Personalized LSTM Models for ECG Lead Transformations Led to Fewer Diagnostic Errors Than Generalized Models: Deriving 12-Lead ECG from Lead II, V2, and V6.

BACKGROUND AND OBJECTIVE: The prevalence of chronic cardiovascular diseases (CVDs) has risen globally, nearly doubling from 1990 to 2019. ECG is a simple, non-invasive measurement that can help identify CVDs at an early and treatable stage. A multi-lead ECG, up to 15 leads in a wearable form factor, is desirable. We seek to derive multiple ECG leads from a select subset of leads so that the number of electrodes can be reduced in line with a patient-friendly wearable device. We further compare personalized derivations to generalized derivations. METHODS: Long-Short Term Memory (LSTM) networks using Lead II, V2, and V6 as input are trained to obtain generalized models using Bayesian Optimization for hyperparameter tuning for all patients and personalized models for each patient by applying transfer learning to the generalized models. We compare quantitatively using error metrics Root Mean Square Error (RMSE), R2, and Pearson correlation (ρ). We compare qualitatively by matching ECG interpretations of board-certified cardiologists. RESULTS: ECG interpretations from personalized models, when corrected for an intra-observer variance, were identical to the original ECGs, whereas generalized models led to errors. Mean performance values for generalized and personalized models were (RMSE-74.31 µV, R2-72.05, ρ-0.88) and (RMSE-26.27 µV, R2-96.38, ρ-0.98), respectively. CONCLUSIONS: Diagnostic accuracy based on derived ECG is the most critical validation of ECG derivation methods. Personalized transformation should be sought to derive ECGs. Performing a personalized calibration step to wearable ECG systems and LSTM networks could yield ambulatory 15-lead ECGs with accuracy comparable to clinical ECGs.

Atrial fibrillation: Utility of CHADS2 and CHA2DS2-VASc scores as predictors of readmission, mortality and resource utilization.

BACKGROUND: CHADS2 and CHA2DS2-VASc scores are widely used for thromboembolic risk assessment in Atrial Fibrillation(AF) cohort, however further utilization to predict outcomes is understudied. METHOD: HCUP's National Readmission Data(NRD) 2013 was queried for AF admissions using ICD-9-CM code 427.31 in principal diagnosis field. Patients with mitral valve disease or repair/or replacement were excluded to estimate population with non-valvular AF only. CHADS2 and CHA2DS2-VASc were calculated for each patient. Hierarchical two-level logistic and linear models were used to evaluate study outcomes in terms of mortality, 30 or 90-day readmissions, length of stay(LOS) and cost. RESULT: Of 116,450 principal non-valvular AF admissions(50.2% female and 43.1% age≥75years) 29,179 patients were readmitted, with total 40,959 readmissions. Higher CHADS2 and CHA2DS2-VASc score were associated with increased mortality from 0.4% for CHADS2 of 0 to 3.2% for score of 6 and from 0.2% for CHA2DS2-VASc of 0 to 3.2% for score≥8. LOS increased from 2.20days for CHADS2 of 0 to 5.08days for score of 6, while cost increased from $7888 to $11,151. 30-day readmission rate increased from 8.9% for CHADS2 of 0 to 26.0% for score of 6, and 90-day readmission rate increased from 15.2% to 39%. CHA2DS2-VASc scoring similarly demonstrated a trend towards increasing readmission rate, LOS and cost for higher scores. Also, similar results were seen in hierarchical modeling with increment of CHADS2 and CHA2DS2-VASc scores. CONCLUSION: CHADS2 and CHA2DS2-VASc scores can be used as quick surrogate markers for predicting outcomes beyond thromboembolic risk. Physician familiarity with these systems makes them easy to use bedside clinical tools to improve outcomes and resource allocation.

The presence of diastolic and systolic dysfunction in patients with impaired relaxation filling pattern.

BACKGROUND: Impaired relaxation filling pattern in the presence of preserved left ventricular ejection fraction is a common finding which is ascribed to early diastolic dysfunction. We thought to determine the distribution of various markers of systolic and diastolic dysfunction in these patients. METHODS: A total of 106 patients were included in this retrospective observational study. Diastolic dysfunction was defined using the American Society of Echocardiography recommendations ("50% rule"). The systolic dysfunction was evaluated by global longitudinal strain and circumferential strain. For further analysis, the patients were divided into the lower strain (lower tertile) and higher strain (upper tertiles) groups based on longitudinal strain. RESULTS: There were marked differences between the groups in essential echocardiographic parameters. Patients in the lower strain group were more likely to have definite diastolic dysfunction (23% vs 7%, P<.01) and less likely to have normal diastolic function (54% vs 80%, P<.01). They also showed lower circumferential strain (-21.2% vs -25.7%, P<.01). In multivariate analysis, left atrial enlargement (odds ratio (OR) 4.6, 95% confidence interval (CI) 1.2-17.1), left ventricular mass index (OR 1.03, 95% CI 1.0-1.1), left ventricular end-diastolic volume (OR 1.1, 95% CI 1.0-1.1), and E-wave deceleration time (OR 0.99, 95% CI 0.98-1.0) were independently associated with lower global longitudinal strain. CONCLUSIONS: Patients with impaired relaxation filling pattern in the presence of preserved left ventricular ejection fraction have a wide spectrum of left ventricular systolic and diastolic performance abnormalities. Isolated assessment of diastolic dysfunction may not fully characterize this group of patients.