Qualitative and Quantitative MRI Analysis in IDH1 Genotype Prediction of Lower-Grade Gliomas: A Machine Learning Approach

Table 2

List of selected VASARI and radiomics features.

Feature selection
Top 5 VASARI features	AUC value	Top 10 radiomics features	AUC value
Enhancement quality	0.752	GLRLM short run low gray-level emphasis	0.756
Deep white matter invasion	0.738	GLRLM low gray-level run emphasis	0.682
Tumor location	0.684	GLRLM run-length variance	0.678
Proportion of necrosis	0.682	Histogram minimum	0.677
T1/FLAIR ratio	0.632	Eccentricity	0.662
		GLSZM large zone high gray-level emphasis	0.641
		GLSZM low gray-level zone emphasis	0.628
		Histogram energy	0.616
		Histogram standard deviation	0.612
		GLSZM zone-size nonuniformity	0.607

VASARI: Visually Accessible Rembrandt Images; AUC: area under the receiver operating characteristic curve; GLRLM: gray-level run-length matrix; GLSZM: gray-level size zone matrix.