Skip to main content
. 2020 Oct 2;10:558162. doi: 10.3389/fonc.2020.558162

TABLE 3.

Summary of the subgroup-specific classification performance of the radiomics model.

Molecular subgroups Cohorts Area under the curve ACC (%) SEN (%) SPE (%)
Wingless Training 0.7576 (0.6113–0.9039) 79.35 (52.15–90.22) 66.67 (46.67–100.00) 81.82 (44.16–97.40)
Testing 0.8264 (0.5597–1) 86.67 (76.67–100.00) 50.00 (50.00–100.00) 95.83 (75.00–100.00)
Sonic hedgehog Training 0.8014 (0.6598–0.943) 91.30 (65.22–95.65) 62.50 (43.75–93.75) 97.37 (59.21–100.00)
Testing 0.6683 (0.361–0.9755) 83.33 (33.33–93.33) 75.00 (50.00–100.00) 80.77 (23.08–100.00)
Group 3 Training 0.8995 (0.8325–0.9665) 83.70 (78.26–91.30) 82.50 (62.50–95.06) 84.62 (69.23–100.00)
Testing 0.6004 (0.3887–0.8122) 60.00 (56.67–80.00) 78.57 (72.86–100.00) 43.75 (18.75–93.75)
Group 4 Training 0.9188 (0.8368–1) 93.48 (79.35–97.83) 76.19 (66.67–100.00) 98.59 (77.46–100.00)
Testing 0.6979 (0.4042–0.9916) 76.67 (60.00–90.00) 83.33 (50.00–100.00) 75.00 (54.17–95.83)

ACC, SEN, and SPE are short for accuracy, sensitivity, and specificity, respectively. The 95% confidence interval for each index is shown.