Table 3 Performance of the three different regression models for VF prediction from OCT data (excluding any augmentation using MICE and SMOTE) with fivefold cross-validation (report on test results with fivefold cross-validation, mean ± standard deviation).

From: Predicting visual field global and local parameters from OCT measurements using explainable machine learning

ML model

Folds

MD

VFI

PSD

R

RMSE (dB)

MAE (dB)

R

RMSE

MAE

R

RMSE (dB)

MAE (dB)

XGBoost

Fold-1

0.78

3.07

1.97

0.79

8.10

4.46

0.76

2.27

1.44

Fold-2

0.80

2.87

1.99

0.76

7.94

4.91

0.71

2.54

1.83

Fold-3

0.77

3.76

2.51

0.82

9.15

5.08

0.74

2.61

1.68

Fold-4

0.63

3.84

2.68

0.68

8.68

5.46

0.65

2.61

1.77

Fold-5

0.81

2.62

2.00

0.83

6.52

4.61

0.63

2.70

1.81

Mean ± STD

0.76 ± 0.08

3.23 ± 0.54

2.23 ± 0.34

0.78 ± 0.06

8.08 ± 0.99

4.91 ± 0.39

0.70 ± 0.06

2.55 ± 0.16

1.70 ± 0.16

95% CI

0.69–0.82

2.76–3.71

1.93–2.53

0.72–0.83

7.21–8.95

4.56–5.25

0.65–0.75

2.4–2.69

1.56–1.84

SVM

Fold-1

0.75

3.45

2.03

0.69

10.18

5.02

0.72

2.45

1.51

Fold-2

0.78

3.24

2.21

0.75

9.34

5.45

0.77

2.36

1.70

Fold-3

0.77

4.23

2.60

0.72

13.37

6.29

0.73

2.76

1.75

Fold-4

0.64

3.65

2.50

0.62

9.64

5.28

0.7

2.37

1.64

Fold-5

0.73

3.18

2.26

0.70

9.14

5.08

0.69

2.48

1.70

Mean ± STD

0.73 ± 0.06

3.55 ± 0.42

2.32 ± 0.23

0.70 ± 0.05

10.34 ± 1.74

5.43 ± 0.52

0.72 ± 0.03

2.48 ± 0.16

1.66 ± 0.09

95% CI

0.68–0.78

3.18–3.92

2.12–2.52

0.65–0.74

8.81–11.86

4.97–5.88

0.69–0.75

2.34–2.63

1.58–1.74

RF

Fold-1

0.84

2.73

1.85

0.76

8.29

4.65

0.79

2.12

1.37

Fold-2

0.78

3.01

2.09

0.78

7.53

4.97

0.79

2.22

1.69

Fold-3

0.83

3.26

2.29

0.78

10.10

5.45

0.77

2.53

1.57

Fold-4

0.64

3.77

2.60

0.66

8.96

5.61

0.72

2.33

1.59

Fold-5

0.78

2.82

2.20

0.81

6.68

4.47

0.71

2.41

1.65

Mean ± STD

0.77 ± 0.08

3.12 ± 0.42

2.21 ± 0.28

0.76 ± 0.06

8.31 ± 1.31

5.03 ± 0.49

0.75 ± 0.04

2.32 ± 0.16

1.57 ± 0.12

95% CI

0.71–0.84

2.75–3.48

1.96–2.45

0.7–0.81

7.16–9.47

4.6–5.46

0.72–0.79

2.18–2.46

1.47–1.68

  1. Boldface numbers indicate the best performance for that metric. Input features included RNFL, GC-IPL and MC thickness data in the spatial and frequency domain. 95% CI = Mean ± 1.96 × Standard Error (SE), SE = STD/\(\sqrt{\text{n}}\), n = 5, CI: Confidence Interval, MD mean deviation, VFI visual field index, PSD pattern standard deviation, RMSE root mean squared error, MAE mean absolute error.
Back to article page