Skip to main content
. 2022 Apr 13;12:6193. doi: 10.1038/s41598-022-10136-9

Figure 5.

Figure 5

Testing performances of the developed models. The model using the deep imaging features performed worse on the testing set than on the validation set (0.70 compared to 0.88), pointing to some overfitting. In contrast, the mortality score model similarly (0.66 compared to 0.62). Combining the imaging features with the mortality score yielded a slightly higher AUC of 0.74, using the same SVM classifier. There are 145 positive samples (survival) and 39 negative samples (Deaths) in the testing set.