Skip to main content

Table 2 Performance of single models and the FMDLS

From: Identification of ocular refraction based on deep learning algorithm as a novel retinoscopy method

 

RM

CM

FMDLS

MAE

Accuracy (95% CI)

Specificity (95% CI)

Sensitivity (95% CI)

AUC (95% CI)

F1-score

MAE

r

Performance improvement

Spherea

0.86

0.790 (0.751–0.842)

0.991 (0.974–0.998)

0.795 (0.745–0.839)

0.798 (0.748–0.842)

0.775

0.63

0.815

27.10%

Sphereb

0.66

0.850 (0.825–0.875)

0.996 (0.99–0.998)

0.859 (0.835–0.883)

0.863 (0.839–0.887)

0.828

0.50

0.949

29.41%

Cylinder

0.38

0.860 (0.836–0.884)

0.989 (0.982–0.996)

0.861 (0.837–0.885)

0.834 (0.808–0.860)

0.863

0.31

0.807

26.67%

Axis

–

0.890 (0.816–0.964)

0.941 (0.849–0.981)

0.882 (0.776–0.944)

0.814 (0.708–0.902)

0.880

–

–

–

  1. RM regression model, CM classification model, FMDLS fusion model-based deep learning system, MAE mean absolute error, AUC area under the curve
  2. aModel without age as an eigenvector
  3. bModel with age as an eigenvector
Back to article page

BioMedical Engineering OnLine

ISSN: 1475-925X

Contact us