Table 2. Comparison of the spatial algorithm with machine learning (SVM) for the classification of glaucoma.

	Spatial algorithm	SVM
Average AUROC [%]	98.3	82.4
Standard deviation AUROC [%]	3.1	2.3
Minimum AUROC [%]	85.1	76.1
Maximum AUROC [%]	99.7	88.0
Average sensitivity [%]	95.4	74.3
Average specificity [%]	94.2	79.3

We used 100 random splits of the ORIGA dataset (70% training, 30% testing). Both the spatial algorithm and SVM used the full pCDR, rather than the simple vertical vCDR alone. These data comes from semi-automated segmentation.