Vertebral Force Prediction during Manual Lifting using Sagittal Plane Posture and Machine Learning

Vertebral force during manual lifting causes lower back pain (LBP) among workers in various occupational fields. Continuous monitoring of vertebral force is required for preventing LBP. Vertebral force can be measured directly by implant devices; however, these implant devices are difficult to install on general workers. Non-invasive technologies such as an optical motion capture system could estimate vertebral force, however, special measurement devices are necessary. The objective of this study is to propose non-invasive measurement method for vertebral force using ordinary digital camera. The proposed method predicts vertebral forces by machine learning and sagittal plane posture during manual lifting. The proposed method predicted vertebral forces for 48 trials manual lifting via leave-one-out-cross-validation. Training and testing of the proposed method used OrthoLoad database that included sagittal 2D images and vertebral force obtained from implant device. The results showed that the proposed method using model tree algorithm could predict vertebral force by greater than 0.7 correlation with actual values. These results indicate possibility that the proposed method can be used for vertebral force assessments using ordinary devices.