Related Experiment Video
Updated: Aug 14, 2026
10:25
Technical Approach for Infrared Tracking for Soft Tissue Navigation with a Holographic Head-Mounted Display and Preclinical Validation
Published on: September 2, 2025
HRV-Based Surrogate Classification in Surgical Environments Using Personalised Labelling and Sequence Modelling
Abdulaziz Almuhini1,2, Viji Ahanathapillai2, Zeeshan Raza2,3
1Department of Biomedical Technology, College of Applied Medical Sciences in Al-Kharj, Prince Sattam bin Abdulaziz University, Al-Kharj 11942, Saudi Arabia.
Sensors (Basel, Switzerland)
|August 13, 2026
Summary
Personalized heart rate variability (HRV) analysis using deep learning effectively identifies stress states in surgeons. This approach, utilizing subject-specific thresholds and sequence-based models, shows promise for enhancing surgical safety and performance.
Area of Science:
- Physiological monitoring in healthcare
- Machine learning applications in surgery
- Wearable sensor technology
Background:
- Accurate stress and workload monitoring is crucial in surgical environments for patient safety and performance.
- Heart rate variability (HRV) offers objective, continuous stress assessment but existing models lack real-world generalizability.
- Current methods often use population-level models and limited temporal analysis, failing to capture individual physiological differences.
Purpose of the Study:
- To develop an end-to-end pipeline for classifying stress-like physiological states in surgeons using HRV.
- To combine personalized feature extraction, surrogate label generation, and deep learning for stress assessment.
- To evaluate sequence-based models (CNNs, LSTMs) for HRV analysis in surgical settings.
Main Methods:
- Collected HRV data from surgeons during colorectal procedures using wearable sensors.
- Processed minute-level HRV features and generated subject-specific stress labels.
- Evaluated deep learning models (CNNs, LSTMs) with varying temporal windows and validation strategies, including leave-one-subject-out (LOSO).
Main Results:
- The Standard Deviation of Normal-to-Normal intervals (SDNN) effectively distinguished physiological states.
- Optimal performance was achieved with temporal windows of 10-20 minutes, with comparable results for CNN and LSTM models.
- Leave-one-subject-out (LOSO) validation demonstrated performance comparable or superior to pooled random splits, emphasizing personalized labeling benefits.
Conclusions:
- Personalized HRV-based modeling and temporal sequence analysis offer a feasible method for identifying stress-like states in surgical environments.
- The findings highlight the critical importance of rigorous evaluation designs, particularly personalized labeling, for physiological machine learning systems.
- This approach holds potential for improving surgical performance and patient safety through objective stress monitoring.
