Accuracy of Deep Learning Algorithms for the Diagnosis of Retinopathy of Prematurity by Fundus Images: A Systematic Review and Meta-Analysis

Table 2

The results of primary and subgroup analyses.

	Sensitivity (95% CI)	Specificity (95% CI)	PLR (95% CI)	NLR (95% CI)	DOR (95% CI)	AUC (95% CI)	Spearman r ( value)
Primary analyses	0.953 (0.946–0.959)	0.975 (0.973–0.977)	19.265 (8.431–44.019)	0.065 (0.040–0.105)	313.73 (115.85–849.60)	0.984 (0.978–0.989)	−0.561 (0.030)
Validation dataset	0.934 (0.922–0.945)	0.973 (0.969–0.977)	26.232 (6.978–98.616)	0.076 (0.046–0.125)	359.58 (94.565–1367.3)	0.977 (0.968–0.986)	−0.612 (0.060)
Test dataset	0.969 (0.961–0.975)	0.977 (0.974–0.979)	22.853 (12.593–41.475)	0.049 (0.026–0.092)	522.92 (213.89–1278.4)	0.987 (0.982–0.992)	−0.280 (0.354)
Define ROP	0.956 (0.949–0.962)	0.979 (0.977–0.981)	30.118 (19.225–47.184)	0.055 (0.033–0.092)	576.21 (238.54–1391.9)	0.9895 (0.9849–0.9941)	−0.503 (0.138)
Distinguish ROP	0.931 (0.906–0.952)	0.856 (0.826–0.882)	7.927 (2.049–30.674)	0.097 (0.038–0.252)	88.655 (13.251–593.13)	0.9820 (0.9641–0.9999)	−0.600 (0.285)

Note. PLR, positive likelihood ratio; NLR, negative likelihood ratio; DOR, diagnostic odds ratios.