Comparing supervised machine learning algorithms for the prediction of partial arterial pressure of oxygen during craniotomy

Objective To evaluate the feasibility of continuous paO 2 prediction in an intraoperative setting among neurosurgical patients with modern machine learning methods. Materials and Methods Data were extracted from routine clinical care of lung-healthy, neurosurgical patients. We used recursive feature elimination to identify relevant features for the prediction of paO 2 . Five machine learning algorithms (gradient boosting regressor, k-nearest neighbors regressor, random forest regressor, support vector regression, multi-layer perceptron regressor) and a multivariate linear regression were then tuned and fitted to the selected features. A performance matrix consisting of Spearman’s ρ, mean absolute percentage error (MAPE) and root mean squared error (RMSE) was finally computed based on the test set and used to compare and rank each algorithm. Results We analyzed 4,180 patients with 12,497 observations. A total of 20 features were selected from analysis of the training dataset comprising 836 patients with 9,992 observations. The best algorithm, random forest, was able to predict paO 2 values with ρ=0.90, MAPE=10.4%, and RMSE=30.9mmHg, closely followed by gradient boosting and multi-layer perceptron. Support vector regression, k-nearest neighbors regressor and the linear regression did not achieve the performance metrics. Discussion We successfully applied and compared several machine learning algorithms to estimate continuous paO 2 values in neurosurgical patients. The random forest regressor performed best over all three categories of the performance matrix. Conclusion PaO 2 can be predicted by perioperative routine data in neurosurgical patients.