Skip to main content

Table 5 Mean and standard deviation of the estimation error of the clinical parameters for both networks and Topcon microscope built-in software

From: Fully convolutional architecture vs sliding-window CNN for corneal endothelium cell segmentation

Dataset

ECD (cells/mm2)

CV (%)

HEX (%)

Error

 Topcon

−4.1±41.7

1.9±2.6

−2.2±7.2

 SW-net

9.9±23.1

0.5±1.6

−0.7±2.0

 U-net

3.2±10.1

0.4±0.7

−0.2±1.0

Absolute error

 Topcon

29.8±29.6

2.3±2.2

5.3±5.3

 SW-net

14.9±20.2

0.8±1.4

1.6±2.5

 U-net

7.8±7.2

0.6±0.5

0.9±0.8

 

ECD (%)

CV (%)

HEX (%)

Relative error

 Topcon

1.3±1.4

10.1±9.0

8.0±8.9

 SW-net

0.8±1.3

3.6±7.3

2.1±2.9

 U-net

0.4±0.4

2.8±2.6

1.3±1.0

  1. Error and absolute error are computed as the difference (and absolute difference) between estimates and gold standard values. Relative error is computed as the percentage of the absolute error with respect to the gold standard values