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

AI can reveal overlooked contributors to pain: The LAYER framework discovered that non-muscle tissues—particularly the deep fascial membrane and fat layers—contribute nearly as much to myofascial low back pain prediction as muscle tissue itself (0.316 vs 0.317 saliency scores), suggesting decades of muscle-focused treatment approaches may have missed important therapeutic targets.

Explainability matters for medical AI credibility: By making visible which specific tissue layers the AI model relied on to make predictions, LAYER demonstrates how explainable AI can challenge established medical assumptions and build clinician trust, rather than operating as a "black box" that doctors cannot verify or understand.

Larger datasets with diverse imaging enable better insights: The study analyzed over 4,000 ultrasound scans using multiple imaging modes (B-mode and shear-wave), showing that responsible AI development requires substantial, multi-faceted data to uncover nuanced biological patterns that smaller studies might miss.

Anatomically grounded AI design improves real-world applicability: Rather than treating tissue layers as abstract data points, LAYER was built with anatomical knowledge embedded into its structure, making findings more interpretable to medical professionals and more likely to translate into actual clinical improvements.

AI-driven discoveries can reshape clinical practice: These findings suggest that pain management strategies may need to shift from muscle-centric interventions to multi-tissue approaches, demonstrating how responsible AI can generate actionable insights that potentially improve patient outcomes.