Table 5.

Harmonized Feature RSF Model Performance. Mean results for performance metrics with associated bootstrap percentile 95% confidence intervals are displayed for deep learning and radiomics features with manually or automatically segmented masks. C-index: concordance index; IBS: integrated brier score.

Modality	C-Index	IBS	Modality	C-Index		IBS
RSF model with deep learning features (Manual masks)			RSF model with deep learning features (Automatic masks)
CT	0.649 (0.596, 0.708)	0.207 (0.189, 0.238)	CT	0.748 (0.700, 0.801)		0.168 (0.139, 0.200)
PET	0.716 (0.657, 0.776)	0.175 (0.138, 0.203)	PET	0.717 (0.671, 0.768)		0.176 (0.150, 0.210)
Clinical	0.591 (0.525, 0.647)	0.175 (0.138, 0.203)	Clinical	0.590 (0.538, 0.663)		0.176 (0.150, 0.21)
PET + CT Ensemble	0.711 (0.661, 0.767)	0.179 (0.154, 0.207)	PET + CT Ensemble	0.729 (0.682, 0.774)		0.169 (0.141, 0.20)
PET + CT + Clinical Ensemble	0.709 (0.655, 0.774)	0.176 (0.148, 0.207)	PET + CT + Clinical Ensemble	0.727 (0.679, 0.774)		0.169 (0.142, 0.201)
RSF model with radiomics features (Manual masks)			RSF model with radiomics features (Automatic masks)
CT	0.738 (0.677, 0.796)	0.183 (0.162, 0.217)	CT	0.680 (0.638, 0.722)		0.196 (0.173, 0.223)
PET	0.629 (0.581, 0.682)	0.225 (0.194, 0.273)	PET	0.653 (0.602, 0.704)		0.204 (0.173, 0.235)
Clinical	0.600 (0.541, 0.673)	0.225 (0.194, 0.273)	Clinical	0.589 (0.534, 0.647)		0.204 (0.173, 0.235)
PET + CT Ensemble	0.680 (0.619, 0.743)	0.197 (0.174, 0.235)	PET + CT Ensemble	0.664 (0.616, 0.705)		0.195 (0.167, 0.223)
PET + CT + Clinical Ensemble	0.676 (0.611, 0.733)	0.187 (0.162, 0.228)	PET + CT + Clinical Ensemble	0.659 (0.613, 0.712)		0.186 (0.159, 0.214)