Signal-Averaged Electrocardiography (SAECG) and Artificial Intelligence Algorithmic Electrocardiogram for Cardiovascular-Related Diseases
Number: 0664
Table Of Contents
PolicyApplicable CPT / HCPCS / ICD-10 Codes
Background
References
Policy
Scope of Policy
This Clinical Policy Bulletin addresses signal-averaged electrocardiography (SAECG) and artificial intelligence algorithmic electrocardiograms for detection of cardiovascular-related diseases.
-
Experimental and Investigational
Aenta considers the following procedures experimental and investigational because the effectiveness of these approaches has not been established:
- Signal-averaged electrocardiography because no prospective clinical studies have demonstrated the utility of this testing in improving clinical outcomes;
- Remote algorithmic analysis of electrocardiographic-derived data (Premier Heart's Multifunction Cardiogram (MCG); also known as 3DMP Computerized EKG System) because the clinical value of the system in managing persons suspected of having significant coronary artery disease has not been established;
- Assistive artificial intelligence algorithmic electrocardiogram assessment (e.g., MyoVista Wavelet ECG Cardiac Testing Device) for evaluation of cardiac dysfunction because the clinical value of this approach has not been established.
-
Related Policies
Background
Signal-Averaged Electrocardiography (SAECG)
Signal-averaged electrocardiography (SAECG) is a technique involving computerized analysis of segments of a standard electrocardiogram that allows the detection of ventricular late potentials. Ventricular late potentials in patients with cardiac abnormalities, especially coronary artery disease or following an acute myocardial infarction (MI), have been associated with an increased risk of ventricular tachyarrhythmias and sudden cardiac death. Proponents of SAECG claim that it can obviate the need for invasive techniques commonly used to identify high-risk patients for interventions that treat or prevent ventricular tachyarrhythmia and sudden death.
An Agency for Healthcare Policy and Research's assessment (AHCPR, 1998) found that the current data on SAECG show relatively consistent high negative-predictive values, poor positive-predictive values, and variable sensitivity and specificity when the technique is used on patients with cardiomyopathy or following a MI. However, the high negative- predictive value of SAECG is largely due to the fact that the incidence of fatal arrhythmic events among post-MI patients is now below 10 %. The incidence of fatal arrhythmias has declined among post-MI patients, a large percentage of whom are on anti-thrombotic therapy, most likely following the trend of decreased mortality rate following MI.
In 1996, an American College of Cardiology (ACC) consensus statement on SAECG concluded that SAECG has "established value" in assessing the risk of development of sustained ventricular arrhythmias in patients recovering from MI. However, subsequently published guidelines from the ACC on management of acute MI (1999) stated that the usefulness of SAECG for risk assessment after MI is less well-established by evidence/opinion. In addition, subsequently published ACC guidelines on implantable anti-arrhythmia devices (1998) do not recommend SAECG for selecting patients for automated implantable cardioverter defibrillators (AICDs).
Although it has been proposed that SAECG may be used to select post-MI patients for anti-arrhythmic drugs or AICD implantation, there are no prospective clinical studies demonstrating the clinical utility of SAECG in selecting patients for these therapies. In addition, there are no prospective clinical studies proving that SAECG can be used successfully to select patients for electrophysiologic studies or Holter monitoring, or to use SAECG for risk stratification in lieu of these other tests.
Grimm et al (2003) studied arrhythmia risk stratification with regard to prophylactic implantable cardioverter-defibrillator patients with in idiopathic dilated cardiomyopathy (IDC). These researchers concluded that reduced left ventricular ejection fraction (LVEF) and lack of beta-blocker use are important arrhythmia risk predictors in IDC, whereas SAECG, baroreflex sensitivity, heart rate variability, and T-wave alternans do not seem to be helpful for arrhythmia risk stratification. Furthermore, in a review on electrocardiographic arrhythmia risk testing, Engel et al (2004) evaluated the various electrocardiographic (ECG) techniques that appear to have potential in assessment of risk for arrhythmia. The resting ECG (premature ventricular contractions, QRS duration, damage scores, QT dispersion, and ST segment and T wave abnormalities), T-wave alternans, late potentials identified on SAECG, and heart rate variability were explored. The authors stated that unequivocal evidence to support the widespread use of any single non-invasive technique is lacking; further research in this area is needed.
Guidelines from the European Society for Cardiology (Brignole, et al., 2004) concluded that the systematic use of SAECG in syncope is "not recommended."
Tamaki and colleagues (2009) prospectively compared the predictive value of cardiac iodine-123 metaiodobenzylguanidine (MIBG) imaging for sudden cardiac death (SCD) with that of the SAECG, heart rate variability (HRV), and QT dispersion in patients with chronic heart failure (CHF). At entry, cardiac MIBG imaging, SAECG, 24-hr Holter monitoring, and standard 12-lead ECG were performed in 106 consecutive stable CHF outpatients with a radionuclide LVEF less than 40 %. The cardiac MIBG washout rate (WR) was obtained from MIBG imaging. Furthermore, the time and frequency domain HRV parameters were calculated from 24-hr Holter recordings, and QT dispersion was measured from the 12-lead ECG. During a follow-up period of 65 +/- 31 months, 18 of 106 patients died suddenly. A multi-variate Cox analysis revealed that WR and LVEF were significantly and independently associated with SCD, whereas the SAECG, HRV parameters, or QT dispersion were not. Patients with an abnormal WR (greater than 27 %) had a significantly higher risk of SCD (adjusted hazard ratio: 4.79, 95 % confidence interval: 1.55 to 14.76). Even when confined to the patients with LVEF greater than 35 %, SCD was significantly more frequently observed in the patients with than without an abnormal WR (p = 0.02). The authors concluded that cardiac MIBG WR, but not SAECG, HRV, or QT dispersion, is a powerful predictor of SCD in patients with mild-to-moderate CHF, independently of LVEF.
Park and colleagues (2009) examined the correlation between parameters of 2-dimensional ECG and SAECG in patients with arrhythmogenic right ventricular cardiomyopathy (ARVC). A total of 33 patients (13 females, 40.3 +/- 14.4 years old) were included in this study. Both the right and left ventricular dimensions and systolic function were assessed with 2-dimensional ECG. The SAECG was performed with high-gain amplification and filtered using bi-directional Butterworth filters between 40 and 250 Hz. The right ventricular (RV) outflow tract was the most frequently (n = 18, 54 %) involved segment. Six (18 %) patients had only mildly decreased RV systolic function. All the other patients had normal RV systolic function. Although localized left ventricular wall motion abnormalities were observed in 14 (42 %) patients, the LVEF was normal in most (n = 32, 97 %). Late potentials were positive in 22 (63 %) patients. There was no significant correlation between parameters of the SAECG and 2-dimensional ECG for the entire patient population. The authors concluded that the SAECG parameters exhibited no correlation to any of 2-dimensional ECG parameters in the patients with ARVC. Fragmented electrical activity may develop with no significant relation to the anatomical changes in the patients with ARVC.
The Agency for Healthcare Research and Quality's systematic review of ECG-based signal analysis technologies for evaluating patients with acute coronary syndrome (Coeytaux et al, 2012) concluded that "Existing research is largely insufficient to confidently inform the appropriate use of ECG-based signal analysis technologies in diagnosing coronary artery disease (CAD) and/or ACS. Further research is needed to better describe the performance characteristics of these devices to determine in what circumstances, if any, these devices might precede, replace, or add to the standard ECG in test strategies to identify clinically significant CAD in the patient population of interest. To fully assess the impact of these devices on diagnostic strategies for patients with chest pain, test performance needs to be linked to clinically important outcomes through modeling or longitudinal studies".
Proclemer et al (2013) examined the current clinical practice of screening and risk evaluation for SCD in ischemic and non-ischemic cardiomyopathy with a focus on selection of candidates for ICD therapy, timing of ICD implantation, and use of non-invasive and invasive diagnostic tests across Europe. A systematic screening program for SCD existed in 19 out of 31 centers (61.3 %). Implantation of ICDs according to the inclusion criteria of MADIT-II and SCD-HeFT trials was reported in 30 and 29 % of centers, respectively, followed by MADIT-CRT (18 %), COMPANION (16 %), and combined MADIT and MUSTT (7 %) indications. In patients with severe renal impairment, ICD implantation for primary prevention of SCD was always avoided in 8 centers (33.3%), was not used only if creatinine level was greater than 2.5 mg/dL in 10 centers (32.2 %), and in patients with permanent dialysis in 8 centers (33.3 %). Signal-averaged electrocardiography and heart rate variability were never considered as risk stratification tools in 23 centers (74.2 %). Implantation of a loop recorder was performed in patients with borderline indications for ICD therapy in 6 centers (19.4 %), for research purposes in 5 (16.1 %), and was never performed in 20 (64.5 %) centers. The authors concluded that the majority of participating European centers have a screening program for SCD and the selection of candidates for ICD therapy was mainly based on the clinical risk stratification and not on non-invasive and invasive diagnostic tests or implantable loop recorder use.
Furthermore, an UpToDate review on "Clinical applications of the signal-averaged electrocardiogram: Overview" (Narayan and Cain, 2014) states that "Guideline Recommendations – We agree with the 2008 American Heart Association (AHA)/American College of Cardiology (ACC)/Heart Rhythm Society (HRS) scientific statement on noninvasive risk stratification and the 2006 ACC/AHA/European Society of Cardiology (ESC) guidelines for management of patients with ventricular arrhythmias, which concluded that the SAECG may be useful to identify patients at low risk for SCD, but its routine use to identify patients at high risk for SCD is not yet adequately supported. Similarly, the 2006 AHA/ACC scientific statement on syncope concluded that routine use of T-wave alternans combined with signal-averaged ECG and assessment of heart rate variability in patients with syncope and a negative initial evaluation is not yet established and currently is not indicated".
An UpToDate review on "Use of the signal-averaged electrocardiogram in nonischemic heart disease and cardiac transplantation" (Narayan and Cain, 2015) concludes that:
- Data are conflicting on the efficacy of SAECG in predicting clinical outcome or ventricular arrhythmias in patients with non-ischemic dilated cardiomyopathy.
- There are insufficient data to recommend the use of the SAECG for risk stratification of patients with non-ischemic cardiomyopathy.
- Although small studies have identified SAECG alterations in patients with cardiac transplant rejection, the utility of SAECG for detection of rejection has not been established.
Dinov and colleagues (2016) correlated SAECG with the endocardial scar characteristics in patients with ischemic ventricular tachycardia (VT). These researchers suggested that successful catheter ablation (CA) can result in normalization of the SAECG. A total of 50 patients (42 men; aged 67 ± 10 years, EF 34 ± 12 %) with ischemic VTs were prospectively enrolled; SAECG was performed before and after CA. Patients with at least 2 abnormal criteria (filtered QRS greater than or equal to 114 ms; root mean square 40 less than 20 μV, and low-amplitude potentials 40 greater than 38 ms) were defined as having positive SAECG. There was a linear correlation between endocardial scar area (less than 1.5 mV) and filtered QRS (r = 0.414; p = 0.003); CA resulted in normalization of the SAECG in 6 patients. In patients with filtered QRS less than or equal to 120 ms, 13 (40.6 %) patients had normal SAECG after CA compared with 7 (21.9 %) before ablation (p = 0.034). Patients with normal or normalized SAECG after CA had better VT-free survival compared with those whose SAECG remained abnormal. Abnormal SAECG after CA was a predictor for VT recurrence: hazard ratio (HR) = 3.64; p = 0.039 for the overall population, and HR = 5.80; p = 0.022 for patients having QRS less than or equal to 120 ms. The authors concluded that there was a significant correlation between the surface SAECG and endocardial scar size in patients with ischemic VTs. A successful CA could result in normalization of SAECG that was associated with more favorable long-term outcomes.
The main drawbacks of this study were its small sample size (n = 50) and the relatively short follow-up (median of 12 months). The authors stated that this study should be considered as a hypothesis-generating one. The presented results are valid for patients with ischemic heart disease and must be confirmed in other clinical conditions (e.g., dilated cardiomyopathy, and arrhythmogenic right ventricular dysplasia). They noted that as long as the post-ablation SAECGs were recorded before the hospital discharge, it remained unclear if the VT recurrences during the follow-up were accompanied by perturbations in the SAECG. The localization of the scar may influence the sensitivity of the method because the abnormal low-amplitude potentials less than 40 μV and root mean square voltage in the last 40 ms of the filtered QRS were less pronounced in patients with anterior or septal MIs.
Gatzoulis and colleagues (2018) noted that SAECG records delayed depolarization of myocardial areas with slow conduction that can form the substrate for monomorphic VT. This technique has been examined mostly in patients with CAD, but its use has been declined over the years. However, several lines of evidence, derived from clinical data in patients with healed MI, indicated that SAECG remains a valuable tool in risk stratification, especially when incorporated into algorithms encompassing invasive and non-invasive indices. Such an approach can aid the more precise identification of candidates for device therapy, in the context of primary prevention of SCD. These investigators examined the value of SAECG as a predictor of arrhythmic outcome in patients with ischemic heart disease and discussed potential future indications. These researchers stated that given the relative paucity of data, clinical studies are needed, examining the prognostic value of SAECG in post‐MI patients treated with primary percutaneous coronary interventions (PCIs), even in the absence of significant LV dysfunction. The authors stated that late potentials (LPs) in primary electrical disorders, such as Brugada syndrome, are intriguing, however, the underlying pathophysiology and clinical significance are still under investigation.
SAECG for Brugada Syndrome
Nagamoto and colleagues (2017) stated that ventricular fibrillation (VF) and atrial fibrillation (AF) are well-known arrhythmias in patients with Brugada syndrome. These researchers evaluated the characteristics of the atrial arrhythmogenic substrate using (SAECG in patients with Brugada syndrome. SAECGs were performed during normal sinus rhythm in 23 normal volunteers (control group), 21 patients with paroxysmal AF (PAF group), and 21 with Brugada syndrome (Brugada group). The filtered P wave duration (fPd) in the control, Brugada, and PAF groups was 113.9 ± 12.9 ms, 125.3 ± 15.0 ms, and 137.1 ± 16.3 ms, respectively. The fPd in the PAF group was significantly longer compared to that in the control and Brugada groups (p < 0.05). The fPd in the Brugada group was significantly longer than that in the control group (p < 0.05) and significantly shorter than that in the PAF group (p < 0.05). The authors concluded that patients with Brugada syndrome had abnormal P waves on the SAECG. The abnormal P waves on the SAECG in Brugada syndrome patients may have intermediate characteristics between control and PAF patients.
The authors stated that this study had 2 drawbacks. First, although there is a well-known potential correlation with a susceptibility to AF in Brugada patients, the abnormal fPD on the SAECG was not correlated with the development of AF in this study as only 2 patients had a history of AF. The patient number in the Brugada group was rather small with 21 patients, which might have affected these findings. Moreover, this study was a retrospective analysis, thus, a prospective analysis is needed to evaluate the development of AF in Brugada patients with an abnormal fPD. Second, a coved- or saddle back-type ST segment elevation was determined by the electrogram (ECG) at the time of the SAECG. There was some temporal variability in the ECGs in Brugada syndrome patients. Thus, this timing issue could explain the lack of a difference in the fPd between the patients in the Brugada group with the 2 types of ST segment elevation.
SAECG for Prediction of Recurrences after Catheter Ablation of Ventricular Arrhythmias in Arrhythmogenic Right Ventricular Dysplasia/Cardiomyopathy
Liao and colleagues (2017) noted that the changes of SAECG in patients with arrhythmogenic right ventricular dysplasia/cardiomyopathy (ARVD/C) undergoing radiofrequency catheter ablation (RFCA) of ventricular arrhythmias (VAs) remains unknown. Between 2010 and 2014, a total of 81 ARVD/C patients underwent endocardial and/or epicardial RFCA for drug-refractory VAs; 70 patients (mean age of 46.2 ± 14.1 years, 37 men) achieving acute procedure success (negative inducibility) were enrolled. Baseline characteristics, non-invasive examinations and SAECG (before and 3 months after RFCA) were collected retrospectively. After successful RFCA, the electrical parameters of SAECG changed in 39 patients (55.7 %), including 28 patients (40 %) with electrical regression (group 1), and 11 patients (15.7 %) with electrical progression (group 3); 31 patients (44.3 %) showed no significant SAECG change (group 2). During a mean follow-up of 17.8 ± 10.7 months, 23 patients (32.9 %) had VA recurrences, including 4 in group 1, 12 in group 2, and 7 in group 3. In comparisons with groups 2 and 3, group 1 patients had a significantly better VA recurrence-free survival (p = 0.02). In multi-variable Cox regression analysis, electrical regression was found to be associated with fewer VA recurrences (p = 0.02, odds ratio [OR]: 0.28, 95 % CI: 0.10 to 0.83). The authors concluded that electrical regression of SAECG after RFCA in ARVD/C was found to be associated with fewer VA recurrences. This was a retrospective study with a relatively small sample size (n = 70). Well-designed studies with larger sample size are needed to validate these findings.
Multifunction Cardiogram
The Premier Heart digital database-driven multi-phase (3DMP) electrocardiograph (EKG) System provides a computer analysis of digitalized 12-lead EKG waveforms in the frequency domain (power spectral estimate) to aid in the detection of significant coronary artery disease. The 3DMP system was cleared by the Food and Drugs Administration (FDA) based on a 510(k) application. Weiss et al (2002) reported on a cross-sectional analysis of the use of the 3DMP system in 136 patients with symptoms of potential coronary artery disease who were scheduled for angiography. Originally, 200 patients were selected for the study, but 64 of the patients were not included in the study because of various technical problems in their 3DMP readings.
Although the 3DMP system was positive for CAD in 76 of 78 patients with greater than 60 % narrowing by angiography, the 3DMP system also read positive in 8 of 12 patients with 40 to 60 % narrowing. None of the 10 patients with greater than 0 to 40 % narrowing read as positive by the 3DMP system, but 8 of 36 patients with 0 % narrowing read as positive for CAD.
As a significant number (2 of 78) of patients with significant angiographic lesions were missed by the 3DMP system, it is not clear that the device is sufficiently accurate to either be used in lieu of angiography or to select patients for angiography.
There are no evidence-based guidelines from national professional organizations that address the clinical utility of 3DMP in evaluating patients suspected of having coronary artery disease. Prospective clinical studies are necessary to demonstrate the clinical utility of the 3DMP system in managing patients suspected of having significant coronary artery disease.
A technology assessment prepared for the AHCPR on ECG-based signal analysis technologies (Coeytaux et al, 2010) stated that the reliability and test performance of 3DMP in subjects at high-risk or with known CAD is promising. The horizon scan identified 7 potentially relevant devices, including 3 that use body surface mapping and 1 that uses mathematical signal analysis. Of the 7 devices, only the PRIME ECG by Heartscape Technologies (body surface mapping) and the 3DMP/MCG/ mfEMT by Premier Heart (mathematical signal analysis; referred to as the 3DMP) are cleared for marketing by the FDA and commercially available. One body surface mapping device (Visual ECG/Cardio3KG by NewCardio) is commercially available but not cleared; the other devices are not commercially available. The assessment concluded: "There is currently little available evidence that pertains to the utility of ECG-based signal analysis technologies as a diagnostic test among patients at low to intermediate risk of CAD who present in the outpatient setting with the chief complaint of chest pain. The limited evidence that is available demonstrates proof of concept, particularly for the PRIME ECG and 3DMP devices. Further research is needed to better characterize the performance characteristics of these devices to determine in what circumstances, if any, these devices might precede, replace, or add to the standard ECG for the diagnosis of CAD among patients who present with chest pain in the outpatient setting. The randomized controlled trial (RCT) study design is best suited for evaluating the impact that ECG-based signal analysis technologies may have on clinical decision-making and patient outcomes, but there are indirect approaches that might be applied to answer these questions".
Kawaji and colleagues (2015) stated that multifunction cardiogram (MCG) is a computer-enhanced, resting electrocardiogram analysis developed to detect hemodynamically relevant CAD. Based on data from previous studies suggesting excellent diagnostic accuracy in detecting CAD, MCG (approved by the FDA) received a Current Procedure Terminology (CPT) code in 2010 in United States. However, there is no previous study validating MCG by using fractional flow reserve (FFR) as the reference standard. Multifunction cardiogram Evaluation in Diagnosis of Functional coronary Ischemia sTudy (MED-FIT) was designed as a single-center, prospective study enrolling 100 stable patients with suspected CAD scheduled for coronary angiography. The primary and secondary analyses evaluated the diagnostic performance of the MCG severity score to detect functional myocardial ischemia by FFR less than or equal to 0.80, and angiographically significant coronary stenosis (percent diameter stenosis greater than or equal to 50 %) by quantitative coronary angiography. The current analysis set consisted of 91 patients in whom MCG data with adequate quality was obtained. The prevalence of positive functional myocardial ischemia and angiographically significant stenosis in the current study was 42.7 % and 41.8 %, respectively. Area under the receiver operating characteristics curve (AUC) of the MCG severity score for functional myocardial ischemia and angiographically significant stenosis was low (AUC 0.51, 95 % confidence interval [CI]: 0.38 to 0.63, and AUC 0.58, 95 % CI: 0.46 to 0.70, respectively). Sensitivity, and specificity of the MCG severity score for functional myocardial ischemia and angiographically significant stenosis was also low (32 %/67 %, and 37 %/72 %) using a cut-off value of 4.0. The authors concluded that diagnostic performance of the MCG severity score was poor for both functional myocardial ischemia, and angiographically significant stenosis.SAECG for Prediction of Adverse Outcomes in Implantable Cardioverter Defibrillator Patients
Chow and associates (2019) stated that current non-invasive risk stratification methods offer limited prediction of arrhythmic events when selecting patients for implantable cardioverter defibrillator (ICD) implantation. The authors’ laboratory has recently developed a signal processing metric called Layered Symbolic Decomposition frequency (LSDf) that quantifies the percentage of hidden QRS wave frequency components in SAECG recordings. In a pilot study, these researchers examined if LSDf can be predictive of ventricular arrhythmia or death in an ICD patient cohort. A total of 52 ICD patients were recruited from 2008 to 2009. These were followed for a mean of 8.5 ± 0.4 years for the primary outcome of first appropriately treated ventricular arrhythmia (VT/VF) or death; 34 subjects met the primary outcome. LSDf was significantly lower, and 12-lead QRS duration was significantly greater in patients meeting the primary outcome (12.14 ± 3.97 % versus 16.45 ± 3.73 %; p = 0.001) and (111.59 ± 14.96 ms versus 97.69 ± 13.51 ms; p = 0.012), respectively. A 13.25 % LSDf threshold (0.74 sensitivity and 0.85 specificity) was selected based on a receiver operating characteristic (ROC) curve. Kaplan-Meier survival analysis was conducted; patients above the 13.25 % threshold demonstrated significantly better survival outcomes (log-rank p < 0.001). In Cox multi-variate regression analysis, the LSDf threshold (13.25 %) was compared to LVEF (28.5 %), 12-lead QRS duration (100 ms), age, % male sex, NYHA classification, and anti-arrhythmic usage. LSDf was a predictor of the primary outcome (p = 0.005) and an independent predictor for solely ventricular arrhythmia (p = 0.002). The authors concluded that Layered Symbolic Decomposition (LSD) is a novel method to perform spectral analysis without basis function selection. They stated that the findings of this pilot study support the notion that Layered Symbolic Decomposition frequency analysis in SAECG recordings may be a viable predictor of negative ICD survival outcomes.
The authors stated that this study had several drawbacks. The limited number of patients in the ICD cohort was highly selected by both LVEF and surface QRS duration using Canadian guidelines for primary prevention and cardiac resynchronization. SAECG is currently not a routine diagnostic procedure in most clinical settings, and performing SAECG testing is not always practical. Accordingly, a future project of the authors’ laboratory is to test LSD analysis in standard 12‐lead ECG data. Furthermore, these researchers were unable to determine the cause of death during patient follow‐up. They would have ideally liked to only incorporate cardiac‐related causes of death in their analyses, but perhaps this can be assessed by future studies.
SAECG for Identification of Recurrent Embolic Stroke
Jung and colleagues (2020) noted that the investigation of the potential association between ischemic stroke and sub-clinical atrial fibrillation (SCAF) is important for secondary prevention. These researchers examined if SCAF can be predicted by atrial substrate measurement with P wave SAECG. They recruited 125 consecutive patients with embolic stroke of undetermined source (ESUS) and 125 patients with paroxysmal AF as controls. All participants underwent P wave SAECG at baseline, and patients with ESUS were followed-up with Holter monitoring and EKC at baseline, 3, 6, and 12 months after discharge and every 6 months thereafter. In the ESUS group, 32 (25.6 %) patients were diagnosed with SCAF during follow-up. There were no significant differences between the groups regarding atrial substrate. P wave duration (PWD) was a significant predictor of SCAF. Stroke recurrence occurred in 22 patients (17.6 %), and prolonged PWD (longer than or equal to 135 ms) predicted stroke recurrence more robustly than SCAF detection. The authors concluded that in ESUS patients, PWD can be a useful biomarker to predict SCAF and to identify patients who are more likely to have a recurrent embolic stroke associated with an atrial cardiopathy. Moreover, these investigators stated that further research is needed for supporting the utility and applicability of PWD.
The authors stated that this prospective analysis had several limitations. First, a main limitation could be the absence of continuous monitoring techniques such as implantable loop recorder (ILR) or telemonitoring systems to detect AF events in the study population. Indeed, these techniques could detect AF recurrence more effectively and avoid the associated negative prognosis with detection failure. Second, this study had relatively short follow-up periods (3 to 12 months) and a small number of patients with a recurrent stroke event. However, these limitations did not negate the ability of PWD to predict both SCAF and stroke recurrence in patients with ESUS. Finally, these researchers defined the clinical AF as episodes of AF that lasted greater longer than 30 seconds. Although this complied with the accepted definition of AF in general, there have been different definitions of AF in many clinical studies. A recent meta-analysis reported that SCAF strongly predicted clinical AF and was associated with elevated absolute stroke risk. Therefore, PWD longer than or equal to 135 ms without SCAF detection may be related to AC with or without AF of shorter duration (shorter than 30 seconds) or of a very rare frequency, thus, explaining the predictive value of PWD for recurrent stroke. Further research is needed to refine the association between PWD and stroke.SAECG for Prediction of the Requirement of Epicardial Ablation in Arrhythmogenic Right Ventricular Cardiomyopathy
Chung and co-workers (2020) aimed to validate the role of SAECG in identifying arrhythmogenic substrates requiring an epicardial approach in arrhythmogenic right ventricular cardiomyopathy (ARVC). A total of 91 patients with a definite diagnosis of ARVC who underwent successful ablation for drug-refractory ventricular arrhythmia were enrolled and classified into 2 groups: group 1 who underwent successful ablation at the endocardium only, and group 2 who underwent successful ablation requiring an additional epicardial approach. The baseline characteristics of patients and SAECG parameters were obtained for analysis. Male predominance, worse right ventricular (RV) function, higher incidence of syncope, and depolarization abnormality were observed in group 2. Moreover, the number of abnormal SAECG criteria was higher in group 2 than in group 1. After a multi-variate analysis, the independent predictors of the requirement of epicardial ablation included the number of abnormal SAECG criteria (OR 2.8, 95 % CI: 1.4 to 5.4; p = 0.003) and presence of syncope (OR 11.7; 95 % CI: 2.7 to 50.4; p = 0.001). Furthermore, greater than or equal to 2 abnormal SAECG criteria were associated with larger RV endocardial unipolar low-voltage zone (p < 0.001), larger RV endocardial/epicardial bipolar low-voltage zone/scar (p < 0.05), and longer RV endocardial/epicardial total activation time (p < 0.001 and p = 0.004, respectively). The authors concluded that the number of abnormal SAECG criteria was correlated with the extent of diseased epicardial substrates and could be a potential surrogate marker for predicting the requirement of epicardial ablation in patients with ARVC.
SAECG for Evaluation of Myocardial Electrophysiological Alterations in Beta-Thalassemia Major Patients
Patsourakos and associates (2020) stated that the majority of beta thalassemia major (β-TM) patients suffer from cardiac disease, while a significant proportion of them die suddenly. The 12-lead and SAECG are simple, inexpensive, readily available tools for identifying an unfavorable arrhythmiological substrate by detecting the presence of arrhythmias, conduction abnormalities, and late potentials (LPs) in these patients. A total of 47 β-TM patients and 30 healthy controls were submitted to 12-lead and SAECG. Basic electrocardiographic parameters and prevalence of LPs were recorded. Basic echocardiographic parameters were estimated by trans-thoracic echocardiography (TTE); T2* was calculated by cardiac magnetic resonance imaging (CMRI) wherever available. β-TM patients demonstrated a more prolonged PR interval (167.74 msec versus 147.07 msec) (p = 0.043), a higher prevalence of PR prolongation (21.05 % versus 0 %) (p = 0.013), and a higher prevalence of LPs (18/47, 38.3 % versus 2/30, 6.7 %) (p = 0.002) compared with controls. The prevalence of AF among b-TM patients was estimated at 10.64 %. Patients had also greater E/e' ratio (8.35, SD = 2.2 versus 7, SD = 2.07) (p =0 .012) and LAVI (30.7 mL/m2, SD = 8.76 versus 24.6 ml/m2, SD = 6.57) (p = 0.002) than controls. Regression analysis showed that QTc and LAVI could correctly predict the presence of LPs in the 80.9 % of the patients. The authors concluded that β-TM patients had a higher prevalence of a prolonged PR interval, AF, and LPs; 12-lead and SAECG performance was feasible in all subjects and constituted a readily available tool for evaluating myocardial electrophysiological alterations in this patient group.
The authors stated that the cross‐sectional nature of the current study was its major drawback. Long‐term observational studies are under way incorporating sophisticated imaging modalities such as novel echocardiographic techniques (speckle tracking imaging) and CMRI. Data from CMRI were available in 38 of 47 patients, preventing these researchers from drawing more reliable conclusions regarding the correlation between T2* and other electrocardiographic and echocardiographic indices. Another drawback was the relatively small number of subjects (n = 47 for β-TM patients). Thalassemia intermedia patients, being a heterogeneous group regarding transfusion dependency, were excluded from the current study; therefore, favoring the extraction of grossly abnormal results.
SAECG for Evaluation of Early Re-Polarization Pattern
Hassanzadeh and colleagues (2021) noted that early repolarization (ER) pattern is diagnosed when the J-point is elevated on the patient's ECG. These researchers examined SAECG in patients with ER pattern. Participants were divided into 3 groups: Patients with normal ECG pattern (control group); patients with J-point elevation in the inferior leads; and patients with J-point elevation in non-inferior leads. The mean filtered QRS duration in groups with J-point elevation in inferior leads and non-inferior leads and in the control, was 86.4 ± 23.4 msec, 84.8 ± 26.6 msec, and 85.8 ± 24.8 msec, respectively, indicating no significant difference across the 3 groups. The mean duration of terminal QRS of less than 40µV was 21.2 ± 4.2 msec, 22.8 ± 4.6 msec, and 23.1 ± 4.5 msec in the mentioned groups, respectively, without a significant difference between the groups. Furthermore, the mean root-mean-square voltage of terminal 40 msec was 34.5 ± 8.3 µV, 35.3 ± 8.6µV, and 35.7 ± 9.2 µV in patients with increased J-point in inferior leads, non-inferior leads, and the control group, respectively, showing no difference among the groups. The authors found that parameters in SAECG did not have any significant difference between patients with ER pattern and healthy individuals. Moreover, these investigators stated that SAECG could not distinguish patients with elevated J-point in inferior leads from non-inferior leads. Overall, SAECG did not appear to be a reliable diagnostic tool for the assessment of ER pattern.
Assistive Artificial Intelligence Algorithmic Electrocardiogram Risk-Based Assessment (MyoVista Wavelet ECG Cardiac Testing Device) for Evaluation of Cardiac Dysfunction
On July 7, 2022, Anumana, Inc. (Cambridge, MA), an artificial intelligence (AI)-driven health technology company announced that the American Medical Association (AMA) has issued new industry-first Category III CPT codes for novel assistive AI algorithmic electrocardiogram (ECG) risk assessment for cardiac dysfunction. These codes are expected to cover the AI cardiac dysfunction algorithm incorporated in the Heart Test Laboratories, Inc. (Southlake, TX)’s MyoVista Wavelet ECG Cardiac Testing Device currently in development for FDA De Novo submission and clearance. The codes are in addition to the long-established CPT codes in place for the conventional ECG. However, there is currently insufficient evidence to support the use of AI algorithmic electrocardiogram for detection of cardiovascular-related diseases.
Lee et al (2022) noted that several AI models for the detection and prediction of cardiovascular-related diseases, including arrhythmias, diabetes mellitus (DM), and sleep apnea, have been reported. In a systematic review and meta-analysis, these investigators identified AI models developed for or applicable to wearable and mobile devices for diverse cardiovascular (CV)-related diseases. The searched databases included Medline, Embase, and Cochrane Library. For AI models for atrial fibrillation (AF) detection, a meta-analysis of diagnostic accuracy was carried out to summarize sensitivity and specificity. A total of 102 studies were included in the qualitative review. There were AI models for the detection of arrythmia (n = 62), followed by sleep apnea (n = 11), peripheral vascular diseases (PVD; n = 6), DM (n = 5), hypertension/hypotension (n = 5), valvular heart disease (n = 4), heart failure (HF; n = 3), myocardial infarction (MI) and cardiac arrest (n = 2), and others (n = 4). For quantitative analysis of 26 studies reporting AI models for AF detection, meta-analyzed sensitivity was 94.80 % and specificity was 96.96 %. Deep neural networks showed superior performance (meta-analyzed area under receiver operating characteristics curve (AUROC) of 0.981) compared to conventional machine learning algorithms (meta-analyzed AUROC of 0.961). However, AI models tested with proprietary dataset (meta-analyzed AUROC of 0.972) or data acquired from wearable devices (meta-analyzed AUROC of 0.977) showed inferior performance than those with public dataset (meta-analyzed AUROC of 0.986) or data from in-hospital devices (meta-analyzed AUROC of 0.983). The authors concluded that this systematic review and meta-analysis demonstrated that AI models for the diagnosis and prediction of various CV-related diseases as well as arrhythmias are being developed, and that they are gradually developing into a form that is suitable for wearable and mobile devices. Many studies have shown that the deep learning algorithm showed very high performance compared to the existing analysis methods that use human visualization or the extraction of hand-made features for bio-signals, such as ECG or photo-plethysmography (PPG) signals. However, there must still be sufficient consideration of various aspects, such as the data acquisition process, characteristics of the acquired data, characteristics of the population to which the algorithm is applied, weight reduction of the algorithm, working principle, and interpretability of the model, to develop a practical medical AI model that can be used in the real world.
In a systematic review, Jemioło et al (2022) examined the current state and quality of publicly available datasets used for automated affect and emotion recognition (AAER) with AI, emphasizing CV signals. The quality of such datasets is essential to create replicable systems for future work to grow. These investigators reviewed 9 sources up to August 31, 2020, using a developed search strategy, including studies considering the use of AI in AAER based on CV signals. Two independent reviewers carried out the screening of identified records, full-text assessment, data extraction, and credibility. All discrepancies were resolved by discussion. These researchers descriptively synthesized the results and examined their credibility. The protocol was registered on the Open Science Framework (OSF) platform. A total of 18 records out of 195 were selected from 4,649 records, focusing on datasets containing CV signals for AAER. Included papers analyzed and shared data of 812 subjects aged 17 to 47; ECG was the most examined signal (83.33 % of datasets); and researchers used video stimulation most frequently (52.38 % of experiments). Despite these results, much information was not reported by researchers. The authors concluded that the quality of the analyzed studies was primarily low; they stated that investigators in the field should concentrate more on methodology. These researchers stated that in the future, more attention should be put into controlling bias in research to ensure incremental knowledge gain. The quality of studies and reporting needs to be improved in order to propose and develop models that do not introduce biases. Preferably, researchers should focus more on methodology and describe procedures thoroughly. The authors recommended following standardized guidelines of reporting.
Rafie et al (2022) stated that the medical complexity and high acuity of patients in the cardiac intensive care unit (CICU) make for a unique patient population with high morbidity and mortality. While there are many tools for predictions of mortality in other settings, there is a lack of robust mortality prediction tools for CICU patients. The ongoing advances in AI and machine learning also pose a potential asset to the advancement of mortality prediction. AI algorithms have been developed for use of ECG interpretation with promising accuracy and clinical application. Furthermore, AI algorithms applied to ECG interpretation have been developed to predict various variables such as structural heart disease, left ventricular systolic dysfunction (LVSD), and atrial fibrillation (AF). These variables can be used and applied to new mortality prediction models that are dynamic with the changes in the patient's clinical course and may lead to more accurate and reliable mortality prediction. The authors concluded that the ability to identify occult disease processes and predict patient outcomes is important in the care of CICU patients. This capability has clinical benefits that include early mortality prediction, identification of patients who may benefit from invasive testing or intervention, and identification of patient who may not need ICU-level of care. The interest in predicting outcomes in ICU settings is exemplified by the recent proliferation of prediction tools. Moreover, these investigators noted that although many of these tools are valuable, many lack validation in the CICU population and there is considerable variability in their performance. These researchers stated that randomized controlled trials (RCTs) are needed to develop AI-ECG models to evaluate these models prospectively and compare them against currently implemented risk scores.
Bjerken et al (2022) noted that screening for LVSD, defined as reduced left ventricular ejection fraction (LVEF), deserves renewed interest as the medical treatment for the prevention and progression of HF improves. In a systematic review, these investigators updated literature to examine the potential and caveats of using AI-enabled ECG (AIeECG) as an opportunistic screening tool for LVSD. They searched PubMed and Cochrane for variations of the terms "ECG", "heart failure", "systolic dysfunction", and "artificial intelligence" from January 2010 to April 2022 and selected studies that reported the diagnostic accuracy and confounders of using AIeECG to detect LVSD. Out of 40 studies, these researchers identified 15 relevant studies -- 11 retrospective cohorts, 3 prospective cohorts, and 1 case series. Although various LVEF thresholds were employed, AIeECG detected LVSD with a median AUC of 0.90 (inter-quartile range [IQR] from 0.85 to 0.95), a sensitivity of 83.3 % (IQR from 73 % to 86.9 %) and a specificity of 87 % (IQR from 84.5 % to 90.9 %). AIeECG algorithms succeeded across a wide range of sex, age, and co-morbidity and appeared particularly useful in non-cardiology settings and when combined with natriuretic peptide testing. In addition, a false-positive AIeECG indicated a future development of LVSD. No studies examined the effect on treatment or patient outcomes. The authors concluded that this systematic review corroborated the arrival of a new generic biomarker, AIeECG, to improve the detection of LVSD. These investigators stated that AIeECG, in addition to natriuretic peptides and ECG, will improve screening for LVSD; however, prospective, randomized trials especially in the primary sector, are needed to show cost-effectiveness and clinical significance, preferably in combination with other biomarkers such as the natriuretic peptides.
The authors stated that this review had several drawbacks. These investigators searched PubMed and Cochrane databases and only included peer-reviewed studies of high quality to focus on clinical aspects rather than technical differences between the algorithms. They mainly focused on LVSD and a 12-lead ECG that could have led to exclusion of otherwise relevant studies that examined AIeECG based on 1- or 3-lead ECG. In addition, although it was a drawback that these researchers did not employ a formal “risk-of-bias assessment tool,” they aimed to minimize the risk of bias by using strict study selection criteria. Furthermore, these researchers noted that due to few studies and lack of statistical power, they summarized data instead of making formal statistical tests. These investigators stated that many more studies and algorithms will without doubt evolve over the coming years, allowing for more accurate estimates of accuracy.
Chung et al (2022) stated that CV diseases are one of the leading global causes of mortality. To-date, clinicians rely on their own analyses or automated analyses of the ECG to obtain a diagnosis. However, both approaches can only include a finite number of predictors and are unable to execute complex analyses. AI has enabled the introduction of machine-learning (ML) and deep-learning (DL) algorithms to compensate for the existing limitations of current ECG analysis methods, with promising results. However, it should be prudent to recognize that these AI-based algorithms are also associated with their own unique set of challenges and limitations, such as professional liability, systematic bias, surveillance, cyber-security, as well as technical and logistical challenges. The objective of this review was to increase familiarity with; and awareness of AI-based algorithms used in ECG diagnosis, and to inform the interested stakeholders on their potential use in addressing present clinical challenges. The authors concluded that AI-based analysis of ECG signals is expected to revolutionize healthcare diagnostic and prognostic services, in both cardiology and non-cardiology-related diseases. To ensure AI can safely enhance the quality of healthcare services, a framework that regulates the implementation of AI is mandatory. Scientists must be cognizant of the existing limitations of AI-based analysis of ECGs, such as potential bias and actively design interventions to safeguard undesirable outcomes. With such measures, AI can be implemented on a global scale for clinical practice. Ultimately, AI has the potential to offer data-driven clinical decision support systems.
Huang et al (2022) noted that CV disease (CVD) is the world's leading cause of mortality. There is significant interest in using AI to analyze data from novel sensors such as wearables to provide an earlier and more accurate prediction and diagnosis of heart disease. Digital health technologies that fuse AI and sensing devices may help disease prevention and reduce the substantial morbidity and mortality caused by CVD worldwide. In this review, these investigators described recent developments in the use of digital health for CVD, focusing on AI approaches for CVD detection, diagnosis, and prediction via AI models driven by data collected from wearables. They examined the available evidence on the use of wearables and AI in CVD diagnosis, followed by a detailed description of the dominant AI approaches used for modelling and prediction using data acquired from sensors such as wearables. The authors discussed the AI-based algorithms and models and clinical applications and found that AI and ML-based approaches are superior to traditional or conventional statistical methods for predicting CV events; however, further studies examining the use of such algorithms in the real world are needed. Additionally, improvements in wearable device data accuracy and better management of their application are needed.
References
The above policy is based on the following references:
- Bae MH, Kim JH, Jang SY, et al. Changes in follow-up ECG and signal-averaged ECG in patients with arrhythmogenic right ventricular cardiomyopathy. Pacing Clin Electrophysiol. 2014;37(4):430-438.
- Bennhagen RG, Sornmo L, Pahlm O, Pesonen E. Serial signal-averaged electrocardiography in children after cardiac transplantation. Pediatr Transplant. 2005;9(6):773-779.
- Bjerken LV, Ronborg SN, Jensen MT, et al. Artificial intelligence enabled ECG screening for left ventricular systolic dysfunction: A systematic review. Heart Fail Rev. 2022 Nov 8 [Online ahead of print].
- Brignole M, Alboni P, Benditt DG, et al. Guidelines on management (diagnosis and treatment) of syncope--update 2004. Europace. 2004;6(6):467-537.
- Cain ME, Anderson JL, Arnsdorf MF, et al. Signal-averaged electrocardiography. ACC Expert Consensus Document. JACC J Am Col Cardiol. 1996;27(1):238-249.
- Chow R, Hashemi J, Torbey S, et al. Novel frequency analysis of signal-averaged electrocardiograms is predictive of adverse outcomes in implantable cardioverter defibrillator patients. Ann Noninvasive Electrocardiol. 2019;24(3):e12629.
- Chung CT, Lee S., King E, et al. Clinical significance, challenges and limitations in using artificial intelligence for electrocardiography-based diagnosis. Int J Arrhythm. 2022;23:24.
- Chung F-P, Lin C-Y, Lin Y-J, et al. Application of noninvasive signal-averaged electrocardiogram analysis in predicting the requirement of epicardial ablation in patients with arrhythmogenic right ventricular cardiomyopathy. Heart Rhythm. 2020;17(4):584-591.
- Coeytaux RR, Leisy PJ, Wagner GS, et al. Systematic review of ECG-based signal analysis technologies for evaluating patients with acute coronary syndrome. Agency for Healthcare Research and Quality's (AHRQ). Rockville, MD; AHRQ; June 2012.
- Coeytaux RR, Williams JW, Chung E, Gharacholou M. ECG-based signal analysis technologies. Technology Assessment. Prepared for the Agency for Healthcare Research and Quality (AHRQ) by the Duke Evidence-based Practice Center (Contract No. HHSA 290-2007-10066I). Rockville, MD: AHRQ; May 24, 2010.
- Dinov B, Bode K, Koenig S, et al. Signal-averaged electrocardiography as a noninvasive tool for evaluating the outcomes after radiofrequency catheter ablation of ventricular tachycardia in patients with ischemic heart disease: Reassessment of an old tool. Circ Arrhythm Electrophysiol. 2016;9(9).
- Engel G, Beckerman JG, Froelicher VF, et al. Electrocardiographic arrhythmia risk testing. Curr Probl Cardiol. 2004;29(7):365-432.
- Gatzoulis KA, Arsenos P, Trachanas K, et al. Signal-averaged electrocardiography: Past, present, and future. J Arrhythm. 2018 34(3):222-229.
- Gregoratos G, Cheitlin MD, Conill A, et al. ACC/AHA guidelines for implantation of cardiac pacemakers and antiarrhythmia devices: A report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Committee on Pacemaker Implantation). J Am Coll Cardiol. 1998;31(5):1175-1209.
- Grimm W, Christ M, Bach J, et al. Noninvasive arrhythmia risk stratification in idiopathic dilated cardiomyopathy: Results of the Marburg Cardiomyopathy Study. Circulation. 2003;108(23):2883-2891.
- Grube E, Bootsveld A, Buellesfeld L, et al. Computerized two-lead resting ECG analysis for the detection of coronary artery stenosis after coronary revascularization. Int J Med Sci. 2008;5(2):50-61.
- Haghjoo M, Arya A, Parsaie M, et al. Does the abnormal signal-averaged electrocardiogram predict future appropriate therapy in patients with implantable cardioverter-defibrillators? J Electrocardiol. 2006;39(2):150-155.
- Hassanzadeh M, Mardani E, Hosseinpour A, et al. Signal averaged ECG in patients with early repolarization. J Arrhythm. 2021;37(2):432-437.
- Horenstein MS, Idriss SF, Hamilton RM, et al. Efficacy of signal-averaged electrocardiography in the young orthotopic heart transplant patient to detect allograft rejection. Pediatr Cardiol. 2006;27(5):589-593.
- Huang J-D, Wang J, Ramsey E, et al. Applying artificial intelligence to wearable sensor data to diagnose and predict cardiovascular disease: A review. Sensors (Basel). 2022;22(20):8002.
- Hunt SA, Baker DW, Chin MH, et al. ACC/AHA guidelines for the evaluation and management of chronic heart failure in the adult. Bethesda, MD: American College of Cardiology Foundation (ACCF); September 2001.
- Husser D, Stridh M, Sornmo L, et al. Analysis of the surface electrocardiogram for monitoring and predicting antiarrhythmic drug effects in atrial fibrillation. Cardiovasc Drugs Ther. 2004;18(5):377-386.
- Jaroszynski AJ, Glowniak A, Sodolski T, et al. Effect of haemodialysis on signal-averaged electrocardiogram P-wave parameters. Nephrol Dial Transplant. 2006;21(2):425-430.
- Jemioło P, Storman D, Mamica M, et al. Datasets for automated affect and emotion recognition from cardiovascular signals using artificial intelligence -- A systematic review. Sensors (Basel). 2022;22(7):2538.
- Jung M, Kim J-S, Song JH, et al. Usefulness of P wave duration in embolic stroke of undetermined source. J Clin Med. 2020;9(4):1134.
- Kawaji T, Shiomi H, Morimoto T, et al. Noninvasive detection of functional myocardial ischemia: Multifunction cardiogram evaluation in diagnosis of functional coronary ischemia study (MED-FIT). Ann Noninvasive Electrocardiol. 2015;20(5):446-453.
- Lee S, Chu Y, Ryu J, et al. Artificial intelligence for detection of cardiovascular-related diseases from wearable devices: A systematic review and meta-analysis. Yonsei Med J. 2022;63(Suppl):S93-S107.
- Liao YC, Chung FP, Lin YJ, et al. The application of signal average ECG in the prediction of recurrences after catheter ablation of ventricular arrhythmias in arrhythmogenic right ventricular dysplasia/cardiomyopathy. Int J Cardiol. 2017;236:168-173.
- Liao YC, Lin YJ, Chung FP, et al. Risk stratification of arrhythmogenic right ventricular cardiomyopathy based on signal averaged electrocardiograms. Int J Cardiol. 2014;174(3):628-633.
- Nagamoto Y, Fujii Y, Morita Y, et al. Atrial electrical abnormality in patients with Brugada syndrome assessed by signal-averaged electrocardiography. Indian Heart J. 2017;69(6):714-719.
- Narayan SM, Cain ME. Clinical applications of the signal-averaged electrocardiogram: Overview. UpToDate [online serial]. Waltham, MA: UpToDate; reviewed May 2015.
- Narayan SM, Cain ME. Use of the signal-averaged electrocardiogram in nonischemic heart disease and cardiac transplantation. UpToDate [online serial]. Waltham, MA: UpToDate; reviewed May 2015.
- Omeroglu RE, Olgar S, Nisli K. Signal-averaged electrocardiogram may be a beneficial prognostic procedure in the postoperative follow-up tetralogy of fallot patients to determine the risk of ventricular arrhythmias. Pediatr Cardiol. 2007;28(3):208-212.
- Park Y, Cho Y, Lee DY, et al. Correlation between the parameters of signal-averaged ECG and two-dimensional echocardiography in patients with arrhythmogenic right ventricular cardiomyopathy. Ann Noninvasive Electrocardiol. 2009;14(1):50-56.
- Patsourakos D, Gatzoulis KA, Aggeli C, et al. Twelve-lead and signal-averaged electrocardiographic parameters among beta-thalassemia major patients. J Arrhythm. 2020;36(5):920-928.
- Priori SG, Aliot E, Blomstrom-Lundqvist C, et al. Task Force on Sudden Cardiac Death of the European Society of Cardiology. Eur Heart J. 2001;22(16):1374-450.
- Proclemer A, Lewalter T, Bongiorni MG, et al. Screening and risk evaluation for sudden cardiac death in ischaemic and non-ischaemic cardiomyopathy: Results of the European Heart Rhythm Association survey. Europace. 2013;15(7):1059-1062.
- Rafie N, Jentzer JC, Noseworthy PA, Kashou AH. Mortality prediction in cardiac intensive care unit patients: A systematic review of existing and artificial intelligence augmented approaches. Front Artif Intell. 2022;5:876007.
- Ryan TJ, Antman EM, Brooks NH, et al. 1999 update: ACC/AHA guidelines for the management of patients with acute myocardial infarction. A report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines. J Am Coll Cardiol. 1999;34(3):890-911.
- Schoenenberger AW, Erne P, Ammann S, et al. Prediction of arrhythmic events after myocardial infarction based on signal-averaged electrocardiogram and ejection fraction. Pacing Clin Electrophysiol. 2008;31(2):221-228.
- Strobeck JE, Shen JT, Singh B, et al. Comparison of a two-lead, computerized, resting ECG signal analysis device, the MultiFunction-CardioGram or MCG (a.k.a. 3DMP), to quantitative coronary angiography for the detection of relevant coronary artery stenosis (>70%) - a meta-analysis of all published trials performed and analyzed in the US. Int J Med Sci. 2009;6(4):143-155.
- Tamaki S, Yamada T, Okuyama Y, et al. Cardiac iodine-123 metaiodobenzylguanidine imaging predicts sudden cardiac death independently of left ventricular ejection fraction in patients with chronic heart failure and left ventricular systolic dysfunction: Results from a comparative study with signal-averaged electrocardiogram, heart rate variability, and QT dispersion. J Am Coll Cardiol. 2009;53(5):426-435.
- U.S. Department of Health and Human Services, Public Health Service, Agency for Healthcare Policy and Research (AHCPR). Signal-averaged electrocardiography. Health Technology Assessment No. 11. AHCPR Pub. No. 98-0020. Rockville, MD: AHCPR; May 1998.
- U.S. Food and Drug Administration (FDA), Center for Devices and Radiologic Health (CDRH). Cardiotron multi-phase EKG information analysis system. 510(k) no. K992703. Rockville, MD: FDA; March 21, 2000.
- Urbanova D, Urban L, Mikuskova E, et al. Frequency-domain analysis of the signal-averaged electrocardiogram in hematological malignancies survivors. Bratisl Lek Listy. 2010;111(3):144-149.
- Weiss MB, Narasimhadevara SM, Feng GQ, Shen JT. Computer-enhanced frequency-domain and 12-lead electrocardiography accurately detect abnormalities consistent with obstructive and nonobstructive coronary artery disease. Heart Dis. 2002;4(1):2-12.