appropriate interventions aimed at reducing disease-related burden. In this context, strategies employing classical artificial intelligence (CAI) and deep learning (DL) have emerged as promising approaches to optimize cardiac disorder screening and detection. Purpose This study aimed to compare a CAI model and a DL model for the detection of AF in patients undergoing electrocardiographic (ECG) examinations in tertiary healthcare centers. Methods Between December 2022 and November 2023, a total of 135,476 ECGs were performed, comprising 5,067 with AF and 130,409 without AF. The ECGs were analyzed using both artificial intelligence models. The obtained results were then compared to the gold standard (cardiologist's report). In the CAI model, signals were extracted from ECG images, analyzing five key parameters: cardiac rhythm, atrial depolarization, atrioventricular conduction, ventricular depolarization, and ventricular repolarization (figure 1A). These parameters were benchmarked against the standard values from the Brazilian Society of Cardiology guidelines for detecting cardiac anomalies. Conversely, the DL model utilized a one-dimensional ResNet-based Convolutional Neural Network (CNN). This model was trained using ADAM optimization and binary cross-entropy loss, enabling the learning of complex patterns in the data (figure 1B). Results The mean age was 54.6 years (71.9 years with AF and 53.9 without AF). In the AF population, 52.2% were male (46% were male in the overall sample). In the analysis conducted, the CAI model showed a sensitivity and specificity of 90% and 62%, respectively, while the DL model had 90% and 69%, respectively. ROC curves were generated for both models, demonstrating the superior performance of the DL model (figure 2A). Conclusions Although the sensitivity remained similar between the models, the DL model distinguished itself with higher specificity. These results suggest that artificial intelligence, particularly the deep learning approach, holds promise as a supportive resource in AF diagnosis. However, further studies are needed to evaluate the models more thoroughly and determine their clinical applicability in a broader context.Artificial intelligence modelsROC curves

Performance of a vectorcardiographic deep learning algorithm compared to single-lead and 12-lead ECG for atrial flutter detection: implications for wearable devices

Abstract Background Wearable devices have developed the capacity to offer single-lead ECG analysis and are already widespread in use, particularly for detection of atrial fibrillation (1,2). Vectorcardiography entails the transformation of electrical vectors into X, Y, and Z axes to constitute a vectorcardiogram (VCG) (3). The VCG has diagnostic promise given the capacity to generate it without requiring clinical presentation for standard 12-lead ECG acquisition. Furthermore, a novel credit-card sized device capable of acquiring a VCG has recently been developed – raising the potential for arrhythmia detection and discrimination beyond wearable single-lead ECG systems. Objective To assess the performance of a deep learning algorithm trained on single lead and 12-lead ECGs and VCGs for detecting atrial flutter. Methods A convolutional neural network (AFL-CNN) was developed on a dataset of 4,000 ECGs and VCGs from a single US academic medical institution (80% training and 20% for testing). Validation was performed on a separate publicly available dataset (2223 ECGs). Model performance was assessed on single lead ECG analysis (lead I, obtained by removing the other leads) and standard 12-lead ECG. The gold standard for AFL diagnosis was physician-confirmed 12-lead ECG diagnoses in the dataset. The VCG was generated from the 12-lead ECG via Kors transformation matrix to which AFL-CNN was also applied. Model performance was assessed with sensitivity, specificity, positive and negative predictive values, accuracy, and F1 statistic. Results On single lead ECG analysis, AFL-CNN achieved 71.2% sensitivity and 96.9% specificity with PPV 84.0% and NPV 93.7% and 92.2% accuracy with F1 83.2%. When AFL-CNN was applied to the VCG, performance improved: 91.0% sensitivity and 98.7% specificity, PPV 94.24%, NPV 98.0%, 97.3% accuracy, and F1 95.31%. Lastly, the performance remained high when AFL-CNN was applied to the 12-lead ECG: 95.4% sensitivity, 98.63% specificity, PPV 94.0%, NPV 99.0%, and accuracy of 98.0%, with F1 97.2%. Conclusion A deep learning algorithm applied to vectorcardiography for the detection of atrial flutter showed improved performance over single lead analysis and comparable performance to standard 12-lead applications. The finding of improved model performance beyond single lead ECG tracings underscores both the limitations of current wearable technologies in arrhythmia detection beyond atrial fibrillation and the potential for VCG-based algorithm prediction to fill gaps in healthcare inequality when standard 12-lead acquisition is challenging to obtain.

Continuing evaluation of atrial fibrillation detection after cryptogenic stroke: one-year findings from a multi-center study with confirm Rx ICM

Abstract Background The detection of atrial fibrillation (AF) after a cryptogenic stroke (CS) carries crucial therapeutic implications. While clinical trials have meticulously explored the incidence of AF in CS patients, real-world data on this topic remain relatively sparse. In this study, we aimed to assess the incidence of AF among patients with cryptogenic stroke, who underwent the implantation of an insertable cardiac monitor (ICM). The abstract reports the interim results from the CS sub-study of the Smart Registry. Methods A prospective, single-arm, multi-center registry was conducted to identify AF in patients with cryptogenic stroke using the Confirm R ICM. Inclusion criteria comprised participants aged 40 years or older who had experienced cryptogenic strokes within a 90-day window. Prior to enrollment, comprehensive diagnostic assessments were performed to meticulously rule out all potential sources of embolism, in accordance with the minimal standards of the American and European Guidelines. At each follow-up visit, expert electrophysiologists reviewed and adjudicated ICM detected AF episodes. The primary endpoint was the cumulative incidence of true device-detected AF (lasting more than 30 seconds) at 6 months. Kaplan-Meier curves were employed to assess the cumulative incidence of true device-detected AF through 12 months. Results In our study, 117 patients were enrolled across 19 global sites from September 2021 to August 31, 2023. Kaplan-Meier analysis revealed a 21.3% probability (confidence interval (CI): 14.5%, 30.7%) of true AF episodes lasting more than 30 seconds at 6 months, increasing to 31.0% (CI: 22.0%, 42.5%) at 12 months. Similar results were observed in episodes lasting over 120 seconds at both 6 months (20.7%) and 12 months (30.5%). In average, 14.4 true AF episodes per patient were detected per year. The median time from implantation to AF detection (> 30 seconds) was 31.5 days (interquartile range, 7 to 191). Conclusion Our ongoing study illustrates the effectiveness of the Confirm Rx ICM in detecting atrial fibrillation after cryptogenic stroke. These results confirm the diagnostics value of long-term continuous AF monitoring in cryptogenic stroke patients.

Non-invasive detection of cardiac arrhythmias using photoplethysmography signals

Abstract Cardiac arrhythmias (CAs) are associated with critical heart-related complications such as stroke or heart failure. Because of the intermittent and asymptomatic presentation of some CAs in their early stages, the screening remains limited using traditional methods based on ECG. Recent advances in photoplethysmography (PPG) have revealed substantial potential for wearable devices to detect CAs within large populations. Although PPG has demonstrated efficiency in discriminating atrial fibrillation (AF) from normal sinus rhythm, it remains unclear whether AF detection methods remains effective in the presence of CAs other than AF. In this study, we applied a recurrent neural network on a dataset containing eight different types of CAs (Table 1). The classifier processes sequences of interbeat intervals (IBIs) as input and discriminates between normal and abnormal rhythms. The algorithm was evaluated on 64 patients (45 males and 19 females , with a mean age of 55.9 ± 16.0) undergoing a diagnostic or therapeutic electrophysiological procedure. This dataset includes simultaneous recordings of PPG signals from a wrist-bracelet and 12-lead ECG, with the latter used as the gold-standard for annotating cardiac rhythms. The classifier achieved 84% accuracy, 77% sensitivity and 88% specificity in detecting abnormal rhythms. Table 1 shows that AF (99.6%), atrial tachycardia (100%) and AVRT (96.4%) were well classified as abnormal rhythm, whereas the reliability of the detection decreased for atrial flutter (65.4%), atrial or ventricular bigeminy (72.4%) and ventricular tachycardia (80.2%). Undetected abnormal rhythms were often characterized by a rather regular rhythm (illustrated by VT in Figure 1) and CAs for which PPG heartbeat detection was initially not optimal (Bigeminy in Figure 1). In conclusion, this study shows the capability of PPG-based technology to detect various CAs extending beyond AF. It highlights the merits and limitations of IBI-based detection of abnormal rhythms, highlighting the need for a better comprehension of the peripheral hemodynamic signature of CAs.

Post-operative atrial fibrillation after noncardiac surgery: diagnostic role of ambulatory rhythm monitoring for recurrent arrhythmia after hospital discharge

Abstract Introduction Postoperative atrial fibrillation (POAF) after noncardiac surgery constitutes 10-15% of all cases of incident atrial fibrillation (AF) in the community and has been associated with an increased risk of stroke. The risk of subsequent AF among patients with POAF is not well defined. Purpose To evaluate the role of systematic 30-day electrocardiographic (ECG) monitoring in the detection of AF recurrence after POAF. Methods In a prospective interventional study, patients who underwent major noncardiac surgeries and had new AF documented by ECG in the post-operative period between May 2022 and April 2023 were identified. For these patients, a 30-day ambulatory ECG monitor was recommended upon discharge, with final ordering deferred to the primary inpatient teams. Rates of recurrent AF of any duration detected by 30-day monitoring plus routine care versus routine care alone after the surgery were compared. Results Eighty-five patients with new-onset POAF were included (mean age 72 years, male 67%). Of these, 35 (41%) patients underwent up to 30 days [median 23 (IQR 6-27) days] of continuous ECG monitoring after hospital discharge. There were no significant differences in baseline characteristics between patients who did or did not undergo monitoring (Table 1). During median 281 (IQR 197-385) days of follow-up, recurrent AF was documented in 14/35 (40%) patients in the monitoring group as opposed to 7/50 (14%) patients in the usual care group (log-rank p=0.003, Figure 1). The median time from day of surgery until documented AF recurrence was 25 (IQR 6-53) days in the monitoring group and 66 (IQR 25-145) in the usual care group. Overall, 81% of recurrent AF events in both groups were detected in the first 3 months after surgery. The median burden of AF in the cardiac monitoring group was 7% (IQR 1-15), with the longest AF event lasting a median of 6 (IQR 2-18) hours. Anticoagulation for AF was prescribed in 24/64 (37%) patients with POAF without recurrence and in 16/21 (76%) patients with recurrent AF. Conclusion Recurrent AF following POAF is common and detected most frequently in the first 3 months after noncardiac surgery. Systematic ambulatory rhythm monitoring is associated with a higher rate of AF detection following hospital discharge and may identify a patient subgroup who might benefit from anticoagulation therapy.Table 1Figure 1

The impact of cardiac rehabilitation on atrial fibrillation recurrence after pulmonary vein isolation: results of a retrospective study

Abstract Background Although pulmonary vein isolation (PVI) is an extensively utilized atrial fibrillation (AF) treatment, many patients still experience AF recurrences. Multimorbidity care, including physical activity, may reduce recurrences. In Belgium, patients undergoing an ablation are entitled to a cardiac rehabilitation (CR) program after PVI, but not all do participate. More evidence that such (standardized) CR programs for AF patients after ablation can positively impact rhythm control is desirable. Purpose This study retrospectively investigated the impact of CR on AF recurrence after PVI. Methods We analysed an extensive database of a large Belgian tertiary care center with patients who underwent a primary PVI between 2007 and 2020 and who had a follow-up of at least 1 year. General cardiology follow-up included a 24h Holter monitoring 3 and 12 months after PVI. Patients were retrospectively divided into an intervention and control group based on their choice to participate in a standardized CR program. The Kaplan-Meier method was used to estimate the time-to-recurrence and event-free survival curves after the primary ablation. Differences between curves were tested with the log-rank test. Mann-Whitney U test was used to analyze continuous variables and the Chi-square test for categorical variables. Results In total, 1767 patients were included, of which 589 participated to CR and 1178 served as controls (median 64 (56-70) y, 68.9% male, median BMI 27.5 (24.7-30.7) kg/m2). There was no significant difference between CR and controls in median age (64y vs. 63y respectively; p=0.203), the proportion of male (57.9% vs. 47.0% respectively; p=0.111) and the proportion of paroxysmal AF (66.4% vs. 70.6% respectively; p=0.210). The proportion with recurrent AF at the end of study follow-up (with a three-month initial blanking period) in the CR and control group was similar (32.6% vs. 35.1% respectively; p=0.288), and also survival analysis demonstrated no significant difference (p=0.827; Figure 1). We noted a cross-over of the recurrence curves after about 1500 days, but the rate of AF recurrence was not different among CR and controls within the first year post-ablation (p=0.136) nor when analysed from day 365 until the end of the study (p=0.341) (Figure 2). There was no significant difference in AF recurrences between CR and controls for either paroxysmal AF (p= 0.551) or persistent AF (p=0.181) based on the survival analysis. Conclusions This retrospective analysis could not demonstrate reduced AF recurrence in post-PVI patients following a structured CR program. The unexpected crossover of curves could point to earlier AF detection in CR patients, while there might be long-term benefit of physical activity. Multivariable analysis will have to evaluate whether impact of CR is related to certain clinical patient characteristics, which might also have contributed to inclusion bias.Figure 1Figure 2

Accuracy of the OMRON HEM-7361T blood pressure monitor in the differentiation between atrial fibrillation and sinus rhythm.

We investigated the accuracy of the OMRON HEM-7361T automated oscillometric blood pressure (BP) monitor in the differentiation between atrial fibrillation and sinus rhythm. An approximately equal number of patients with persistent atrial fibrillation and individuals with sinus rhythm were recruited from outpatients and inpatients of Ruijin Hospital, Shanghai, China. BP was measured three times consecutively with a 30-s interval with the OMRON HEM-7361T automatic electronic BP monitor for atrial fibrillation detection. A hand-held single lead electrocardiogram device was used for simultaneous electrocardiogram recordings. The device accurately identified atrial fibrillation in 100 (99.0%) of the 101 patients, with only 1 patient incorrectly classified as non-atrial fibrillation. The device correctly identified 99 (95.2%) of the 104 participants with sinus rhythm as non-atrial fibrillation, with five participants incorrectly classified as atrial fibrillation. The device had a positive predictive value of 95.2%, negative predictive value of 99.0%, and overall accuracy of 97.1%. Among the six misclassified participants, one with atrial fibrillation had a heart rate of 65 beats/min, and four of the five participants with sinus rhythm had cardiac arrhythmias (atrial or ventricular premature beat in one participants, sinus tachycardia in one participant, and both arrhythmias in one participant). The OMRON HEM-7361T BP monitor is accurate in the differentiation between atrial fibrillation and sinus rhythm. Whether the device is sufficiently accurate in the differentiation between atrial fibrillation and other cardiac arrhythmias remains under investigation.

MANAGEMENT OF REMOTE MONITORING FOR IMPLANTABLE DEVICES CONTROLS: SINGLE CENTER EXPERIENCE

Abstract Background Nowadays the benefits of remote monitoring (RM)of cardiac implantable electronic devices (CIEDs) are Known: an improved both device surveillance and patient clinical management and a decrease in office follow up. So RM is used as supplement in clinic follow up for patients with CIED.However despite decades to RM utilization there are challenges in managing patients with RM including large number of trasmissions and lack of trained staff. There are many different model of care to RM management. Methods We describe our model of care to management CIED recipients with RM considering data collected from January 2021 to November 2023. Results Our model of care consists of 4 electrophysiologists Physicians and 4 Tecnicians. RM enrollment is performed at scheduled in office visit after ten days from hospital discharge after device implantation. implantable cardiac device(ICD)recipients with RM are annualy followed in office and biennal for pacemaker (PM) recipients and they send every three months transmissions. Trained tecnicians gives home monitor, take care of patients education and collected and screen transmissions. Tecnicians inform immediately physicians in case of urgent alert (shock or antitachicardia pacing therapy (ATP), sustained ventricular tachycardia(TV), ventricular fibrillation (Fv) indicator of elective replacement (ERI), lead releted alert or noise). Physician contact patients the same day (exept Saturday and Sunday). In addiction Physicians twice a week review less urgent and scheduled transmissions and call the patients by phone for visit in office for patients with low rate of biventricular pacing, detection of atrial fibrillation. We have 1000 patients with RM: 746 ICD recipients (103 Biotronik, 150 Abbott, 200 Boston scientific, 293 medtronic), 50 implantable cardiac monitor (ICM) recipients and 204 PM recipients. We analyze 12721 transmissions.Considering ICD recipient we detect 96 episode of ATP and 31 of shock. Shock of ICD was registered in 52 patients (6,9%)We received 1982 atrial fibrillation (AF) detection, with 35 for new detectable AF.The most of transmisisone were for low bIV pacing. We contacted these patients by phone about clinical status and in some case we called them in office. Conclusions Our model of care allows to control a large number of trasmissions with fast management of urgent alert trasmissions.

Machine Learning Modeling to Predict Atrial Fibrillation Detection in Embolic Stroke of Undetermined Source Patients.

In patients with embolic stroke of undetermined source (ESUS), occult atrial fibrillation (AF) has been implicated as a key source of cardioembolism. However, only a minority acquire implantable cardiac loop recorders (ILRs) to detect occult paroxysmal AF, partly due to financial cost and procedural inconvenience. Without the initiation of appropriate anticoagulation, these patients are at risk of increased ischemic stroke recurrence. Hence, cost-effective and accurate methods of predicting AF in ESUS patients are highly sought after. We aimed to incorporate clinical and echocardiography data into machine learning (ML) algorithms for AF prediction on ILRs in ESUS. This was a single-center cohort study that included 157 consecutive patients diagnosed with ESUS from October 2014 to October 2017 who had ILR evaluation. We developed four ML models, with hyperparameters tuned, to predict AF detection on an ILR. The median age of the cohort was 67 (IQR 59-74) years old and the median monitoring duration was 1051 (IQR 478-1287) days. Of the 157 patients, 32 (20.4%) had occult AF detected on the ILR. Support vector machine predicted for AF with a 95% confidence interval area under the receiver operating characteristic curve (AUC) of 0.736-0.737, multilayer perceptron with an AUC of 0.697-0.708, XGBoost with an AUC of 0.697-0.697, and random forest with an AUC of 0.663-0.674. ML feature importance found that age, HDL-C, and admitting heart rate were important non-echocardiography variables, while peak mitral A-wave velocity and left atrial volume were important echocardiography parameters aiding this prediction. Machine learning modeling incorporating clinical and echocardiographic variables predicted AF in ESUS patients with moderate accuracy.