Table 4.

Performance of MIL models in EGFR mutation prediction

	Ours	MHIM	DSMIL	ABMIL
AUROC	0.7680 (0.7607–0.7720)	0.7441 (0.7381–0.7508)	0.7210 (0.7170–0.7302)	0.7021 (0.6982–0.7094)
AUPRC	0.8391 (0.8326–0.8430)	0.8098 (0.7983–0.8124)	0.7904 (0.7862–0.8022)	0.7622 (0.7584–0.7712)

Table comparing the performance of multiple‐instance learning (MIL) models developed for EGFR mutation prediction. AUROC and AUPRC values of the proposed model (Ours) and existing methods (MHIM, DSMIL, and ABMIL) are presented.

AUPRC, area under the precision‐recall curve.