Visual Computing for Industry, Biomedicine, and Art

Table 2 Medoid features and feature selection result

From: CT-based radiomics: predicting early outcomes after percutaneous transluminal renal angioplasty in patients with severe atherosclerotic renal artery stenosis

Feature type	Feature name	P value	AUC	Selected by LASSO	Signature input
Renal feature	XH_H_uniformity	0.052	0.673 (0.502–0.843)	√	√
	XL_H_mean_absulute_deviation	0.034	0.675 (0.519–0.830)	√
	XH_GLCM_correlation	0.013	0.722 (0.573–0.870)	√	√
	XL_GLRLM_LRHGLE	0.090	0.651 (0.491–0.811)
	D1_GLCM_correlation	0.096	0.675 (0.500-0.849)	√	√
Perirenal adipose feature	XL_H_uniformity	0.005	0.737 (0.578–0.896)
	X_GLCM_contrast	0.005	0.740 (0.592–0.889)	√
	XL_H_krutosis	0.053	0.665 (0.489–0.841)
	XL_H_standard_deviation	0.005	0.740 (0.583–0.898)
	XH_H_mean	< 0.001	0.815 (0.688–0.941)	√
	XH_GLCM_cluster_shade	< 0.001	0.815 (0.689–0.942)	√	√
	D5_GLCM_homogeneity2	0.006	0.748 (0.595–0.902)	√	√
	XH_GLCM_correlation	0.070	0.618 (0.440–0.795)	√
	D1_H_maximum	0.012	0.714 (0.563–0.865)
	D5_GLCM_entropy	0.026	0.685 (0.526–0.843)
	D1_GLCM_energy	0.004	0.745 (0.585–0.906)	√

Dx Xth deep learning feature map, H Histogram feature, GLCM Gray-level co-occurrence matrix-based feature, GLRLM Gray-level run length matrix feature, LRHGLE Low run high gray-level emphasis

Back to article page