Table 3.

Differentiating non-arteritic anterior ischaemic optic neuropathy versus optic neuritis using magnetic resonance images after adjusting for significant demographic factors.

Factor in the model	AUC*	p value
Basic model: Reference model
Age and sex	0.94	N/A
Any location model: Basic model+
Post-contrast enhancement	0.95	0.26
DWI signal	0.95	0.31
Post-contrast enhancement or DWI signal	0.95	0.44
Post-contrast enhancement positive and DWI signal negative	0.94	0.81
Disc model: Basic model+
Post-contrast enhancement	0.95	0.14
DWI signal	0.94	0.81
Post-contrast enhancement or DWI signal	0.95	0.087
Post-contrast enhancement positive and DWI signal negative	N/A	N/A
Retrobulbar intraorbital segment model: Basic model+
Post-contrast enhancement	0.98	0.002
DWI signal	0.95	0.018
Post-contrast enhancement or DWI signal	0.98	0.003
Post-contrast enhancement positive and DWI signal negative	0.97	0.061

Note. Disc = within 2 mm of the intraocular optic disc. DWI = diffusion-weighted imaging; NAION = non-arteritic anterior ischaemic optic neuropathy; ON = optic neuritis; OR = odd ratio.

*The area under ROC curve, AUC, measures the probability that presence of image characteristics, i.e., PCE and/or DWI at disc head and/or retrobulbar intraorbital segment, can correctly classify NAION and ON patients. An AUC of 1.0 represents a perfect classification; an AUC of 0.5 represents a worthless image characteristic. In general, a characteristic is excellent if 0.9 ≤ AUC ≤ 1.0; good if 0.80 ≤ AUC < 0.90; fair if 0.70 ≤ AUC < 0.80; poor if 0.60 ≤ AUC < 0.70; and failed if AUC < 0.60.