Dynamic analysis of iris changes and a deep learning system for automated angle-closure classification based on AS-OCT videos

Table 2 Comparison of the classification performance on private anterior segment optical coherence tomography (AS-OCT) video datasets and image datasets

	AUC (95% CI)	Sensitivity	Specificity	Accuracy	Precision	Recall	F1 Score	Kappa
Casia dataset
Images	0.766 (0.610–0.923)	0.790	0.750	0.771	0.771	0.771	0.771	0.539
Original videos	0.820 (0.680–0.961)	0.833	0.765	0.800	0.801	0.800	0.800	0.599
Aligned videos	0.905 (0.802–1.000)	0.947	0.690	0.829	0.844	0.829	0.824	0.648
Zeiss dataset
Images	0.767 (0.620–0.914)	0.750	0.933	0.821	0.852	0.821	0.823	0.643
Original videos	0.837 (0.713–0.961)	0.826	0.813	0.821	0.823	0.821	0.821	0.633
Aligned videos	0.919 (0.831–1.000)	0.913	0.933	0.921	0.923	0.921	0.921	0.837

AUC = area under the receiver operating characteristic curve; CI = confidence interval
The bold numbers indicate optimal results

ISSN: 2326-0254