Table 5.

Performance (area under the ROC curve with 95% confidence interval, area under the PR curve with 95% confidence interval) of models trained on both clinical data and visual fields, exclusively visual fields, and exclusively clinical data given 4 timesteps of data with respect to ground truth from traditional algorithms corresponding to Figure 4.

Reference	MD slope		VFI slope		PLR
Model	AUROC	AUPR	AUROC	AUPR	AUROC	AUPR
Clinical Data and Visual Fields	0.939 (±0.008)	0.772 (±0.121)	0.892 (±0.012)	0.786 (±0.045)	0.915 (±0.009)	0.794 (±0.060)
Visual Fields Only	0.815 (±0.015)	0.519 (±0.128)	0.794 (±0.020)	0.490 (±0.101)	0.794 (±0.015)	0.526 (±0.079)
Clinical Data Only	0.743 (±0.018)	0.507 (±0.058)	0.756 (±0.019)	0.467 (±0.151)	0.792 (±0.014)	0.625 (±0.121)

MD = Mean Deviation, PLR = Pointwise Linear Regression, VFI = Visual Field Index.