Deep learning in occupational safety and health and its contribution to workplace risk management: A systematic review

Occupational health and safety (OHS) is essential in all industrial sectors due to the significant impact of accidents and occupational illnesses. Traditionally, risk management has relied on periodic assessments and reactive control measures; however, artificial intelligence and deep learning have enabled the development of more proactive and efficient approaches. These technologies facilitate the analysis of large volumes of data to identify hidden patterns, predict risks, and improve workplace safety. In this context, this study analyzes the current use of deep learning in OHS, focusing on its implementation, perceived effectiveness, and the challenges it presents. Furthermore, it examines its application in sectors such as construction, logistics, and mining, where it contributes to risk prevention and the detection of unsafe behaviors. To this end, a systematic review was conducted following the PRISMA protocol, including 14 open-access articles selected from an initial pool of 274 publications retrieved from Scopus, PubMed, Web of Science, and DOAJ. The analysis concludes that deep learning has great potential to reduce unsafe behaviors by identifying and detecting key variables, such as information management, that influence the occurrence of incidents and hazardous conditions. Therefore, this technology emerges as a valuable tool to support OHS professionals in prevention, control, and decision-making in various work environments.