Skip to main content
. 2020 Oct 30;3:143. doi: 10.1038/s41746-020-00350-y

Table 1.

Performance of the DLIFS in detecting poor-quality images.

CMAAI test set Zhongshan ophthalmic centre dataset Xudong ophthalmic hospital dataset
AUC (95% CI) 0.996 (0.995–0.997) 0.994 (0.989–0.997) 0.997 (0.995–0.998)
Sensitivity (95% CI) 96.9% (96.3–98.3) 95.6% (92.9–98.3) 96.6% (94.5–98.7)
Specificity (95% CI) 96.6% (96.1–97.1) 97.9% (97.1–98.7) 98.8% (98.4–99.2)

AUC area under the receiver operating characteristic curve, CI confidence interval, CMAAI Chinese Medical Alliance for Artificial Intelligence, DLIFS deep learning-based image filtering system.