An artificial intelligence algorithm for the detection of pulmonary ground-glass nodules on spectral detector CT: performance on virtual monochromatic images

Table 3 ROC curve analysis for identifying IAC from MIA in convention group and VMIs group

		AUC	95% CI	P value	Sensitivity	Specificity	Youden index	Cutoff value	DeLong test (vs. Convention)
Total mass	Convention	0.862	0.790–0.926	0.000	82.60%	75.00%	0.576	120.65	—
	80	0.871	0.798–0.932	0.000	82.60%	77.30%	0.599	114.90	0.11
	70	0.869	0.798–0.932	0.000	82.60%	79.50%	0.621	117.35	0.24
	60	0.867	0.796–0.930	0.000	76.10%	84.10%	0.602	128.35	0.32
	50	0.866	0.794–0.928	0.000	76.10%	81.80%	0.579	125.85	0.48
	40	0.851	0.776–0.919	0.000	78.30%	77.30%	0.556	117.40	0.24
Total volume
	Convention	0.823	0.736–0.899	0.000	69.60%	81.80%	0.514	359.12	—
	80	0.846	0.766–0.917	0.000	82.60%	72.70%	0.553	276.45	0.01
	70	0.842	0.759–0.914	0.000	82.60%	70.50%	0.531	263.84	0.03
	60	0.843	0.760–0.914	0.000	71.70%	81.80%	0.535	322.91	0.02
	50	0.837	0.752–0.910	0.000	71.70%	81.80%	0.535	325.48	0.13
	40	0.807	0.719–0.888	0.000	71.70%	77.30%	0.49	312.16	0.23

MIA Minimally invasive adenocarcinoma, IAC Invasive adenocarcinoma, VMIs Virtual monochromatic images, ROC Receiver operating characteristic, AUC Area under curve, CI Confidence interval

ISSN: 1471-2342