Pulmonary Nodule Detection from 3D CT Image with a Two-Stage Network

Table 4

Detection sensitivity and the CPM score achieved by different methods.

Methods	1/8 FPs/scan	1/4 FPs/scan	1/2 FPs/scan	1 FPs/scan	2 FPs/scan	4 FPs/scan	8 FPs/scan	CPM
NoduleNet (2020) [26]	0.732	0.763	0.818	0.875	0.893	0.917	0.928	0.846
S4ND (2019) [14]	0.781	0.822	0.85	0.898	0.928	0.943	0.955	0.882
DBResNet (2020) [13]	0.817	0.852	0.885	0.912	0.936	0.947	0.955	0.900
V-Net (2020) [15]	0.702	0.801	0.844	0.881	0.927	0.939	0.942	0.862
H-DenseUNet (2018) [22]	0.753	0.802	0.832	0.88	0.909	0.935	0.953	0.866
Swin-UNetr (2021) [27]	0.761	0.827	0.867	0.902	0.933	0.949	0.956	0.885
UNetr (2022) [28]	0.726	0.754	0.801	0.853	0.878	0.912	0.937	0.837
Ours	0.833	0.87	0.898	0.921	0.938	0.951	0.96	0.910

The best results are highlighted in bold.