LIFTIX: Software for Evaluation of Lifting Posture Integrating Ergonomics and Machine Learning

Lumbar load in the lifting posture has been evaluated and improved to prevent lower back pain in various occupational fields, as awkward postures lead to lower back pain during manual lifting.Generally, lifting posture and lumbar load are evaluated through human observation or technical musculoskeletal simulation.Human observation has limitations in terms of repeatability and accuracy.Musculoskeletal simulators require professional skills and operational time for accurate analyses.Therefore, an easy and accurate system for evaluating lifting postures is required.Thus, the objective of this study was to develop and verify easy and accurate evaluation software for lifting postures that can be used by beginners for ergonomic posture assessment.The developed LIFTIX system calculates the compression force of L5-S1 using a combination of the Hand-Calculation Back Compressive Force Estimation Model (HCBCF) and machine learning-based pose estimation using the MediaPipe model.In addition, LIFTIX has functions for computational simulation between parameters of the biomechanical model, such as the joint angle and compression force of the L5-S1.Furthermore, these calculations and functions can be performed using a graphical user interface (GUI).In the experiment, the accuracy of the developed system, LIFTIX, was verified by comparing it with an existing ergonomic musculoskeletal simulator (3DSSPP) for 10 lifting postures.The results showed that LIFTIX could automatically evaluate the compression force of L5-S1 with a significantly high correlation (r=0.794,p<0.05) with the existing musculoskeletal simulator.These results indicate that the developed LIFTIX system can be used to evaluate lifting postures in workplaces.