The Effect of Preoperative Medication on Intraoperative Neuromonitoring Signals During Spine Surgery: A Machine Learning Analysis

Introduction: Machine learning (ML) is a subset of artificial intelligence (AI) that is being increasingly applied in spine surgery research. The application of ML to intraoperative neuromonitoring (IONM) has the potential to develop automated signal monitoring which could improve patient safety. Due to the multitude of variables that can alter IONM signals, specialized training is required to interpret IONM signals. Pharmacological agents are one of these variables whose effects on IONM signals has yet to be fully explored by ML. This study utilizes ML to analyze the effects of preoperative medications on IONM signals to enhance the predictive accuracy of IONM for improved intraoperative decision-making.

Methods: A retrospective review of spine surgery cases with IONM data was conducted. Preoperative opioids, anxiolytics, antidepressants, and gabapentinoids prescribed within 90 days of surgery were obtained from the local state prescription drug monitoring program. IONM signal characteristics were used to identify relevant features for ML algorithms. ML models were assessed for their ability to predict the effect of preoperative medication on IONM signal integrity.

Results: ML analysis demonstrated limited predictive capacity for baseline IONM changes with preoperative opioid data alone, with a classification accuracy of 64.4%. Inclusion of additional demographic, comorbidity, and medication variables may enhance model performance. Future directions include the analysis of preoperative anxiolytics, antidepressants, gabapentinoids in order to identify their possible effects on IONM signals.

Conclusion : This study explores the potential of ML to analyze the effect of preoperative medications on IONM signals. Although the initial results indicate limited predictive capacity, the incorporation of larger data sets and more variables may allow for more thorough analysis which may pave the way for more robust models. Future research expanding ML analysis to comprehensive preoperative medications, comorbidities, demographics, and other past medical history data may aid in to the creation of ML algorithms capable of supporting real-time IONM analysis.