Fig. 3: ROC curves for MIL models trained using different strategies. | npj Precision Oncology

Fig. 3: ROC curves for MIL models trained using different strategies.

From: Annotation-free deep learning for predicting gene mutations from whole slide images of acute myeloid leukemia

Fig. 3

a Models trained with NPM1 mutations. b Models trained with FLT3-ITD. “Non-upsampling patch-level MIL" refers to patch-level MIL without upsampling and ensemble learning, “non-upsampling cell-level MIL" refers to cell-level MIL without upsampling and ensemble learning, “upsampling cell-level MIL" means cell-level MIL with upsampling bags but without ensemble learning and “ensemble + upsampling cell-level MIL" refers to cell-level MIL with both upsampling and ensemble learning.

Back to article page