Table 3 Refractive errors of SVR-assisted calculations for long and normal eyes.

From: Training data size and predication errors in the use of machine-learning assisted intraocular lens power calculation

Date set size (eye)

4800

2400

1200

600

300

150

75

Mean (SD), D

 Long AXL

− 0.03 (0.48)

− 0.02 (0.49)

− 0.03 (0.49)

− 0.03 (0.50)

− 0.03 (0.51)

− 0.02 (0.54)

− 0.01 (0.71)

 Normal AXL

0.00 (0.43)

0.00 (0.43)

0.00 (0.43)

0.00 (0.44)

0.00 (0.44)

0.00 (0.46)

0.01 (0.49)

Within ± 0.25 D error

 Long AXL

50.9%

48.5%

50.9%

48.3%

46.8%

45.0%

39.6%

 Normal AXL

50.4%

50.2%

50.5%

49.9%

49.1%

48.0%

47.0%

Within ± 0.50 D error

 Long AXL

78.9%

78.2%

77.6%

78.2%

75.8%

74.9%

69.2%

 Normal AXL

79.7%

79.0%

79.0%

79.4%

78.1%

77.5%

75.2%

Within ± 1.00 D error

 Long AXL

96.9%

97.1%

97.3%

96.7%

96.5%

95.1%

92.2%

 Normal AXL

97.2%

97.1e%

97.2%

97.1%

96.9%

96.3%

95.3%

  1. D: diopter, SD: standard deviation, AXL: axial length.