Time-domain heart rate variability features for automatic congestive heart failure prediction

Area of Science:

• Cardiology
• Biomedical Engineering
• Data Science

Background:

• Heart failure is often diagnosed late in primary care due to limited diagnostic tools and overlapping symptoms.
• Non-invasive monitoring of heart rate variability (HRV) offers a potential solution for early detection.
• HRV reflects autonomic nervous system activity, crucial for cardiovascular health assessment.

Purpose of the Study:

• To evaluate the feasibility of using machine learning (ML) on HRV data for classifying healthy individuals and heart failure patients.
• To explore HRV as a non-invasive biomarker for heart failure detection.

Main Methods:

• Utilized digitized ECG recordings from 54 healthy adults and 44 heart failure patients (NYHA classes 1-3).
• Performed time-domain HRV analysis, calculating parameters like RMSSD, SDNN, SDSD, NN50, pNN50, and mRRi.
• Applied ML algorithms (SVM, KNN, Naïve Bayes, DT) with five-fold cross-validation to classify subjects based on HRV parameters.

Main Results:

• The best overall classification accuracy was 77% using KNN and DT with a 5-minute HRV window.
• KNN and DT achieved high accuracy (91%) in classifying severe congestive heart failure.
• Area Under the Curve (AUC) values for KNN and DT reached 0.77 and 0.78 overall, and 0.88 and 0.92 for severe cases, respectively.

Conclusions:

• HRV analysis combined with ML accurately predicts severe congestive heart failure.
• This non-invasive approach using HRV shows potential for primary care screening of heart failure.
• Further research could integrate ML-driven HRV analysis into routine clinical practice for early heart failure detection.