Predicting the Grade of Prostate Cancer Based on a Biparametric MRI Radiomics Signature

Table 4

Discrimination for differentiating low-grade and high-grade PCa in the training and test group and PI-RADS V2.1 scores.

Model	Machine learning	AUC	Accuracy	Sensitivity	Specificity	LR+	LR−
T2WI combine ADC (training)	RF	0.982	0.857	0.92	0.714	3.2168	0.112
	Logistic regression	0.886	0.847	0.968	0.538	2.0952	0.0595
	SVM	0.943	0.857	0.957	0.621	2.5251	0.0692
T2WI combine ADC (test)	RF	0.918	0.795	0.935	0.462	1.7379	0.141
	Logistic regression	0.886	0.841	0.968	0.538	2.0952	0.06
	SVM	0.913	0.841	0.935	0.615	2.4286	0.106
PI-RADS V2.1	Reader 1	0.767	0.725	0.703	0.78	3.1955	0.381
	Reader 2	0.813	0.76	0.713	0.878	5.8443	0.327

ADC, apparent diffusion coefficient; T2WI, T2-weighting imaging; AUC, area under the curve; RF, random forest tree; SVM, support vector machine; LR+, positive likelihood ratio; LR−, negative likelihood ratio; PI-RADS V2.1, prostate imaging reporting and data system version 2.1.