Table 4.

Performance of different deep neural network (DNN) models on the T1^a-T2^b-T1c^c-fused images for image-based classification.

DNN models	Sensitivity (95% CI)	Specificity (95% CI)	Accuracy (95% CI)	Area under the curve (95% CI)
VGGd16	0.858 (0.834-0.880)	0.826 (0.791-0.858)	0.847 (0.828-0.865)	0.864 (0.842-0.886)
VGG19	0.852 (0.828-0.874)	0.704 (0.662-0.744)	0.801 (0.780-0.821)	0.828 (0.804-0.852)
ResNete-50	0.899 (0.879-0.918)	0.663 (0.620-0.704)	0.818 (0.797-0.837)	0.866 (0.844-0.888)
Inception-v3	0.844 (0.819-0.866)	0.716 (0.675-0.755)	0.800 (0.778-0.820)	0.845 (0.822-0.868)
Inception-ResNet-v2	0.925 (0.907-0.941)	0.755 (0.716-0.792)	0.867 (0.848-0.884)	0.913 (0.895-0.931)
ERN-Netf	0.820 (0.794-0.844)	0.789 (0.751-0.824)	0.809 (0.788-0.829)	0.915 (0.895-0.932)

^aT1: T1-weighted magnetic resonance imaging (MRI).

^bT2: T2-weighted MRI.

^cT1c: gadolinium-contrast-enhanced T1-weighted MRI.

^dVGG: Visual Geometry Group.

^eResNet: residual neural network.

^fERN-Net: efficient radionecrosis neural network.