Integrating multidimensional data analytics for precision diagnosis of chronic low back pain

Area of Science:

• Biomedical research
• Orthopedics
• Data science in healthcare

Background:

• Low back pain (LBP) is a global health issue, with a significant portion becoming chronic (cLBP).
• The multifactorial nature of cLBP makes identifying key contributors challenging.
• Understanding these contributors is crucial for effective management and treatment.

Purpose of the Study:

• To identify the most effective modalities for differentiating individuals with chronic low back pain (cLBP).
• To leverage machine learning for variable importance selection in cLBP classification.
• To guide the development of targeted diagnostics and personalized treatment strategies for cLBP.

Main Methods:

• Utilized a comprehensive dataset of 1,161 adults, including questionnaire data, clinical/functional assessments, and spino-pelvic MRI (144 parameters).
• Employed Boruta and random forest algorithms for variable importance selection and cLBP classification.
• Validated findings using an unseen holdout dataset.

Main Results:

• A multimodal model incorporating questionnaire, clinical, and MRI data demonstrated the highest efficacy in differentiating cLBP.
• Identified nine robust variables: psychosocial factors, neck/hip mobility, lower lumbar disc herniation, and degeneration.
• These key predictors remained consistent in the holdout dataset.

Conclusions:

• A multi-dimensional approach is essential for understanding and managing chronic low back pain.
• Specific factors including psychosocial elements, physical mobility, and lumbar spine conditions are critical differentiators for cLBP.
• Findings support the development of precise diagnostic tools and tailored therapeutic interventions for cLBP patients.