Background and aims: Machine learning helps predict and make decisions about the diagnosisand treatment of diseases using data and previous experiences through statistical techniques andcomputer programs. Machine learning models have been designed to predict outcomes, diagnoses,and mortality within health care. One of the
machine learning applications is the diagnosisand management of burn wounds. However,
machine learning for burn wound management isstill questionable and challenging. Considering the growth of ML in medicine and the complexitiesand challenges of burn care, this review specializes on
machine learning applications in burnwound management. This study aimed to assess the potential roles and applications of machinelearning for burn wound management.Method: The systematic review protocol followed the Preferred Reporting Items for SystematicReviews and Meta-Analyses (PRISMA) guidelines. An extensive search was conducted in onlinedatabases including PubMed, ISI, Scopus, Google Scholar, and Science direct with the keywordssuch as “Machine Learning”, “Transfer Learning”, “Artificial Intelligence”, “Burns”, “Woundsand Injuries”, “Wound Healing “, from the earliest records up to October ۲۰, ۲۰۲۲. Also, all originalEnglish articles related to the purpose of the present study were included in the study. Lettersto the editor, opinions, conference abstracts, interventions, and reviews were excluded from thisstudy. The appraisal tool for cross-sectional studies (AXIS tool) was used to assess the quality ofincluded studies. All stages of search and quality evaluation of articles were conducted by tworesearchers independently.Results: Six out of ۵۲۴ studies were included in the study. Various roles include preparing machinelearning algorithms for burn assessment and burn wound management (n=۵), improvingthe accuracy and sensitivity of burn-related complications such as acute kidney injury (n=۵), andpreparing
machine learning algorithms for burn assessment and burn wound management (n=۴).Using laboratory images’ color and texture characteristics makes it possible to classify and identifyburns at different depths (n=۳). A concept of spatial frequency domain imaging to diagnoseburn wounds and decide whether to perform skin grafts has been developed to predict burn spaceand severity (n=۲).Conclusion: Overall, the results of this study show that
machine learning could be considereda potentially new and promising technology for the management of burn wounds in the future.However, there is a lack of evidence to support this claim. Accordingly, it is recommended thatfuture researchers design good studies that evaluate the role of
machine learning in the accurateassessment of patients and diagnostic and therapeutic measures for patients with burn woundswithin future studies.