Early Detection of Low Back Pain: A Machine Learning Approach with Enhanced Data Techniques

Low back pain is a condition quite common to millions across the globe, usually leading to a high degree of disability and poor quality of living. It may result from posture faults, some deformities of the spine, injuries, or degenerative ailments. Worldwide, it leads to high healthcare and economic hindrances. Most people with low back pain have spinal deformities, such as deviations in pelvic tilt and lumbar angles, which may help in early detection and intervention. It is that acute detection would avert chronic complications, alleviation of pain, as well as improvement of the outcome of the patient. In this research, we intend to investigate the application of multiple machine-learning techniques toward early identification of LBP. We used a Kaggle dataset having 310 instances with 12 numeric attributes indicating spinal anomalies for addressing intrinsic class imbalance by SMOTE creation of more instances for the minority class. Moreover, to improve the robustness and diversity of the dataset, we adopted the bootstrapped resampling method to add reliability into model training by replicating those data points. Advanced machine learning models were trained on this enhanced dataset, and their performances were evaluated rigorously. Advanced Gradient Boosting model was exceptionally capable, overtaking the other techniques and those of previous research with perfect accuracy of 1.00. Each model underwent systematic fine-tuning to optimize its performance further, ensuring reliable and actionable results. This research comes as an excellent contribution to the field of LBP detection by providing strong and effective protocol which could change healing practice from one type of diagnosis and treatment to another.