Table 3 Sensitivity of classification algorithm in detection of visually significant cataractsa at different specificity levels

From: Detecting visually significant cataract using retinal photograph-based deep learning

Testing sets

Sensitivity (%) (95% CI)

At 70% specificity

At 80% specificity

At 90% specificity

Internal

 SIMES (n = 72; N = 1,692)

98.8 (97.6, 100.0)

98.8 (97.6, 100.0)

92.7 (87.2, 97.8)

External

 SCES (n = 141; N = 5,747)

100.0 (100.0,100.0)

98.9 (97.7, 100.0)

91.0 (87.2, 94.7)

 SINDI (n = 138; N = 5,626)

97.7 (96.4, 98.9)

97.0 (95.2,98.9)

94.0 (90.8, 96.7)

 BES (n = 48; N = 4,632)

95.3 (91.8, 97.9)

85.7 (80.4, 90.9)

71.5 (65.1, 77.8)

  1. aCataract with BCVA < 20/60.
  2. n, number of eyes with visually significant cataracts with BCVA cut-off <20/60; N, total number of eyes
Back to article page