Table 7.

Performance of network architectures with and without transfer learning applied to the NTUH_ROCF^a data set.

Metric	Sketch-a-Net: with transfer learning, mean (SD)	Our proposed model, mean (SD)
		Without transfer learning	With transfer learning
Sensitivity	0.641 (0.040)	0.749 (0.030)	0.847 (0.017)
Specificity	0.810 (0.022)	0.814 (0.017)	0.905 (0.009)
Accuracy	0.725 (0.010)	0.781 (0.014)	0.876 (0.010)
AUROCb	0.819 (0.010)	0.846 (0.005)	0.913 (0.004)

^aNTUH_ROCF: National Taiwan University Hospital_Rey-Osterrieth Complex Figure.

^bAUROC: area under the receiver operating characteristic curve.