Interpretable Model to Support Differential Diagnosis Between Ischemic Heart Disease, Dilated Cardiomyopathy and Healthy Subjects

The differential diagnosis between Ischemic Heart Disease (IHD) and Dilated Cardiomyopathy (DCM) can often be challenging, because only invasive, and not largely available exams can provide a definite diagnosis. The echocardiographic left ventricular ejection fraction (LVEF) and global longitudinal strain (GLS) as well as ECGheart rate variability (HRV) analysis are shown to be helpful tools for diagnosing several cardiac diseases. There is also a growing interest in application of interpretable machine learning techniques to guide the diagnosis. We aimed to produce an interpretable model applied for differential diagnosis between DCM, IHD and healthy subjects (HC) based on LVEF, GLS and HRV features. The study encompassed three groups: 130 DCM, 164 IHD, and 152 HC subjects. The novel GLS, LVEF, and linear and non-linear HRV features were extracted for each subject. Then, the interpretable models were produced by a logistic regression algorithm considering a set of features chosen with the ReliefF method. The results showed that the most informative features for classification between IHD, DCM e HC were: GLS, LVEF, age, FD, SD1/SD2 and sex, listed in order of importance. The obtained classification accuracy was 70% and the area under the ROC curvewas 83.4%. The study demonstrates that a logistic regression model and its nomograms allow detailed clinical interpretation of the model and may be a powerful tools support differential diagnosis between IHD, DCM and HC.