Optimal Length of Heart Rate Variability Data and Forecasting Time for Ventricular Fibrillation Prediction Using...

Area of Science:

• Cardiology
• Biomedical Engineering
• Data Science

Background:

• Ventricular fibrillation (VF) is a primary cause of global mortality.
• Heart rate variability (HRV) serves as a key biomarker for detecting and predicting dangerous arrhythmias.
• Early prediction of VF is critical for preventing sudden cardiac death.

Purpose of the Study:

• To evaluate the efficacy of HRV features in predicting the onset of ventricular fibrillation.
• To determine optimal HRV data lengths and forecast times for VF prediction.
• To assess the performance of an artificial neural network (ANN) for VF prediction using HRV.

Main Methods:

• Extracted features from seven different HRV data lengths.
• Utilized an artificial neural network classifier trained and validated with 10-fold cross-validation.
• Tested prediction accuracies across nine different forecast times.

Main Results:

• Maximum prediction accuracies of 88.18% and 88.64% were achieved with 10s and 20s HRV data lengths, respectively, at a 0s forecast time.
• The lowest prediction accuracy was observed at a 70s HRV data length and an 80s forecast time.
• Features from HRV signals close to VF onset demonstrated high predictive power.

Conclusions:

• HRV analysis, particularly using short data segments preceding the event, can effectively predict ventricular fibrillation.
• Artificial neural networks are suitable for predicting VF based on HRV features.
• Near-event HRV features offer a promising approach for real-time VF prediction and prevention strategies.