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

Background and Objectives: Brain tissue oxygenation is usually inferredfrom arterial partial pressure of oxygen (paO2), which is in turn often inferredfrom pulse oximetry measurements or other non-invasive proxies. Our aim was toevaluate the feasibility of continuous paO2 prediction in an intraoperative set-ting among neurosurgical patients undergoing craniotomies with modern machinelearning methods. Methods : Data from routine clinical care of lung-healthy neurosurgical patientswere extracted from databases of the respective clinical systems and normalized. We used recursive feature elimination to identify relevant features for theprediction of paO2. Six machine learning regression algorithms (gradient boosting, k-nearest neighbors, random forest, support vector, neural network, linear model with stochastic gradient descent) and a multivariable linear regression were then tuned and fitted to the selected features. A performance matrix consisting of Spearman’s ρ, mean absolute percentage error (MAPE), adjusted R2, mean absolute error (MAE) 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 N=4,175 patients with n=14,495 observations. Between 5 and 23 features were selected from the analysis of the training dataset comprising 3,131 patients with 10,896 observations. The best algorithm, a regularized linear model with stochastic gradient descent, could predict paO2 values with adjusted R2=0.76, Spearman’s ρ=0.83, MAPE=15.2 % and RMSE=41.7 mmHg. Further improvement was possible by calibrating the algorithm with the first measured paO2/F iO2 ratio during surgery. Conclusion : PaO2 can be predicted by perioperative routine data in neurosurgical patients even before blood gas analysis. The prediction improves further when including the first measured paO2/F iO2 ratio, realizing quasi-continuous paO2 monitoring.