Machine Learning Models for Predicting Facial Nerve Palsy in Parotid Gland Surgery for Benign Tumors

BackgroundDespite the increasing use of intraoperative facial nerve monitoring during parotid gland surgery (PGS) and improvement in preoperative radiological assessment, injury (FNI) remains most severe complication after PGS. Until now, no studies have been published regarding application machine learning (ML) for predicting FNI We hypothesize that ML would improve prediction patients at risk.MethodsPatients who underwent PGS benign tumors between June 2010 2019 were included.ResultsRegarding accuracy performance each algorithm, K-nearest neighbor random forest achieved highest sensitivity, specificity, positive predictive value, negative value F-score, receiver operating characteristic (ROC)–area under ROC curve, globally. The algorithm values above 0.9 F-score ROC–area sensitivity value.ConclusionsThis study demonstrates models can provide evidence-based predictions about risk to otolaryngologists patients. It is hoped such algorithms, which clinical, radiological, histological, cytological information, information given before so they be better informed any potential complications.