Advanced machine learning applications in fibromyalgia to assess the relationship between 3D spinal alignment with clinical outcomes.

This study leveraged machine learning (ML) models to explore the relationship between three-dimensional (3D) spinal alignment parameters and clinical outcomes in patients suffering from fibromyalgia syndrome (FMS). A cohort of 303 FMS patients, diagnosed according to the 2016 American College of Rheumatology criteria, underwent comprehensive assessments of sagittal imbalance, coronal imbalance, vertebral rotation, pelvic obliquity, pelvic torsion, and pelvic rotation using a validated 3D imaging system. Clinical outcomes, included the fibromyalgia impact questionnaire (FIQ), pain catastrophizing scale (PCS), Pittsburgh sleep quality index (PSQI), and algometric pain scores. Five ML models were employed: Fast Kolmogorov-Arnold Networks with Bee Colony Optimization (FastKAN-BCO), FastKAN with LBFGS, Multilayer Perceptron with LBFGS (MLP-LBFGS), Multilayer Perceptron with ADAM (MLP-ADAM), and linear regression. Among the models tested, FastKAN-BCO demonstrated the highest R-squared value (0.95) for algometric pain, while the MLP-LBFGS model achieved superior performance for PCS (R2 = 0.94), FIQ (R2 = 0.88), and PSQI (R2 = 0.97) predictions. Sagittal imbalance and pelvic obliquity were identified as key predictors of symptom severity. Stratification revealed that individuals with more pronounced pelvic asymmetry and vertebral rotation exceeding 10° experienced increased symptom intensity. The contribution of vertebral rotation was nonlinear, indicating a threshold-dependent impact. This study illustrates the potential of ML techniques to uncover complex associations between 3D spinal alignment and FMS outcomes, offering a foundation for personalized diagnostic and therapeutic approaches. The results emphasize the critical role of postural dysfunction in FMS and highlight the potential of advanced ML models.