Skip to main content
. 2023 Jan 17;29(4):1024–1033. doi: 10.1111/cns.14071

FIGURE 3.

FIGURE 3

The OI input in different ML algorithms performed better than other inputs. In the ML models for good outcome, the models built with II alone had significantly lower predictive power for prognosis than the other input combinations. Mann–Whitney U test with Bonferroni correction was used. (A), (C) The AUC of ML models with various combination of input in SVO‐AIS population for excellent and good outcome; (B), (D) The AUC of ML models with various combination of input in LAA‐AIS population for excellent and good outcome. The mean ± 95% CI of the receiver operating characteristic area under the curve (AUC) is displayed as a bar graph. AIS, acute ischemic stroke; GOA‐RF, Grasshopper Optimization Algorithm Random Forest; GOA‐XGBoost, Grasshopper Optimization Algorithm eXtreme Gradient Boosting; GPR, Gaussian Process Regression; II, image index; LAA, large artery atherosclerosis; OI, optimal index; RI, regular index; SVO, small vessel occlusion; TI, total index.