Pain Intensity Recognition Rates via Biopotential Feature Patterns with Support Vector Machines

Area of Science:

• Biomedical Engineering
• Computational Neuroscience
• Pain Medicine

Background:

• Current pain diagnosis relies on subjective patient reports, lacking objective measurement.
• Existing tools like verbal, visual analog (VAS), and numeric rating scales (NRS) have limitations in reliability, validity, and applicability to certain patient groups.
• Biopotential signal analysis using machine learning presents a promising alternative for objective pain assessment.

Purpose of the Study:

• To develop and evaluate an automated pain recognition system using biopotential signals.
• To create a comprehensive biopotential database for pain research.
• To optimize the performance of machine learning algorithms for pain intensity measurement.

Main Methods:

• Collected electromyography, skin conductance level, and electrocardiography signals from 85 participants under controlled painful heat stimuli.
• Extracted 159 features from signal analysis, including amplitude, frequency, and similarity.
• Utilized machine learning algorithms to classify pain intensity levels.

Main Results:

• Achieved high classification rates: 90.94% for baseline vs. pain tolerance and 79.29% for baseline vs. pain threshold.
• Facial electromyography signals, particularly features from amplitude and similarity groups, were most effective for pain recognition.
• The developed system demonstrates significant potential for automated pain assessment.

Conclusions:

• Machine learning-based pain measurement using biopotentials can provide objective data for clinical teams.
• This technology can enhance treatment assessment and patient care.
• Further development could lead to more reliable and efficient pain diagnostics.