Improving Surgical Security using Artificial Intelligence and Deep Learning

The safety of the patient in the different stages of a surgical procedure is an important aspect as concerns risks that need to be catered for as well to avert untoward effects. This research considers Artificial Intelligence and Machine Learning technologies as a core mechanism that can be used to augment the surgical safety systems already in place that determine surgical patients’ risk of death while hospitalized and to also ameliorate pre-operative safety checks. The first approach focuses on traditional checklists by proposing an automated voice validation system of checklist items that uses the Gemini Application Programming Interface (API) on python. This voice validation seeks to mitigate the coverage of human error in relation to the notice of the absence or overweight of the required documents. The second approach describes how machine learning algorithms are used to model a surgical risk predictor across wide bases of patients’ directories, past medical history, and lab tests. Categorical values were made engineering values, and patient were standardized of age with focus on elective surgeries only since emergency surgeries are excluded. For missing values, mean and median imputation was used and PCA for dimensionality reduction. To address the class imbalance issue, the Synthetic Minority Over-sampling Technique (SMOTE) was employed. For the fast and effective binary outcome classification, Support Vector Machine Classification (SVC) with hyperparameters optimized through the GridSearchCV. The concomitant use of speech recognition for checklist verification and ML methods to assess postoperative risk can significantly improve safety during surgical procedures, making it possible to entirely change the scope of preoperative procedures and patient organizational aspects of hospitals.