Table 4.

Predictive performance of VOI_P models

*VOI	#AUC	Accuracy	Sensitivity	Specificity	F1
Training and validation sets
VOI_P1	0.821	77.3%	78.3%	76.3%	78.8%
VOI_P2	0.849	78.9%	82.6%	74.6%	80.9%
VOI_P3	0.735	59.4%	27.5%	96.6%	42.2%
VOI_P4	0.835	75.0%	81.2%	67.8%	77.8%
VOI_P5	0.747	57.0%	27.5%	91.5%	40.9%
VOI_P10	0.778	68.8%	62.3%	76.3%	68.3%
VOI_P15	0.861	76.6%	81.2%	71.2%	78.9%
Internal testing set
VOI_P1	0.700	67.2%	79.7%	52.5%	72.4%
VOI_P2	0.687	61.7%	72.5%	49.2%	67.1%
VOI_P3	0.689	52.3%	21.7%	88.1%	33.0%
VOI_P4	0.676	68.0%	79.7%	54.2%	72.8%
VOI_P5	0.664	50.0%	17.4%	88.1%	27.3%
VOI_P10	0.703	60.9%	55.1%	67.8%	60.3%
VOI_P15	0.716	66.4%	75.4%	55.9%	70.7%
External testing set
VOI_P1	0.655	59.0%	73.2%	49.4%	59.0%
VOI_P2	0.686	56.8%	78.6%	42.2%	59.5%
VOI_P3	0.606	59.7%	7.1%	95.2%	12.5%
VOI_P4	0.635	59.0%	75.0%	48.2%	59.6%
VOI_P5	0.609	59.0%	1.8%	97.6%	3.4%
VOI_P10	0.601	63.3%	48.2%	73.5%	51.4%
VOI_P15	0.704	61.9%	78.6%	50.6%	62.4%

^*VOI—volume of interest

^#AUC—area under the curve