Table 3. Results of the multivariate analyses considering all "significant" variables from univariate analyses, comparing stroke patients to non-stroke patients (Analysis #1), and infarct side with contralateral side (Analysis #2), and comparison using a ROC analysis.

A ROC analysis allows to assess the "quality" of a model. More specifically, the area under the ROC curve is a measure of the accuracy of the model. A model with a area under the ROC curve equal to 1 is a perfect model, where the predicted values match exactly the observed values; if the area under to ROC curve decreases below 1, this corresponds to the accumulation of false positive and false negative predicted values; finally, if the area under the ROC curve is 0.5, the model is as good in its predictive ability as flipping a coin.

"Carotid Stroke Patients" versus "Non-Carotid Stroke Patients" (Analysis #1)			Infarct Side versus Contralateral Side (Analysis #2)
CT features	Odds ratio (95% CI)	p value	CT features	Odds ratio (95% CI)	p value
Lumen area [mm2]	0.79 (0.71 – 0.88)	<0.001
wall volume [100 mm3]	4.08 (2.24 – 7.43)	<0.001	wall volume [100 mm3]	1.58 (1.00 – 2.49)	0.017
number of calcium clusters	0.31 (0.15 – 0.63)	0.001	fibrous cap thickness	0.11 (0.06 – 0.16)	0.013
number of lipid clusters	1.22 (1.08 – 1.39)	0.002	number of lipid clusters	1.58 (1.15 – 2.18)	0.005
location of largest lipid cluster	0.74 (0.61 – 0.89)	0.002	location of largest lipid cluster	0.83 (0.67 – 1.02)	0.048
Area under ROC curve = 0.796			Area under ROC curve = 0.832

CI stands for confidence interval