Proof-of-concept: Differentiating upper trapezius muscle with myofascial trigger point using deep learning model on a...

Area of Science:

• Biomedical Engineering
• Neuroscience
• Pain Management

Background:

• Myofascial pain syndrome often originates from myofascial trigger points (MTrPs).
• Needle EMG (iEMG) detects spontaneous activity at MTrPs, but surface EMG (sEMG) validation is inconsistent.
• Non-invasive methods are needed to identify MTrPs.

Purpose of the Study:

• To determine if sEMG signals alone can differentiate MTrPs.
• To investigate the discriminative potential of sEMG for identifying latent MTrPs.
• To compare different machine learning models and signal processing techniques for MTrP detection.

Main Methods:

• sEMG signals were collected from healthy controls and patients with myofascial pain.
• An automated algorithm segmented pre- and post-intervention contraction data.
• A 1D Convolutional Neural Network (CNN) was trained and evaluated using cross-validation.

Main Results:

• The best subject-level performance (AUC 0.940) was achieved using pairwise logistic loss on pre-intervention sEMG signals.
• Latent-only analysis showed the highest AUC (0.916) with combined pre- and post-intervention signals.
• sEMG data demonstrated discriminative information for distinguishing MTrP subjects from controls.

Conclusions:

• Upper trapezius sEMG signals contain sufficient information to identify latent MTrPs.
• This proof-of-concept study supports the use of sEMG for non-invasive MTrP detection.
• Further large-scale validation studies are warranted.