Skip to main content
. 2023 Feb 21;13(4):1342–1354. doi: 10.7150/thno.81784

Figure 5.

Figure 5

Comparison between weakly-supervised MIL-CNN and supervised CNN model. Dependences of the model prediction (A) accuracy, (B) precision and (C) recall on the threshold values. (D) ROC curves of the two models. The shades represent the standard deviation from 5-fold cross validation.