Table 3 Predictive performance of the five Models.

Model	AUC	Sensitivity	Specificity	PPV	NPV	Accuracy	p-value
Training Set
Clinical Model	0.712 (0.612 ~ 0.812)	0.789	0.636	0.517	0.86	68.7	< 0.001*
Radiomics Model	0.752 (0.658 ~ 0.846)	0.658	0.753	0.568	0.817	72.17	< 0.001*
Combined model
LR Algorithm	0.825 (0.740 ~ 0.911)	0.789	0.779	0.638	0.882	78.26	< 0.001*
RF Algorithm	0.823 (0.744 ~ 0.902)	0.842	0.688	0.571	0.898	73.91	< 0.001*
SVM Algorithm	0.841 (0.758 ~ 0.925)	0.842	0.805	0.681	0.912	81.74	< 0.001*
Validation Set
Clinical Model	0.731 (0.574 ~ 0.889)	0.813	0.656	0.542	0.875	70.83	0.004*
Radiomics Model	0.666 (0.498 ~ 0.834)	0.625	0.688	0.5	0.786	66.67	0.052
Combined Model
LR Algorithm	0.828 (0.706 ~ 0.950)	0.813	0.813	0.684	0.897	81.25	< 0.001*
RF Algorithm	0.587 (0.409 ~ 0.765)	0.313	0.906	0.625	0.725	70.83	0.038*
SVM Algorithm	0.510 (0.336 ~ 0.684)	0.813	0.375	0.394	0.8	52.08	0.912

AUC area under the ROC curve, RF Random Forest, SVM Support Vector Machine. *Indicates statistical significance. PPV Positive Predictive Value, NPV Negative Predictive Value.

Quick links

Search