Current Trends in Artificial Intelligence for Recognizing Work Postures to Prevent Work-Related Musculoskeletal Disorders: Systematic Review and Meta-Analysis by Occupational Activity

The use of artificial intelligence (AI) to recognize postures is a promising approach for the prevention of work-related musculoskeletal disorders (WMSDs). The aim was to conduct a systematic review with meta-analysis to assess the performance of work posture recognition systems during occupational activity. The results were reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. The Google Scholar, IEEE Xplore, PubMed/MedLine, and ScienceDirect databases were screened without date restrictions. Two authors independently selected articles and extracted data. Studies were included if they presented a performance analysis of an AI deep learning (DL) or machine learning (ML) method that assessed the WMSD risk associated with working postures. Only peer-reviewed studies written in English including accuracy, precision, specificity, sensitivity, or F1-score values were included. The risk of bias was assessed using the Prediction Model Study Risk of Bias Assessment Tool. Of the 157 unique records, 58 studies were selected. The five performance parameters were investigated and averaged for seven occupational activities, eight posture categories, and the AI methods (ML vs. DL). Statistical analyses showed that DL methods produced better results. The reported systems detected sitting and standing postures with high accuracy. The solutions proposed in Manufacturing and Construction were the most numerous and the most effective on average. The major limitation lies in the wide variety of methods used. This analysis is a valuable source of information for designing new detection systems that are effective, ergonomic, easy to use, and acceptable so that humans remain at the center of the production process as defined by Industry 5.0.