Amide proton transfer-weighted imaging and stretch-exponential model DWI based 18F-FDG PET/MRI for differentiation of benign and malignant solitary pulmonary lesions

Table 5 Predictive performance for identifying benign and malignant SPL

Parameters	AUC (95% CI)	P-value	Cutoff	Sensitivity	Specificity	Comparison with a combined diagnosis
Training set
SUV_max	0.853 (0.759 ~ 0.921)	< 0.001	6.690	73.21%	96.43%	Z = 3.080, P = 0.002
MTV (ml)	0.768 (0.663 ~ 0.853)	< 0.001	6.930	67.86%	78.57%	Z = 4.030, P < 0.001
TLG (g)	0.809 (0.709 ~ 0.887)	< 0.001	16.653	73.21%	82.14%	Z = 3.653, P < 0.001
ADC (× 10⁻³mm²/s)	0.888 (0.800 ~ 0.946)	< 0.001	1.573	76.79%	92.86%	Z = 2.761, P = 0.006
DDC (× 10⁻³mm²/s)	0.786 (0.683 ~ 0.868)	< 0.001	2.721	78.57%	85.71%	Z = 3.844, P < 0.001
α	0.786 (0.683 ~ 0.868)	< 0.001	0.534	67.86%	89.29%	Z = 3.773, P < 0.001
MTRasym(3.5 ppm) (%)	0.691 (0.581 ~ 0.788)	0.002	2.225	55.36%	89.29%	Z = 4.938, P < 0.001
Prediction model	0.976 (0.916—0.997)	< 0.001	/	94.64%	92.86%	/
Test set
SUV_max	0.885 (0.734 ~ 0.967)	< 0.001	3.330	86.96%	84.62%	Z = 1.741, P = 0.081
MTV (ml)	0.793 (0.625 ~ 0.909)	< 0.001	8.395	69.57%	84.62%	Z = 2.337, P = 0.020
TLG (g)	0.833 (0.671 ~ 0.936)	< 0.001	17.924	69.57%	92.31%	Z = 2.317, P = 0.021
ADC (× 10⁻³mm²/s)	0.756 (0.584 ~ 0.883)	0.004	1.324	60.87%	92.31%	Z = 2.263, P = 0.024
DDC (× 10⁻³mm²/s)	0.729 (0.555 ~ 0.863)	0.006	1.824	52.17%	92.31%	Z = 2.697, P = 0.007
α	0.776 (0.606 ~ 0.898)	< 0.001	0.485	86.96%	61.54%	Z = 2.514, P = 0.012
MTRasym(3.5 ppm) (%)	0.732 (0.559 ~ 0.866)	0.007	1.880	65.22%	100.00%	Z = 2.562, P = 0.010
Prediction model	0.957 (0.831—0.997)	< 0.001	/	91.30%	92.31%	/

SPL solitary pulmonary lesion, SUV_max Maximum standardized uptake value, MTV Metabolic tumor volume, TLG Total lesion glycolysis, ADC Apparent diffusion coefficient, DDC Distributed diffusion coefficient, α diffusion heterogeneity index, MTRasym(3.5 ppm) magnetization transfer ratio asymmetry at 3.5 ppm. The prediction model represents SUV_max. + ADC + MTRasym (3.5 ppm)

ISSN: 1470-7330