GLEAM: A multimodal deep learning framework for chronic lower back pain detection using EEG and sEMG signals

Area of Science:

• Biomedical Engineering
• Computational Neuroscience
• Artificial Intelligence in Healthcare

Background:

• Low back pain (LBP) is a global health challenge, causing significant disability and economic burden.
• Current diagnostic methods for LBP intensity can be subjective and lack precision.
• There is a critical need for objective and accurate tools to assess LBP severity.

Purpose of the Study:

• To develop and validate a novel deep learning framework for precise LBP intensity classification.
• To investigate the efficacy of combining electroencephalography (EEG) and surface electromyography (sEMG) signals for LBP diagnosis.
• To introduce an advanced model capable of distinguishing between four levels of LBP: none, mild, moderate, and intolerable.

Main Methods:

• A hybrid deep learning model, GLEAM (GAN-Convolution-sElf Attention-ETLSTM), was designed.
• A denoising Generative Adversarial Network (GAN) was utilized to enhance the quality of EEG and sEMG signals.
• Convolutional Neural Networks (CNNs), self-attention mechanisms, and an ETLSTM network were integrated for feature extraction and classification.

Main Results:

• The GLEAM model achieved a high accuracy of 98.95% in classifying LBP intensity.
• The denoising GAN effectively removed noise from EEG and sEMG signals, improving data quality.
• The integrated architecture successfully captured both local and global patterns in the physiological signals.

Conclusions:

• The proposed GLEAM framework offers a robust and reliable method for objective LBP intensity assessment.
• This deep learning approach demonstrates significant potential for improving LBP diagnosis and patient management.
• The novel integration of denoising GANs and ETLSTM networks represents a significant advancement in biosignal analysis for pain assessment.