Risultati per: Passo 12. Ripulire in modo mirato le cartelle cliniche degli assisiti

Circulation, Volume 148, Issue Suppl_1, Page A15117-A15117, November 6, 2023. Introduction:The 2022 AHA Guidelines for the Management of Heart Failure (HF) emphasized initiating guideline-directed medical therapy (GDMT) as early as possible during acute hospitalization. GDMT significantly differs between HF patients with reduced and preserved ejection fraction (EF). The electrocardiogram (ECG) is one of the first tests performed to assess cardiac function among HF patients.Purpose:To develop a linear model to estimate EF using the ECG and clinical covariates.Methods:Medical record data from 2020-21 of HF patients >18 years and positive Framingham Heart Failure Diagnostic Criteria excluding those with ventricular assist devices were extracted. Demographics, NYHA class, EF, and comorbidities, 12-lead ECG closest to the time of the echocardiogram were analyzed. We extracted 312 different features from the ECG using the Philips DXL Algorithm. Recursive LASSO regression was used to identify a subset of features in domain blocks. All analyses were completed in RStudio (4.3.0).Results:Among 126 patients (age 69.4+15.5 years; BMI 31.2+8.2 kg/m2; EF 40+19%; the time between ECG-echocardiogram 13.30+40.67 hours: minutes), the predominant cardiac rhythm (70%, n=88) was sinus. Among the 312 ECG features, 39 (13%) were significant; the final linear model included 9 (3%) features: S Duration V1, QRS Duration III, Horizontal T Angle, Mean Heart Rate, Q Duration aVL, Maximum QRS Angle, R Duration V3, S Amplitude V3, and P amplitude V3. The final linear model achieved an R2=0.498 and adjusted R2=0.459. We adjusted the model for age, body mass index, and sex, and model performance improved by R2=0.107 adjusted R2=0.104. The addition of time between ECG-echocardiogram to the model did not significantly change performance. From the predicted values, 50% (n=63) were within the margin of error, and the correlation between predicted and true values was moderate (r=0.629 [95%CI, 0.510-0.724]). The mean difference between predicted and true EF values varied between+12.06%.Conclusions:Our initial results demonstrate the ability to estimate EF using a 12-lead ECG. While the results are limited due to the small sample size and significant time between ECG and echocardiogram, these preliminary results support the feasibility of future research.

Circulation, Volume 148, Issue Suppl_1, Page A16473-A16473, November 6, 2023. Introduction:Artificial Intelligence-based 5-lead 3D-vectorcardiography (5L3DVCG-AI) offers additional information over 12-lead electrocardiography (ECG) in the detection of significant coronary stenoses. 5L3DVCG-AI is under investigation as a new screening tool for coronary vascular disease (CVD). Hypothesis: We tested the hypothesis of variables from the reconstructed “12-lead ECG” (5L12L-ECG, modified Dower transformation) corresponding with the standard 12-lead ECG (ECG).Methods:In this monocentric exploratory retrospective study, raw data of 331 patients with 5L3DVCG-AI and ECG were included. Cardiac pathology (CP) was categorised as exclusion of any CP (control), mild CP or overt CP by 2 independent cardiologists. The following variables were compared: RR-interval, P, PQ, QRS, QT, QTcB, QTcF, QRS-morphology, and ST-morphology. Cardiovascular risk factors (CVRF) were quantified with the modified PROCAM score.Results:From 331 patients (m:w 60:40%, 50.0 ± 19.8 years) of mixed ethnicity and moderate CVRF (2.1 ± 1.2), 70% were controls, 21% had mild CP and 9% overt CP. All variables from reconstructed 5L12L-ECG correlated to the corresponding individual variables from ECG (r= 0.49 to 0.7, p

Circulation, Volume 148, Issue Suppl_1, Page A18177-A18177, November 6, 2023. Background:In patients undergoing cardiac surgery who have ventricular arrhythmias (VAs) such as VT, VF, or PVCs, concurrent surgical ablation is an attractive therapeutic strategy. However, electrophysiologic mapping systems are not routinely available in the OR and it may be difficult to localize the VA origin. We developed a workflow incorporating novel 12-lead ECG computational model-based mapping to localize VAs.Hypothesis:We hypothesized that use of pre-operative computational ECG mapping can help guide surgical ablation of VAs in patients undergoing cardiac surgery.Methods:Patients undergoing cardiac surgery with pre-existing VT, VF or PVCs were enrolled with informed consent. A standard 12-lead ECG of the VA was recorded in the clinic or during non-invasive programmed stimulation in the EP lab. ECG mapping localized the VA and visualized on a 3D model. During circulatory arrest, surgical ablation was performed using either cryoablation or irrigated RF ablation probe. Follow-up was performed with event monitors or ICD monitoring.Results:A total of 7 patients (mean age 55±15 years, female 29%, EF 31%±20%) were enrolled (Table). Surgical indications included mitral annuloplasty (Pt 1), pulmonary thromboendoarterectomy + CABG (Pt 2), coronary artery unroofing (Pt 3), left ventricular assist device (Pt 4-6, Fig 1), and CABG (Pt 7). In this cohort, 5 PVCs, 3 monomorphic VT and 1 VF morphologies were localized using ECG mapping and surgically ablated at time of cardiac surgery. There was a 100% (18 to 0) decrease in VT/VF episodes and 97.2 ± 0.03% reduction in PVC burden at median 9.5 month (IQR 3.7-29.5) follow-up. No intra- or post-operative complications occurred.Conclusions:This case series illustrates feasibility and excellent efficacy of a novel preoperative ECG mapping workflow using a forward-solution algorithm to guide successful and safe concomitant ventricular arrhythmia ablation during cardiac surgery.

Circulation, Volume 148, Issue Suppl_1, Page A13412-A13412, November 6, 2023. Introduction:Multi-vessel coronary vasospastic angina (multi-VSA) is life-threatening disease. We tried to predict multi-VSA by machine learning (ML) on 12-lead electrocardiogram (ECG).Hypothesis:Machine learning on 12-lead ECG has powerful diagnostic value for multi-VSA.Methods:We recruited 227 consecutive sinus-rhythm patients (63.6±12.9years, 136men) who underwent acetylcholine-provocation test in coronary angiography (CAG). Multi-VSA was defined as spasm in at least 2 major branches. ECG was recorded before CAG in no chest pain period. ML was performed on table data of ECG parameters using several ensemble learning methods.Results:79 patients (35%) showed multi-VSA, and univariate logistic regression analysis extracted 23 significant but weak predictors, the highest area under receiver operating characteristics curve (AUROC) was 0.673. Conversely, ML demonstrated high diagnostic performance (AUROC of extra trees classifier: 0.817). Shapley additive explanation method showed male, QTc, J wave in lead II, and low amplitude of Q wave in lead I/aVL played essential roles to build the ML model.Conclusion:Several parameters of 12-lead ECG in multi-VSA patients contains potential features of VSA, and their aggregation and ensemble learning can predict VSA with high diagnostic performance.

Circulation, Volume 148, Issue Suppl_1, Page A12785-A12785, November 6, 2023. Introduction:The ECG has always been evaluated as static risk factor for sudden cardiac death (SCD). The importance of dynamic ECG remodeling has not been investigated.Hypothesis:Abnormal ECG remodeling over time is associated with increased risk of SCD.Methods:We investigated pre-SCD ECG remodeling in SCD cases from 2 ongoing population-based studies of out-of-hospital SCD in Portland, OR (discovery) and Ventura County, CA (validation). Two archived pre-SCD ECGs performed at least 1 year apart were obtained from lifetime health records. Controls were matched on geographical region, age, sex, and duration between the 2 ECG recordings. Dynamic ECG remodeling was measured as the change in a previously validated cumulative 6-variable ECG electrical risk score (ERS) between the 1st and 2nd ECG.Results:A total of 231 SCD cases (66.5±13.6 years, 61% male), and 234 controls (65.8±11.1 years, 61% male) were included in the discovery cohort, and 203 SCD cases (70.3±14.4 years, 54% male), and 203 controls (68.4±11.8 years, 54% male) in the validation cohort. The mean time between the 2 ECG recordings in SCD cases and controls was 6.0±4.0 years vs 6.2±4.5 years (discovery) and 3.7±2.6 years vs. 3.7±1.6 years (validation), respectively. In both cohorts, SCD cases compared to controls had greater dynamic ECG remodeling over time: Discovery cohort ERS change +1.06 (95% CI +0.89 to +1.24) vs. -0.05 (-0.21 to +0.11; p

Circulation, Volume 148, Issue Suppl_1, Page A13921-A13921, November 6, 2023. Introduction:Increased left ventricular myocardial mass (LVM) is associated with adverse cardiovascular outcomes. Various rule-based criteria using limited electrocardiogram (ECG) features lacks sensitivity for evaluating LVM. Recent studies using deep learning methods have made progress in LVM evaluation taking advantage of ECG amplitude data.Hypothesis:We hypothesized that LVM prediction with deep learning models are improved by adding inter-lead information and building sex-specific models.Methods:This study proposed a novel deep learning-based method, the eLVMass-Net, by using ECG-LVM (n=1,459) paired data. ECG signals, QRS duration and axis, demographic features were used as input data. ECG signals were encoded by a temporal convolutional network (TCN) encoder. We adopted a total of four models (non-lead-grouping v. lead-grouping, non-sex-specific v. sex-specific) for LVM estimation. For lead-grouping models, the encoders were constructed based on segregated limb and precordial leads. For sex-specific models, we constructed models on both genders separately. Encoded ECG features and demographic features were concatenated for LVM prediction. To evaluate the performance, we utilized a 5-fold cross-validation approach with the evaluation metrics, mean absolute error (MAE) and mean absolute percentage error (MAPE).Results:For non-sex-specific models, eLVMass-Net has achieved an MAE of 14.33±0.71 and an MAPE of 12.90%±1.12%, outperforming the best state-of-the-art method (MAE 19.51±0.82; MAPE 17.62%±0.78%; P < 0.01). Sex-specific models achieved even lower MAPE for both males and females respectively (male MAPE 12.55%±0.88%; female MAPE 12.52%±0.34%), which also surpassed state-of-the-art methods. Adding the information of QRS axis and duration did not significantly improve the model performance (P = 0.28). The saliency map showed that T wave in precordial leads and QRS complex in limb leads are important features with increasing LVM.Conclusions:This study proposed a novel LVM estimation method, outperforming previous methods by emphasizing relevant heartbeat waveforms, inter-lead information, and non-ECG demographic features. The sex-specific analysis is crucial in improving LVM prediction.

Circulation, Volume 148, Issue Suppl_1, Page A12591-A12591, November 6, 2023. Introduction:Acute myocardial infarction (AMI) is the most prevalent cause of mortality and morbidity in patients with type 2 diabetes mellitus (T2DM). However, strict blood glucose control does not always prevent the development and progression of AMI.Aims:Therefore, the present study aimed to explore potential new biomarkers associated with the occurrence of AMI in T2DM patients.Methods:A total of 82 participants were recruited, including the control group (n=28), T2DM without AMI group (T2DM, n=30) and T2DM with initial AMI group (T2DM+AMI, n=24). The untargeted metabolomics using LC-MS analysis was performed to evaluate the changes in serum metabolites. Then, candidate metabolites were determined using ELISA method in the validation study (n=126/T2DM group, n=122/T2DM+AMI group).Results:The results showed that 146 differential serum metabolites were identified among the control, T2DM and T2DM+AMI, Moreover, 16 differentially-expressed metabolites were significantly altered in T2DM+AMI compared to T2DM. Furthermore, three candidate differential metabolites, 12,13-diHOME, noradrenaline (NE) and estrone sulfate (ES), were selected for validation study. Serum levels of 12,13-diHOME and NE in T2DM+AMI were significantly higher than those in T2DM. Multivariate logistic analyses showed that 12,13-diHOME (OR, 1.491; 95% CI, 1.230-1.807,P

Circulation, Volume 148, Issue Suppl_1, Page A17454-A17454, November 6, 2023. Introduction:Over 100 FDA-approved medications, electrolyte perturbations, and many disease states can prolong the QT interval in up to 10% of patients. In contrast, approximately 1 in 2,000 people have congenital long QT syndrome (LQTS) hallmarked by pathological QT prolongation secondary to LQTS-causative mutations.Hypothesis:an artificial intelligence (AI) deep neural network (DNN) analysis of the 12-lead ECG can distinguish patients with LQTS from those with acquired QT prolongation.Methods:The study cohort included 1599 patients with genetically confirmed LQTS and over 2.5 million controls from Mayo Clinic’s ECG data vault. Every patient/control with ≥ 1 ECG above age- and sex- specific 99thpercentile values for QTc [ > 460 ms for all patients (male/female) < 13 years of age, or > 470 ms for men and > 480 ms for women above this age] was included. An AI-DNN involving a multi-layer convolutional neural network was developed. To simulate screening conditions, patients were matched at a ratio of 1:2,000 (incidence of LQTS) or 1:200 (balance of LQTS vs acquired QT prolongation in a tertiary referral center).Results:Among the 1,599 patients with LQTS, 808 ( > 50%) had ≥ 1 ECG with QTc above the aforementioned QTc thresholds (2,987 ECGs) compared to 361,069/2.5M controls (14% of Mayo Clinic patients getting an ECG, 989,313 ECGs). Following age- and sex- matching and splitting, the model successfully distinguished LQTS from those with acquired QT prolongation at 1:2,000 matching with an AUC of 0.937 (accuracy 89%, sensitivity 81%, specificity 90%, PPV 0.05, NPV 0.99). Furthermore, when matching at a rate that genetically-mediated QT prolongation would be encountered in clinic (1:200), the model still successfully separated the two groups (AUC 0.912, accuracy 88%, sensitivity 78%, specificity 89%, PPV 0.1, NPV 0.99).Conclusion:For patients with a QTc exceeding its 99thpercentile values seen in health, this novel AI-DNN 12-lead ECG distinguishes abnormal QT prolongation stemming from LQTS versus acquired QT prolongation with high performance characteristics (AUC > 0.93). Even when scaled to referral center and LQTS incidence ratios, a negative AI-DNN signal rules out the presence of a LQTS disease-causative mutation with 99% negative predictive value.

Circulation, Volume 148, Issue Suppl_1, Page A15690-A15690, November 6, 2023. Introduction:Frailty is a syndrome of functional decline characterized by an increased risk for adverse health outcomes. The association between frailty and left atrial appendage closure (LAAC) outcomes has not been extensively studied yet. This study aims to analyze the impact of frailty on adverse health outcomes following LAAC.Method:Retrospective review of electronic medical records from June 2016 to December 2021 at the University of Illinois, Chicago identified LAAC patients, who were included if they were followed-up within 6 and 12 months of the procedure. Patients were stratified into frail and non-frail groups based on the Johns Hopkins Claims-based Frailty Indicator, an externally validated index. Outcomes included 6- and 12-months major bleeding event defined by VARC, death, and hospitalization indexes. Two-sample t-tests and chi-squared tests were used to compare continuous and categorical variables, respectively.Results:Our cohort included 101 patients (age 72 ± 9 years, 75% male, 24% white): frail (N=34) and non-frail (N=67) groups. There were no statistically significant differences in baseline demographics, comorbidities, and medication except for older age in frail patients (79.6 vs 68.3 years, p=0.000). Although frail patients had higher mortality rate at 6 months (8.82% vs 0%, p=0.014), no differences were seen for nonhome discharges (6.67% vs 1.54%, p=0.184), 6-months major bleeding events (2.94% vs 2.99%, p=0.990), 6-month emergency department visits (38.2% vs 28.4%, p=0.313), and 6-month hospital admission (20.6% vs 25.4%, p=0.593). There were also no differences seen for mortality rate at 12 months (11.8% vs 5.97%, p=0.308), length of hospital stays (1.35 vs 1.37 days, p=0.966), 12-months major bleeding event (2.94% vs 4.48%, p=0.708), and 12-months hospital admission (29.4% vs 31.3%, p=0.842).Conclusion:Our results suggest that frail patients are at a higher risk for death within 6 months following LAAC despite no observable differences in comorbidities, medications, other adverse events, and hospitalization indexes, including nonhome discharges. Additional analysis is needed to determine factors that may preclude frail patients from hospital admissions or precipitate earlier death.

Circulation, Volume 148, Issue Suppl_1, Page A18141-A18141, November 6, 2023. Background:Unstable ventricular tachycardia is difficult to map and ablate, with high recurrence rates. We developed a 12-lead ECG mapping algorithm based on computational models to localize VT to help guide ablation of unstable VT.Hypothesis:We hypothesized that prospective ECG-mapping can facilitate invasive activation mapping and improve the success of unstable VT ablation.Aims:To compare the time to ICD shock + death in patients undergoing ECG-mapping guided ablation of unstable VT compared to standard ablation controls.Methods:Consecutive patients from 2 centers with unstable VT undergoing ECG-mapping guided VT ablation were prospectively enrolled. Using the ECG, computational mapping localized the VT onto a 3D model. A multielectrode catheter was placed at the predicted site and activation mapping was performed during VT reinduction. Ablation was performed per standard protocol. Time to ICD shock or death was compared using Cox regression (adjusting for age, EF and ICM) between ECG-mapping guided ablation vs standard VT ablation controls with a minimum 3 month f/u. Accuracy of ECG mapping was compared with activation mapping.Results:Out of 32 consecutive patients who underwent ECG-mapping guided VT ablation, 26 had unstable VT (age 66±10 yr, EF 34±17%). All 26 (100%) patients with unstable VT (median VT CL: 326±81ms) underwent successful activation mapping, Fig 1A. Using Cox regression adjusting for covariates, ECG-mapping guided VT ablation had a significant reduction in ICD shock or death compared to standard ablation controls (p=0.01, HR=0.23 [CI 0.07-0.70], Fig 1B). There was a 99% reduction in total ICD shocks during mean 7.5 month f/u. For all 32 patients, the mean accuracy of ECG mapping was 1.3±0.7cm when compared to invasive activation±entrainment mapping.Conclusions:Use of computational 12-lead ECG mapping to guide ablation of unstable VT significantly improved freedom from ICD shocks and death resulting in excellent accuracy.

Circulation, Volume 148, Issue Suppl_1, Page A12401-A12401, November 6, 2023. Background:Anticoagulant RCTs are thought to have enrolled younger and less comorbid patients with atrial fibrillation (AF) compared to the general population. We developed a representation score summarizing patient characteristics to describe how well RCT participants with AF reflect a nationally representative cohort.Methods:We studied adults >=65 years old with AF by harmonizing two data sources: (1) patient-level data from 12 landmark RCTs testing anticoagulants vs. placebo or antiplatelets from the Atrial Fibrillation Investigators (AFI) consortium and (2) the Health and Retirement Study AF cohort (HRS-AF), a representative cohort of older adults with AF in the U.S. We fit a logistic regression model to estimate the probability of inclusion in the HRS-AF cohort in the pooled sample using age, height, weight, gender, heart failure, hypertension, diabetes, prior stroke, and prior myocardial infarction. This estimate, the Trial Benchmark Score, reflected the probability of belonging to the HRS-AF cohort and ranged from 0 to 1, with higher scores reflecting a greater likelihood of belonging to the HRS-AF cohort. We plotted the distribution of scores for HRS-AF and AFI participants and compared the mean scores using a t-test.Results:Compared to the HRS-AF cohort (n=3542), AFI participants (n=7933) were younger (72 vs. 76yrs, standardized mean difference [SMD] -0.7), more frequently male (64% vs. 46%, SMD 0.3), and had a lower likelihood of prior stroke (19% vs. 23%, SMD -0.4). The mean Trial Benchmark Score differed significantly between the two cohorts (HRS-AF mean 0.47 vs. AFI mean 0.23, p0.47 (the HRS-AF mean score).Conclusion:Differences in the Trial Benchmark Scores distributions indicate a substantial difference in the distribution of observable characteristics and that RCT participants were not fully representative of the benchmark population.

Circulation, Volume 148, Issue Suppl_1, Page A17037-A17037, November 6, 2023. Introduction:The electrocardiogram (ECG) is a widely available diagnostic tool for evaluating cardiac patients. Although automated ECG interpretation has made significant progress, it has yet to match the accuracy demonstrated by physicians.Hypothesis:In this study, we hypothesized that an artificial intelligence based ECG system can achieve comparable performance to physicians in accurately identifying 20 essential ECG patterns.Methods:An AI-powered system comprising six deep neural networks (DNNs) was trained to identify 20 diagnostic patterns from 12-lead ECGs categorized into six groups: rhythm, infarction, conduction abnormalities, ectopy, chamber enlargement, and axis. An independent test set with the consensus of two expert cardiologists was used as a reference standard. We compared the system’s performance to that of three General Practitioners (GPs) and six individual cardiologists, using F1 scores as the evaluation metric.Results:The AI system was trained on 932,711 standard 12-lead ECGs from 173,949 patients. The independent test set comprised 11,932 annotated ECG labels. Figure 1 shows the respective F1 scores of the DNNs, average GP and average cardiologist as follows: Rhythm: 0.957 vs. 0.771 vs. 0.905; Infarction: 0.925 vs. 0.780 vs. 0.852; Conduction abnormalities: 0.893 vs. 0.714 vs. 0.851; Ectopy: 0.966 vs. 0.896 vs. 0.951; Chamber enlargement: 0.972 vs. 0.562 vs. 0.773; Axis: 0.897 vs. 0.601 vs. 0.685. The AI system’s diagnostic performance exceeded that of GPs and was on par with cardiologists for all individual diagnostic patterns.Conclusions:The AI-powered ECG system is able to accurately identify electrocardiographic abnormalities from the 12-lead ECG, highlighting its potential as a clinical tool for healthcare professionals.

Circulation, Volume 148, Issue Suppl_1, Page A16503-A16503, November 6, 2023. Introduction:Left coronary dominance is associated with worse clinical outcomes in the setting of coronary artery disease (CAD). Little is known about the determinants of dominance, including whether it is solidified in utero or sometime after birth.Hypothesis:Investigating the genetic determinants of dominance will identify new arteriogenesis pathways targetable for enhancing coronary artery growth and collateralization in humans.Methods:We conducted a stratified multi-population genome-wide association study of dominance using REGENIE and TopMed-imputed genotypes available in 43813 White, 9532 Black, and 3550 Hispanic participants of the Million Veteran Program (MVP) who had undergone a cardiac catheterization. We modelled left and co-dominant combined vs. right, as well as left vs. right dominance, adjusting for sex and 10 genetic principal components. We also explored whether dominance and CAD were genetically correlated using LD score regression. Lastly, we sought to validate compelling population genetic signals in mice using whole organ imaging of the heart, which allows for the quantification of coronary anatomy and variation.Results:Thirteen loci reached genome wide significance (GWS). The most intriguing was a top hit located downstream of C-X-C Motif Chemokine Ligand 12 (CXCL12) which, remarkably, reached GWS in both White and Black Veterans. While the CXCL12 signal robustly colocalized with the signal in this region for CAD (COLOC PPH4: 87.1%), the genome wide genetic correlation of dominance with CAD was otherwise modest in both MVP (rg=-0.18, p=0.007) and CardiogramplusC4D (rg=-0.16, p=0.018). To gain evidence for causality, we modeled dominance in Cxcl12-deficient mice (n=64 wildtype; n=62 heterozygous). Murine coronary arteries develop in a field of Cxcl12 expression and choose right or left dominance during embryogenesis, a time when we found human dominance to also be evident. Reminiscent of our human GWAS, dominance was skewed away from the common phenotype in Cxcl12 heterozygous mice (p=0.038).Conclusions:CXCL12 is a primary determinant of coronary artery dominance in humans and mice. The modest negative genetic correlation between CAD and dominance supports a developmental component to CAD susceptibility.

Circulation, Volume 148, Issue Suppl_1, Page A18317-A18317, November 6, 2023. Introduction:ECG construction with deep learning can help standardize ECGs and remove noise and artifacts, transforming potentially unusable ECGs into clinically useful ones. However, noise levels, often unpredictable in clinical settings, may impact model performance.Hypothesis:This study explores the relationship between ECG noise levels and deep learning model performance in constructing noise-free ECGs. We hypothesize that beyond a certain noise threshold, the model’s output ceases to be clinically usable.Aims:Our study aims to determine noise effects on the performance of deep learning models in constructing 12-Lead ECGs, providing insights into the model’s robustness and the optimal conditions for its application.Methods:Using 250 digital 12-Lead ECGs from the PTB database, we injected Gaussian noise (mean 0.0, standard deviations between 0.0 and 1.0) into these ECGs, all processed and scaled to unitless values (-1 to +1) and resampled at 100Hz. A model was asked to construct a noise-free ECG from a full 12-lead ECG with noise injection, which was compared to the original noise-free ECG.Results:The findings reveal two distinct linear phases in the relationship between noise and model performance (measured by mean absolute error, MAE). Between standard deviations of 0.0 and 0.03, the MAE increased marginally, while a sharp increase occurred between 0.03 and 1.0. Importantly, critical features of the original ECG were retained for noise levels up to 0.08, implying a noise threshold for clinical usability.Conclusions:Deep learning models, though not entirely resistant to noise, demonstrate efficacy in constructing ECGs up to certain noise levels (standard deviation 0.03). Beyond this, performance declines markedly. These findings outline the limits and potential of deep learning models like ECGio in clinical practice. Future research should focus on resilience to higher noise levels or noise-reduction preprocessing strategies.

Circulation, Volume 148, Issue Suppl_1, Page A17094-A17094, November 6, 2023. Introduction:As echocardiographic screening is limited by access to technology, structural heart disease (SHD) is often diagnosed after the development of clinical symptoms. To enable automated and accessible screening of SHD, we developed and validated a deep-learning model for 12-lead electrocardiograms (ECGs) for various screening populations.Methods:We used 12-lead ECGs with paired TTEs at the Yale New Haven Hospital (2015-2021) to develop convolutional neural networks for detecting SHD. SHD was defined as the presence of any one of the following on a transthoracic echocardiogram (TTE) performed within 30 days of the ECG: hypertrophic cardiomyopathy (IVSd > 1.5cm and diastolic dysfunction), LV ejection fraction < 40%, or severe left-sided valvular disease (aortic/mitral stenosis or regurgitation). We developed an ensemble XGBoost model based on predictions for individual SHDs and patient age and sex as a single screen across all SHDs. We also simulated the model performance across cohorts with varying disease prevalence.Results:The model was developed in 456,927 ECGs/118,623 individuals (66.5±16.0 years, 45.6% women, 16.5% Black) and validated in 13,181 ECGs/13,181 individuals (65.1±17.3 years, 49.5% women, 14.5% Black). In the held-out validation set, the ensemble XGBoost model achieved an AUROC of 0.92 (95% CI: 0.91-0.94) and an AUPRC of 0.64 (95% CI: 0.57-0.70), with a sensitivity of 85% and specificity of 88%. The AUROCs for the individual disease models ranged from 0.72-0.95. (Panel A) In the test set with 10% disease prevalence, the ensemble model had a PPV 44% and an NPV of 98%. (Panel B) In simulated cohorts with 5% and 20% disease prevalence, the model had reached NPV & PPV of 99% & 27%, and 98% & 64%, respectively. (Panel C)Conclusion:We developed a novel ensemble deep-learning model for detecting several SHDs directly from ECGs with high PPV and NPV. This approach represents a robust, scalable, and accessible modality for automated SHD screening.

Circulation, Volume 148, Issue Suppl_1, Page A13765-A13765, November 6, 2023. Introduction:There is an increasing clinical interest using wearable single-lead ECG sensors for continuous cardiac monitoring.Hypothesis:However, there is limited to qualify the reliability of those. The purpose of this study is to assess ECG signal quality assessments of real-time wearable SmartPatch (HiCardi Plus) compared to 12-lead ECG.Methods:To analyze the similarity between the standard 12-Lead ECG and HiCardi Plus signals, 31 subjects were simultaneously acquired with SmartPatch (HiCardi Plus) and 12-Lead ECG. The 12-lead ECG equipment used in this experiment is IntelliVue Patient Monitor MX700 (Philips)Results:It was found that even if the distance between electrodes was short, the ECG signal measured from the electrode pair located diagonal (direction of the electrical axis of the heart) and vertical direction with chest had a very high correlation with the standard 12-lead ECG signal and was sufficient for diagnosisConclusions:Our device (HiCardi Plus) has substantial equivalence with standard 12-Lead ECG, in terms of QRS-complex detection and arrhythmia classification in the intended use condition.