LAYER: A Quantitative Explainable AI Framework for Decoding Tissue-Layer Drivers of Myofascial Low Back Pain.

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Artificial Intelligence

Background:

• Myofascial pain (MP) is a major cause of chronic low back pain.
• Current research lacks reliable biomarkers and focuses mainly on muscle, neglecting other soft tissues.

Purpose of the Study:

• To develop and validate an explainable AI framework (LAYER) for analyzing soft tissue contributions to MP.
• To identify specific tissue layers and their roles in MP prediction using 3D ultrasound.

Main Methods:

• Developed LAYER, an AI framework for Layer-wise Analysis for Yielding Explainable Relevance Tissue.
• Analyzed over 4,000 3D ultrasound scans across six tissue layers.
• Quantified tissue layer contributions to MP prediction using B-mode and shear-wave imaging.

Main Results:

• Non-muscle tissues, particularly the deep fascial membrane (DFM), significantly contribute to MP prediction.
• The collective saliency of non-muscle layers nearly matched muscle's saliency in combined imaging.
• LAYER demonstrated high saliency for DFM in B-mode imaging (0.420).

Conclusions:

• Challenges the traditional muscle-centric paradigm in myofascial pain research.
• Highlights the importance of fascia and other non-muscle tissues in low back pain.
• Provides a novel, interpretable framework for soft-tissue imaging analysis and identifies new therapeutic targets.