Feasibility of Using High-Resolution Computed Tomography Features for Invasiveness Differentiation of Malignant Nodules Manifesting as Ground-Glass Nodules

Table 1

Patient demographic information comparison between the training and testing cohorts and semantic imaging feature comparison between the invasiveness subtypes and differentiation ability in terms of AUC.

Patient demographical information
	Training			Testing			value
Age	54.9 (±10.80)			55.6 (±10.89)			0.69
Gender							0.24
Female	155 (70.1%)			43 (78.2%)
Male	66 (29.9%)			12 (21.8%)
GGO characteristic
	Training		value	AUC	Testing		value	AUC
	Pre-MIA	Invasive-			Pre-MIA	Invasive-
Nodule type			<0.0001	0.67			0.0009	0.67
Pure	102 (98.08%)	75 (64.1%)			26 (100%)	19 (65.52%)
Part-solid	2 (1.92%)	42 (35.90%)			0 (0%)	10 (34.48%)
Lesion margin			0.02	0.78			<0.0001	0.75
Smooth	74	18			24	12
Lobular or spiculated	30	99			2	17
Air bronchial sign present	3 (2.88%)	8 (6.84%)	0.18		2 (7.69%)	4 (13.79%)	0.67
Pleural indentationpresent	2 (1.92%)	51(43.52%)	<0.0001	0.71	1(3.85%)	11 (37.93%)	0.0027	0.67
Vascular convergence present	1 (0.96%)	32 (27.35%)	<0.0001		0(0%)	4 (13.79%)	0.11
DGGN	10.01 (±4.65)	15.38 (±7.22)	<0.01		18.11 (±8.80)	9.32 (±2.45)	<0.0001
GGN<10 mm	62	11			15	3
GGN>10 mm	42	106			11	26

.