TITLE:
Research Progress and Challenges of Machine Learning Algorithms in the Prediction Model of Perioperative Nausea and Vomiting
AUTHORS:
Lantian Li, Haidong Zhou, Jihua Wei, Jinghao Huang, Dinggui Lu
KEYWORDS:
Machine Learning, Prediction, Perioperative Nausea and Vomiting
JOURNAL NAME:
Journal of Biosciences and Medicines,
Vol.13 No.11,
November
28,
2025
ABSTRACT: Postoperative nausea and vomiting (PONV) is a common complication after anesthesia and surgery. Traditional predictive models, such as Apfel scores, rely on linear assumptions and limited risk factors, and predictive efficacy is difficult to meet the needs of precision medicine. Machine learning (ML) significantly improves prediction accuracy through high-dimensional data processing and nonlinear modeling and reveals new risk factors. Supervised learning (random forest, support vector machine), ensemble learning (XGBoost, LightGBM) and deep learning have performed outstandingly in dynamic prediction and multimodal data fusion, improving the accuracy of risk warning. However, ML clinical transformation faces data heterogeneity, model interpretability controversy, and ethical compliance challenges. Federated learning, interpretability tools and causal inference frameworks (Bayesian networks) have become the key to breaking the deadlock. In the future, we need to promote the integration of multi-omics data, real-time biosensing technology and “enhanced intelligence” model, which refers to a human-AI collaborative system that continuously learns from clinician feedback and real-time patient data to adaptively refine predictions and interventions to achieve closed-loop management from prediction to prevention. This paper systematically reviews the algorithm innovation and clinical verification progress of ML in PONV prediction, providing theoretical and practical references for precise treatment and postoperative management.