Cancer classification using machine learning and HRV analysis: preliminary evidence from a pilot study.

Area of Science:

• Cardiology
• Oncology
• Biomedical Engineering

Background:

• Cancer patients often present with autonomic dysfunction, characterized by reduced heart rate variability (HRV).
• Distinguishing cancer patients from healthy individuals using physiological signals remains a challenge.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for discriminating between cancer patients and healthy controls.
• To assess the efficacy of HRV parameters derived from short ECG recordings for this discrimination.

Main Methods:

• Extracted 12 HRV features from 5-minute ECG recordings.
• Utilized Recursive Feature Elimination (RFE) to identify key discriminating features.
• Trained three base ML classifiers and an ensemble model using a stacking method.

Main Results:

• All selected HRV features (SDNN, RMSSD, pNN50%, HRV triangular index, SD1) were significantly different between groups.
• Base ML models performed above chance, with Random Forest achieving 83% accuracy.
• The ensemble model demonstrated a classification accuracy of 86% and an AUC of 0.95.

Conclusions:

• HRV parameters derived from short ECG recordings are promising inputs for ML-based differentiation of cancer patients.
• The developed ensemble ML model shows high potential for non-invasive cancer detection.
• Further validation with larger sample sizes is recommended.