Classification of lbp patients using imu signal and machine learning approaches

Low back pain (LBP) remains a critical issue in primary care. Currently, clinical practitioners rely on subjective measures/questionnaires to categorize LBP patients to pursue specific treatment regimens. However, the need for new objective scales to assess the level of patients’ prognosis risk is evident. Our goal was to develop a model to discriminate the patients with high vs. low/medium risk of LBP. To this end, an inertial measurement unit (IMU) was placed on the patients’ chest while they were performing trunk flexion/extension task. Machine learning algorithms such as support vector machine (SVM) and multi-layer perceptron (MLP) were implemented, where we obtained the accuracy of 89% and 71%, respectively. Our preliminary efforts in replacing the STarT Back Screening Tool (SBST) with an objective approach can facilitate the development of an approach for subgrouping the LBP patients, which may lead to the enhanced decision making of physicians in prescribing the treatment procedures.