Cascaded-Recalibrated Multiple Instance Deep Model for Pathologic-Level Lung Cancer Prediction in CT Images

Table 3

Performance of lung cancer prediction (pmal) for the proposed recalibrated MIL model by transferring from each attribute.

Recalibrated MIL	Accuracy	AUC	F1-score	Ranking
Tex ⟶ Pmal	0.846 ± 0.037	0.849 ± 0.027	0.813 ± 0.037	1
Sph ⟶ Pmal	0.848 ± 0.059	0.831 ± 0.080	0.803 ± 0.073	2
Mal ⟶ Pmal	0.837 ± 0.035	0.830 ± 0.021	0.800 ± 0.037	3
Lob  ⟶ Pmal	0.822 ± 0.065	0.818 ± 0.050	0.797 ± 0.058	4
Spi ⟶ Pmal	0.838 ± 0.061	0.811 ± 0.056	0.796 ± 0.075	5
Mar ⟶ Pmal	0.813 ± 0.061	0.819 ± 0.038	0.789 ± 0.056	6
Cal ⟶ Pmal	0.803 ± 0.044	0.796 ± 0.032	0.768 ± 0.038	7
Sub ⟶ Pmal	0.802 ± 0.048	0.775 ± 0.086	0.767 ± 0.057	8
Int ⟶ Pmal	0.794 ± 0.076	0.704 ± 0.079	0.743 ± 0.064	9

The bold values mean the superiority to others.