Fig. 5: Sensitivity analysis of CMR-based SISSO & XGBoost models. | npj Digital Medicine

Fig. 5: Sensitivity analysis of CMR-based SISSO & XGBoost models.

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

Fig. 5

a Sensitivity analysis results from SISSO. Each feature corresponds to four bars, from left to right: the Sobol main effect index Si, the total effect index \({S}_{i}^{{\rm{T}}}\), SHAP, and LIME. The left y-axis represents the scale of the Sobol indices, while the right y-axis corresponds to the mean absolute values of the SHAP and LIME indices. b Sensitivity analysis results from XGBoost. The blue bars indicate the mean absolute SHAP values for each feature, while the yellow bars represent the mean absolute LIME values. Here, {x1, x2, …, x8} correspond sequentially to the features listed in Table 4.

Back to article page