Table 2.

Summary of model performance for the different models. Model A consisting of standard clinicopathological variables demonstrated the best performance with the lowest values for AIC and BIC. The addition of CT (Model B, D, D2, and E) and pathology (Model F1–F3) variables to Model A did not improve baseline performance

Model	Log-likelihood	c statistic	R2	Adj R2	D statistic (SE)	Adj D statistic (SE)	Calibration slope	Model degrees freedom	AIC	BIC
A: baseline clinical variables	128.4	0.76 (0.70, 0.82)	0.38	0.37	1.61 (0.26)	1.57 (0.26)	1.27 (0.89, 1.65)	2	260.8	267.5
B: plus local PCT using two PCA scores	128.4	0.77 (0.71, 0.83)	0.38	0.35	1.61 (0.26)	1.51 (0.26)	1.28 (0.89, 1.66)	4	264.8	278.2
D: plus local PCT using two individual measures (BF, BV)	128.3	0.76 (0.70, 0.82)	0.37	0.34	1.57 (0.25)	1.46 (0.26)	1.27 (0.89)	4	264.6	278.1
D2: plus local PCT using four individual measures (BF, BV, PS, and MTT)	128.1	0.77 (0.71, 0.82)	0.39	0.33	1.63 (0.26)	1.45 (0.26)	1.32 (0.92, 1.72)	6	268.2	288.4
E: plus central PCT using two PCA scores	127.7	0.78 (0.72, 0.84)	0.38	0.35	1.60 (0.25)	1.49 (0.25)	1.20 (0.84, 1.56)	4	263.3	276.8
F1: plus four main genes only	122.9	0.77 (0.71, 0.83)	0.47	0.41	1.94 (0.28)	1.72 (0.279)	1.25 (0.91, 1.60)	7	259.7	283.2
F2: plus IHC (CD105, HIF1, and MMR)	127.2	0.78 (0.72, 0.84)	0.41	0.37	1.70 (0.26)	1.55 (0.264)	1.31 (0.92, 1.70)	5	264.3	281.1
F3: plus four main genes and IHC (CD105, HIF1, and MMR)	121.9	0.78 (0.72, 0.84)	0.50	0.41	2.04 (0.30)	1.70 (0.29)	1.26 (0.92, 1.60)	10	263.8	297.3

AIC akaike information criterion, BIC Bayesian information criterion, PCT CT perfusion imaging, PCA principal components analysis, IHC immunohistochemistry, BF blood flow, BV blood volume, PS permeability surface, MTT mean transit time, MMR mismatch repair, HIF hypoxia-inducible factor, SE standard error