Skip to main content
. Author manuscript; available in PMC: 2019 Nov 29.
Published in final edited form as: IEEE Access. 2018 Nov 29;6:77796–77806. doi: 10.1109/ACCESS.2018.2884126

TABLE 8:

Classifiers, features used, and performance statistics for best AUC of risk prediction model (experiment 2) using RIDER features

Experiment Features Set Classifier Feature Selector Features Selected Performanc Statistics
AUC
(95% CI)
Specificity
(95% CI)
Sensitivity
(95% CI)
Train C1T1 &
Test C2T1
Rider features Random Forests ReliefF (15) Mean [HU], Roundness, StdDev [HU], 8a_3D_Is_Attached_To_Pleural_Wall,
3D Laws features L5 W5 L5 Layer 1, 9b_3D_Circularity, Asymmetry, 1Longest Diameter [mm], Short Axis [mm], 9g_3D_MAX_Dist_COG_To_Border_[mm],
3D Laws features E5 E5 R5 Layer 1,
8c_3D_Relative_Border_To_PleuralWall,
8b_3D_Relative_Border_To_Lung,
3D Laws features E5 W5 L5 Layer 1, 9e_3D_SD_Dist_COG_To_Border_[mm]
0.773
(0.74-0.81)
0.89
(0.84-0.93)
0.56
(0.46-0.65)
Train on C1T1 and C1 delta features &
Test on C2T1 and C2 delta features
(i.e., C2T1 - C2T0)
Rider features Random Forests ReliefF (20) Mean [HU],Roundness, 9b_3D_Circularity,
Longest Diameter [mm], Short Axis [mm],
9g_3D_MAX_Dist_COG_To_Border_[mm],
8a_3D_Is_Attached_To_Pleural_Wall,
3D Laws features L5 W5 L5 Layer 1, Asymmetry,
3D Laws features E5 E5 R5 Layer 1,
9e_3D_SD_Dist_COG_To_Border_[mm],
8c_3D_Relative_Border_To_PleuralWall,
8b_3D_Relative_Border_To_Lung,
Short Axis * Longest Diameter [mm_],
9g 3D MAX Dist COG To Border [mm] Delta,
Asymmetry_Delta, ShortAxis[mm]_Delta,
StdDev[HU]_Delta, Mean[HU]_Delta,
LongestDiameter[mm]_Delta
0.822
(0.79-0.85)
0.93
(0.89-0.96)
0.49
(0.40-0.59)
p value (p <0.05) 0.303