Naive Bayesian-based Nomogram for Identification of Early Asymptomatic Dilated Cardiomyopathy

Abstract

Dilated cardiomyopathy (DCM) is one of the leading causes of heart failure. The most used parameter for measuring heart function and predicting outcomes in DCM patients is the left ventricular ejection fraction (LVEF). However, it has some inherent drawbacks. Recent studies have reported that left ventricular global longitudinal strain (GLS) and heart rate variability (HRV) can be used for the prediction of DCM. Therefore, we aimed to investigate a Naive Bayesian-based nomogram produced on clinical and instrumental features, which can be used to support the diagnosis of early asymptomatic DCM. The study encompassed 49 DCM and 50 healthy subjects (HC).The models were produced by naive Bayes algorithms considering the set of selected HRV features, LVEF, GLS, age, and sex. The results showed that the most informative parameters were: GLS, LVEF, meanRR, SD2 age, and sex, listed in order of importance. The obtained classification accuracy was 80%. A naive Bayesian-based nomogram highlighted that GLS brings more information than LVEF, followed by HRV features, age and sex. In conclusion, this study demonstrates that a Naive Bayesian-based nomogram is a powerful tool for the prediction of early asymptomatic DCM and that allows detailed clinical interpretation of the produced model.