Table 7. Performance comparison on LUNA 2016 dataset.

Pretext tasks indicated with an asterisk (*) are reproduced by the same author. Pretext tasks indicated with two asterisks (**) are implemented using different backbones.

No.	Author	Pretext task	Category	Random init.: AUC	SSL: AUC
1	Zhu et al. (2020b)	TCPC ∗∗	Contrastive	0.982	0.996
2	Zhu et al. (2020b)	TCPC ∗∗	Contrastive	0.911	0.987
3	Haghighi et al. (2020)	Semantic Genesis	Multi-tasking	0.943	0.985
4	Zhou et al. (2019)	Models Genesis	Generative	0.942	0.982
5	Haghighi et al. (2020)	Rubik Cube*	Predictive	0.943	0.955
6	Haghighi et al. (2020)	Context Restoration*	Generative	0.943	0.919
7	Haghighi et al. (2020)	Image Inpainting*	Generative	0.943	0.915
8	Haghighi et al. (2020)	Auto-encoder*	Generative	0.943	0.884
9	Tajbakhsh et al. (2019)	3D patch Reconstruct.	Generative	0.724	0.739