Skip to main content
. 2021 Feb;9(4):298. doi: 10.21037/atm-20-4076

Table 2. Comparison of the proposed method with the models based on different methodologies and modalities.

Method Validation AUC Sensitivity (%) Specificity (%) Testing AUC Sensitivity (%) Specificity (%)
R_T2 FLAIR 0.88 (0.75, 0.95) 76 (16/21) 85 (22/26) 0.81 (0.77, 0.85) 80 (84/105) 71 (163/229)
R_T1ce 0.87 (0.74, 0.95) 81 (17/21) 77 (20/26) 0.80 (0.75, 0.84) 71 (75/105) 79 (180/229)
R_T2 FLAIR + T1ce 0.92 (0.80, 0.98) 86 (18/21) 85 (22/26) 0.85 (0.81, 0.89) 85 (89/105) 76 (175/229)
D_T2 FLAIR 0.86 (0.73, 0.95) 81 (17/21) 77 (20/26) 0.79 (0.75, 0.84) 75 (79/105) 76 (175/229)
D_T1ce 0.88 (0.75, 0.95) 71 (15/21) 88 (23/26) 0.80 (0.76, 0.84) 80 (84/105) 69 (158/229)
D_T2 FLAIR + T1ce 0.91 (0.79, 0.97) 81 (17/21) 88 (23/26) 0.84 (0.80, 0.88) 82 (86/105) 77 (176/229)
R + D_ T2 FLAIR 0.90 (0.77, 0.97) 81 (17/21) 85 (22/26) 0.82 (0.78, 0.86) 78 (82/105) 76 (173/229)
R + D_T1ce 0.90 (0.80, 0.98) 86 (18/21) 81 (21/26) 0.82 (0.77, 0.86) 81 (85/105) 75 (172/229)
Proposed method 0.94 (0.85, 0.99) 86 (18/21) 92 (24/26) 0.88 (0.85, 0.92) 88 (92/105) 81 (186/229)

Note. R_: Radiomics, D_: Deep learning, R + D_: Radiomics + Deep learning