Potential Prognostic Markers in the Heart Rate Variability Features for Early Diagnosis of Sepsis in the Pediatric...

Area of Science:

• Biomedical Engineering
• Pediatric Critical Care
• Computational Biology

Background:

• Sepsis is a life-threatening condition requiring timely diagnosis and treatment.
• Early detection of sepsis is crucial for improving patient survival rates.
• Heart rate variability (HRV) shows potential as an early prognostic biomarker for sepsis.

Purpose of the Study:

• To investigate the utility of HRV features for early sepsis detection in neonates and pediatrics.
• To identify specific HRV features that can predict sepsis onset before clinical symptoms appear.
• To develop a machine learning model for automated sepsis classification based on HRV.

Main Methods:

• Extracted 52 hourly linear and non-linear HRV features from electrocardiogram (ECG) data of sepsis patients (n=14) and controls (n=14).
• Analyzed HRV feature variations in the 24 hours preceding clinical sepsis diagnosis.
• Trained a convolutional neural network (CNN) using HRV features from 22 hours prior to diagnosis for classification.

Main Results:

• Variations in specific HRV feature sets within 24 hours effectively provided prognostic information about sepsis evolution.
• A particular set of HRV features at 22 hours pre-diagnosis showed significant differences between sepsis and control groups.
• The CNN model achieved 88.89±7.86% accuracy in classifying sepsis versus non-sepsis individuals.

Conclusions:

• HRV analysis offers a promising non-invasive method for the early prediction of sepsis in pediatric populations.
• Machine learning models trained on HRV data can reliably distinguish between sepsis and non-sepsis states.
• This approach has the potential to significantly improve sepsis diagnosis timing and patient outcomes.