Radiomics analysis of contrast-enhanced CT predicts lymphovascular invasion and disease outcome in gastric cancer: a preliminary study

Table 2 Selected radiomics features in A, V and A + V models

Model	Selected features (Total features)	Individual features	Coefficients
A	4 (180)	Intercept	−0.269
		A_Max Intensity	−0.581
		A_LongRunLowGrayLevelEmphasis_angle135_offset1	−0.440
		A_Maximum3DDiameter	0.236
		A_SurfaceVolumeRatio	−0.648
V	8 (180)	Intercept	−0.335
		V_Percentile5	4.855
		V_Quantile0.5	−0.838
		V_Quantile0.975	0.215
		V_StdDeviation	1.925
		V_Uniformity	0.599
		V_LongRunHighGrayLevelEmphasis_angle0_offset1	−0.287
		V_ShortRunEmphasis_angle45_offset1	0.110
		V_IntensityVariability	0.316
A + V	7 (360)	Intercept	−0.257
		A_MaxIntensity	−0.566
		A_Quantile0.975	−0.206
		A_LongRunLowGrayLevelEmphasis_angle135_offset1	−0.412
		A_Maximum3DDiameter	0.213
		A_SurfaceVolumeRatio	−0.731
		V_Quantile0.975	0.612
		V_LongRunHighGrayLevelEmphasis_angle0_offset1	−0.154

Note: A, V and A + V indicate the predicted model based on arterial phase images, venous phase images and the combination of arterial phase and venous phase images, respectively. The prefix “A_” and “V_” represent the features from arterial phase images and venous phase images. For the suffix of “_angle(n)_offset(m)”, n means the rotation angle (i.e., 0 or 45) and m represent the displacement vectors (distance to the neighbor pixel, i.e., 7)

ISSN: 1470-7330