Predicting the Impacts of the Environment and Severity of High-risk Burns by AIML

A burn injury’s effects might vary from mild to potentially fatal. Burn treatment depends on their degree and location. Extreme burns need medical attention, usually at specialized burn centres with significant follow-up treatment, whereas light burns may occasionally be managed at home. The goal of this research is to apply AI/ML to predict potential complications for burn patients. Burn patients who were hospitalized were included in this retrospective analysis. The study built predicted models for graft surgery, hospital stay, and complications using data from a total of 10 variables. These factors included things like the patient’s medical histories and laboratory findings. The AI models were trained using 65% of the information set, while the remaining 35% was used for evaluation. Precision, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) were used to evaluate the three machine learning (ML) techniques of randomized forests, Light GBM, and logistic regression. The results showed that of the three models tested, the model based on random forests had the most excellent AUC (82.2%) for predicting lengthy hospital stays $( \gt15$ days), followed by the XGBoost (80.8%), and the Light GBM (80.6%). Additionally, the most excellent AUC (79.9%) was shown in the randomized forest model for the requirement for a skin transplant, and the most significant area under the curve (88.3%) was seen in both the randomized forest and XGBoost models for the incidence of unfavourable consequences. The best-performing models with the maximum AUC values are used to create and integrate an AI prediction system into healthcare information systems. The use of AI approaches has shown remarkable promise in predicting whether or not a burn patient would need a skin transplant, how long they will be hospitalized, and the likelihood of other, more severe issues. Hope for developing a new prediction model that can integrate with information systems in hospitals, improving clinical choices and reinforcing doctor-patient conversations, is stoked by our study’s findings.