Predicting malignant risk of ground-glass nodules using convolutional neural networks based on dual-time-point 18F-FDG PET/CT

Table 1 Clinical characteristics for training and testing sets

		Training set (n = 239, m = 318)	Testing set (n = 72, m = 798)	p
Gender, n	Male	84	29
	Female	155	43	0.56
Age (years, mean)		60.87	59.07	0.17
Smoking history, n	Yes	48	16
	No	191	56	0.82
Lung cancer family history, n	Yes	41	15
	No	198	57	0.59
GGOs risk Stratification	High risk	226	53
	Low risk	92	26	0.58
GGOs Position	Left upper lobe	84	18
	Left lower lobe	41	6
	Right upper lobe	110	29
	Right middle lobe	26	7
	Right lower lobe	57	19	0.53
Early-phase PET SUV_max (kBq/ml/MBq/kg)		2.42	1.99	0.13
Delayed-phase PET SUV_max (kBq/ml/MBq/kg)		3.03	2.46	0.17
GGOs diameter		17.88	16.67	0.20

ISSN: 1470-7330