Early Detection of Sciatica Using Ensemble Learning Algorithms

Sciatica is a musculoskeletal disorder caused by compression or irritation of the lumbar nerve roots. Gait and posture abnormalities that appear before clinical symptoms can be detected early by continuously monitoring biomechanical parameters such as pelvic movement and plantar pressure. This paper proposes an ensemble learning approach that uses a lumbar inertial measurement unit (IMU), which contains an accelerometer and a gyroscope, together with a multi-node plantar pressure insole. The sensor signals are fused, and engineered time-domain and statistical features are used by a random forest (RF) model to classify short windows as either normal or at risk. On a held-out test set of 250 windows, the model achieved an overall accuracy of 95%.