The utility of texture analysis based on quantitative synthetic magnetic resonance imaging in nasopharyngeal carcinoma: a preliminary study

Table 4 ROC curve analysis of three single functional models in discrimination between OC invasion and OC non-invasion

Models	OC invasion	OC non-invasion	P	Model fit	AUC (95% CI)	Sensitivity%	Specificity%	Accuracy%
T1 map derived model				1.000	0.997 (0.990, 1.000)	96.9	97.9	97.5
T1_Mean	1570.61 (1348.53, 1892.24)	651.15 (588.75, 756.02)	0.000
T1_Total Energy (× 10⁶)	3621.43 (1488.39, 7822.86)	1398.09 (1048.77, 2008.46)	0.000
T2 map derived model				0.885	0.913 (0.851, 0.975)	93.8	72.9	80.0
T2_10th percentile	72.85 (61.46, 80.28)	83.06 (75.80, 89.22)	0.000
T2_Energy (× 10⁶)	12.14 (7.51, 32.71)	33.98 (18.90, 54.38)	0.000
T2_MAD	15.96 (11.83, 20.44)	12.79 (10.95, 14.03)	0.002
PD map derived model				0.209	0.975 (0.945, 1.000)	93.7	95.8	93.8
PD_Maximum	107.00 (105.00, 108.15)	133.5 (122.95, 140.05)	0.000

Data was expressed as median (interquartile range)
ROC, receiver operating characteristic; OC, occipital clivus; AUC, area under the curve; 95% CI, 95% confidence interval; MAD, mean absolute deviation

ISSN: 1471-2342