Fig. 2: Performance of XGBoost on mPAP via ten-fold cross-validation across four different input feature spaces. | npj Digital Medicine

Fig. 2: Performance of XGBoost on mPAP via ten-fold cross-validation across four different input feature spaces.

From: Artificial intelligence-driven multivariate integration for pulmonary arterial pressure prediction in pulmonary hypertension

Fig. 2

a The original complete dataset including echocardiography features; b Dataset without echocardiography features; c Dataset retaining the top 8 features selected by MI; d Dataset with RWMaxWSS and LWMaxWSS (from the MI feature set) removed and BMI and UA manually added. Here, \({\mathcal{P}}\) denotes the true values of mPAP, and \(\hat{{\mathcal{P}}}\) represents the model estimates. Scatter points correspond to the mean predicted values from ten-fold CV on the y-axis scale, and error bars represent the variability of the predictions.

Back to article page