Non-Specific Low Back Pain: An Inductive Exploratory Analysis through Factor Analysis and Deep Learning for Better...

Area of Science:

• Biomedical Engineering
• Rehabilitation Science
• Data Science

Background:

• Non-specific low back pain (NSLBP) is a widespread health issue with limited understanding of its causes and effective treatments due to population diversity.
• Current treatment approaches for NSLBP are often generic, leading to varied patient responses and highlighting the need for personalized interventions.
• Clustering NSLBP patients based on shared characteristics is a promising strategy, but previous attempts faced challenges due to subjective data and complexity.

Purpose of the Study:

• To investigate the influence and importance of objective, continuous variables for clustering NSLBP patients into meaningful subgroups.
• To explore the utility of high-density electromyography (HD EMG) and motion capture data in differentiating NSLBP from healthy individuals.
• To assess the predictive power of different data types (anthropometric, biomechanical, neuromuscular, balance) in classifying NSLBP.

Main Methods:

• Acquired high-density electromyography (HD EMG) and motion capture data from 46 subjects performing six movement tasks at two speeds.
• Employed deep neural network and factor analysis for exploratory analysis and variable importance assessment.
• Trained classification models using various data subsets (all variables, anthropometric, biomechanical, neuromuscular, balance) to distinguish between healthy individuals and those with NSLBP.

Main Results:

• Classification models achieved high accuracy (up to 93.30% overall, 94.40% for anthropometric data) in identifying NSLBP.
• Factor analysis indicated that NSLBP patients exhibit distinct movement patterns, characterized by slower and more rigid motions compared to healthy individuals.
• Anthropometric variables (age, sex, BMI) showed significant correlations with NSLBP components, and neuromuscular and biomechanical data also contributed to accurate classification.

Conclusions:

• Objective data, including anthropometric measurements, movement patterns, and neuromuscular activity, are valuable for identifying individuals with NSLBP.
• A comprehensive understanding of NSLBP requires integrating multiple data domains rather than analyzing them in isolation.
• Simplifying dynamic data acquisition and further exploring specific movements like back flexion and trunk rotation could enhance future research and clinical applications.