Table 2.

Radiomics features selected for each risk factor. Features are presented in order of importance (accuracy using only one feature) in the model for each risk factor.

CV risk factor	Radiomics signature	Feature type	ROI	Phase	Alone
High cholesterol	Spherical disproportion	Shape	MYO	ED	0.61
	Compactness	Shape	MYO	ED	0.60
	Skewness	First-order	LV	ED	0.59
	Informal measure of correlation	Texture	LV	ES	0.57
	Gray level non-uniformity	Texture	RV	ED	0.55
	Contrast	Texture	RV	ES	0.52
Diabetes	Median	First-order	MYO	ES	0.65
	Surface area to volume ratio	Shape	MYO	ED	0.61
	Energy	First-order	LV	ED	0.61
	Surface area	Shape	MYO	ES	0.58
	Dependence variance	Texture	LV	ED	0.57
	Large area high gray level emphasis	Texture	MYO	ED	0.57
	Energy	First-order	LV	ES	0.57
	Flatness	Shape	RV	ED	0.56
	Surface area	Shape	LV	ES	0.55
	Max 2D diameter column	Shape	RV	ED	0.50
	Difference average	Texture	LV	ES	0.44
Hypertension	Surface area to volume ratio	Shape	MYO	ED	0.61
	Percentile 10	First-order	RV	ES	0.58
	Informal measure of correlation	Texture	LV	ES	0.55
	Dependence non-uniformity normalized	Texture	LV	ED	0.54
	Size zone non-uniformity normalized	Texture	RV	ED	0.54
Current smokers	Gray level non-uniformity	Texture	MYO	ES	0.60
	Dependence entropy	Texture	LV	ED	0.57
	Standard deviation	First-order	MYO	ED	0.53
	Max 2D diameter column	Shape	RV	ED	0.50
	Large dependence low gray level emphasis	Texture	RV	ED	0.45
Previous smokers	Surface area to volume ratio	Shape	MYO	ED	0.57
	Busyness	Texture	LV	ES	0.54
	Run entropy	Texture	MYO	ES	0.50
	Skewness	First-order	RV	ES	0.50
	Run length non-uniformity	Texture	RV	ED	0.49
	Zone variance	Texture	LV	ED	0.49

ROI, region of interest, Alone: model performance using each radiomic feature individually; LV, left-ventricle; RV, right-ventricle; MYO, left ventricle myocardium; ED, end-diastolic.