Scalp EEG-Based Pain Detection Using Convolutional Neural Network.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Pain is a complex sensory and emotional experience influenced by brain activity.
• Objective pain detection is challenging, necessitating advanced neurophysiological methods.
• Recent advancements in brain recording and AI offer new avenues for pain assessment.

Purpose of the Study:

• To propose and evaluate a novel pain detection framework using Electroencephalogram (EEG) and deep convolutional neural networks (CNN).
• To objectively distinguish between pain states and resting states in chronic back pain patients.
• To analyze brain activity patterns associated with different pain induction methods.

Main Methods:

• Utilized Electroencephalogram (EEG) signals from 10 chronic back pain patients.
• Induced pain through two distinct methods: movement stimulation (MS) and video stimulation (VS).
• Employed a multi-layer deep convolutional neural network (CNN) for classifying EEG segments into resting and pain states.

Main Results:

• The CNN-based framework achieved high performance in pain detection, with an Area Under the Curve (AUC) of 0.83 ± 0.09 for MS and 0.81 ± 0.15 for VS.
• Performance exceeded current state-of-the-art approaches.
• Analysis revealed distinct brain topography differences between MS-induced (generalized) and VS-induced (partial) pain.

Conclusions:

• The proposed EEG and CNN framework offers a robust and effective solution for objective pain detection.
• This neurocomputing approach has significant potential for clinical applications and research in pain management.
• Understanding brain activity patterns associated with different pain stimuli can refine diagnostic and therapeutic strategies.