Enhancing differential diagnosis of IHD and DCM using interpretable machine learning in mildly reduced ejection...

Area of Science:

• Cardiology
• Biomedical Engineering
• Machine Learning in Medicine

Background:

• Accurate etiological diagnosis is vital for left ventricular dysfunction, as dilated cardiomyopathy (DCM) and ischemic heart disease (IHD) share early-stage symptoms.
• Differentiating between DCM and IHD is challenging, especially in patients with mildly reduced left ventricular ejection fraction (LVEF).

Purpose of the Study:

• To assess the discriminative capability of global longitudinal strain (GLS) and heart rate variability (HRV) parameters.
• To utilize interpretable machine learning models for differentiating DCM and IHD in patients with LVEF between 40% and 50%.

Main Methods:

• A retrospective study included patients with LVEF 40-50% and recent 24-h Holter ECG.
• HRV features and GLS were extracted from ECG and echocardiography.
• Interpretable predictive models were built using HRV, GLS, sex, and age for classification.

Main Results:

• The logistic regression model achieved 76% classification accuracy and an 83% area under the curve.
• Key differentiating features included sex, age, mean RR, FD, HFn, GLS, pNN50, SD1/SD2, SD1, and LFn.
• The study included 97 DCM and 91 IHD patients.

Conclusions:

• A novel approach integrating HRV metrics and myocardial deformation parameters aids in differential diagnosis.
• This predictive model supports distinguishing between DCM and IHD in patients with mildly reduced ejection fraction.