Table 2.

Deep learning algorithms applied to the LIDC-IDRI database.

Author	Year	Malignant	Benign	Accuracy (%)	Sensitivity (%)	Specificity (%)	AUC	Noduli Type	Architecture
Chen et al. [26]	2018	NA	NA	NA	93.7	NA	NA	All types	CNN
Sun et al. [33]	2017	47576	41372	NA	NA	NA	0.890	All types	CNN
Wang et al. [34]	2017	NA	NA	NA	83.1	NA	NA	All types	CNN
Da Silva et al. [29]	2018	3415	8742	97.6	92.2	98.2	0.955	All types	CNN
Da silva et al. [28]	2017	1413	1830	94.75	94.7	95.1	0.940	All types	CNN
Causey et al. [6]	2018	NA	NA	94.6	94.8	94.3	0.984	All types	CNN
Ramachandran et al. [31]	2018	3300	3300	93.0	89.0	NA	NA	All types	CNN
Zhu et al. [36]	2018	450	554	90.4	NA	NA	NA	All types	CNN
Da Nóbrega et al. [27]	2018	NA	NA	88.4	85.3	NA	0.931	All types	CNN
Song et al. [32]	2017	2311	2265	84.2	84.0	84.3	0.910	All types	CNN
Han et al. [30]	2018	538	622	82.5	96.6	71.4	NA	GGO	CNN
Zhao X. et al. [35]	2018	375	368	82.2	NA	NA	0.877	All types	CNN
Zhang et al. [37]	2017	40800	32000	95.0	93.5	90.2	0.930	> 30 mm	DBN
Xie et al. [39]	2018	648	1324	89.53	84.2	92.0	0.960	All types	DCNN
Li et al. [40]	2016	40772	21720	89.0	87.1	NA	NA	All types	DCNN
Shaffie et al. [42]	2018	NA	NA	91.2	85.0	95.8	0.95	All types	Deep autoencoder
Gruetzemacher et al. [43]	2018	NA	NA	NA	94.2	NA	NA	All types	DNN
Abbas et al. [44]	2017	1300	1300	95.0	94.0	96.0	0.950	All types	DNN
Hamidian et al. [45]	2017	NA	NA	NA	80.0	NA	NA	All types	FCN + CNN
Xie et al. [38]	2018	644	1301	91.6	86.5	94.0	0.95	All types	MV-KBC
Nibali et al. [46]	2017	420	411	89.9	91.1	88.6	NA	All types	ResNet
Naqi et al. [47]	2018	NA	NA	96.9	95.6	97.0	NA	All types	SA + softmax