TABLE 2.

Binary classification results on test set*.

Performance	1CTR (RF)	2PETR (RF)	3CPR (SMO)
(a) Adc vs. NAdc
Accuracy	81.6	85.0	100.0
True positive rate	0.81	0.89	1.00
True negative rate	0.82	0.80	1.00
Mean of precision	0.88	0.85	1.00
AUC	0.90	0.95	1.00
Performance	CTR	PETR	CPR
(b) Sqc vs. NSqc
Accuracy	76.3	83.4	98.5
True positive rate	0.94	0.90	0.97
True negative rate	0.57	0.77	1.00
Mean of precision	0.80	0.84	0.99
AUC	0.89	0.94	0.99
(c) Primary vs. Mts
Accuracy	86.6	80.9	98.0
True positive rate	0.96	0.92	0.96
True negative rate	0.75	0.69	1.00
Mean of precision	0.88	0.82	0.98
AUC	0.98	0.94	0.98

*Adc, Adenocarcinoma, Sqc, squamous cell carcinoma, Mts, metastases, NAdc, not adenocarcinoma, NSqc, not squamous cell carcinoma. ¹CT-based radiomics (with Random Forest classification). ²PET-based radiomics (with Random Forest classification). ³The combination of CT- and PET-based radiomics (with Sequential minimal optimization classification). The best performance metrics for each classification are highlighted in bold.