Prediction of the clinicopathological subtypes of breast cancer using a fisher discriminant analysis model based on radiomic features of diffusion-weighted MRI

Table 5 Comparison of ROC analysis results

IHC status	AUROC	Z statistic	p value
ER(+) VS. ER(−)	Histogram VS. Texture	4.531	< 0.001
	Histogram VS. GLCM	0.462	0.644
	Histogram VS. RLM	0.312	0.7548
	Texture VS. GLCM	−4.147	< 0.001
	Texture VS. RLM	−4.322	< 0.001
	GLCM VS. RLM	−0.171	0.864
PR(+) VS. PR(−)	Histogram VS. Texture	3.676	< 0.001
	Histogram VS. GLCM	−0.412	0.680
	Histogram VS. RLM	0.058	0.954
	Texture VS. GLCM	−3.941	< 0.001
	Texture VS. RLM	−3.607	< 0.001
	GLCM VS. RLM	0.462	0.644
HER2(+) VS. HER2(−)	Histogram VS. Texture	3.189	0.001
	Histogram VS. GLCM	−0.236	0.814
	Histogram VS. RLM	−2.267	0.023
	Texture VS. GLCM	−3.407	< 0.001
	Texture VS. RLM	−4.918	< 0.001
	GLCM VS. RLM	−2.099	0.036
Ki-67 low VS. high	Histogram VS. Texture	3.493	< 0.001
	Histogram VS. GLCM	−1.542	0.123
	Histogram VS. RLM	−0.559	0.576
	Texture VS. GLCM	−4.795	< 0.001
	Texture VS. RLM	−4.066	< 0.001
	GLCM VS. RLM	1.174	0.240

ISSN: 1471-2407