Table 3 The predictive performance of 5 models

From: Interpretable longitudinal glaucoma visual field estimation deep learning system from fundus images and clinical narratives

 

Internal dataset

External dataset

RMSE

MAE

R2

RMSE

MAE

R2

Original CFP

 Pointwise

5.724 (5.588 ~ 5.825)

4.131 (4.018 ~ 4.243)

0.534 (0.508 ~ 0.559)

6.421 (6.129 ~ 6.592)

4.860 (4.672 ~ 5.049)

0.262 (0.241 ~ 0.282)

 Pointwise-mean

3.635 (3.241 ~ 4.085)

2.632 (2.361 ~ 2.933)

0.730 (0.663 ~ 0.779)

3.840 (3.573 ~ 4.161)

2.917 (2.705 ~ 3.165)

0.473 (0.378 ~ 0.553)

 MS

3.561 (3.196 ~ 4.029)

2.632 (2.379 ~ 2.931)

0.740 (0.674 ~ 0.790)

3.958 (3.664 ~ 4.301)

2.977 (2.746 ~ 3.248)

0.440 (0.320 ~ 0.541)

ROI

 

 Pointwise

5.353 (5.208 ~ 5.456)

3.903 (3.795 ~ 4.012)

0.594 (0.575 ~ 0.612)

5.575 (5.403 ~ 5.693)

4.417 (4.292 ~ 4.543)

0.409 (0.383 ~ 0.436)

 Pointwise-mean

3.186 (2.839 ~ 3.611)

2.301 (2.066 ~ 2.577)

0.792 (0.722 ~ 0.834)

2.907 (2.758 ~ 3.076)

2.385 (2.229 ~ 2.538)

0.698 (0.646 ~ 0.744)

 MS

3.328 (3.012 ~ 3.729)

2.465 (2.234 ~ 2.739)

0.773 (0.719 ~ 0.814)

3.714 (3.466 ~ 3.971)

2.897 (2.696 ~ 3.131)

0.507 (0.411 ~ 0.590)

ROI + OD/OC Segmentation

 Pointwise

5.496 (5.588 ~ 5.825)

3.980 (3.862 ~ 4.098)

0.573 (0.551 ~ 0.595)

6.007 (5.827 ~ 6.132)

4.684 (4.544 ~ 4.824)

0.313 (0.290 ~ 0.336)

 Pointwise-mean

3.426 (3.056 ~ 3.946)

2.430 (2.183 ~ 2.726)

0.760 (0.689 ~ 0.809)

3.484 (3.260 ~ 3.729)

2.762 (2.570 ~ 2.957)

0.566 (0.499 ~ 0.626)

 MS

3.481 (3.077 ~ 3.982)

2.460 (2.207 ~ 2.754)

0.752 (0.670 ~ 0.801)

4.018 (3.755 ~ 4.314)

3.170 (2.955 ~ 3.414)

0.423 (0.302 ~ 0.527)

ROI + Clinical Information

 Pointwise

4.984 (4.836 ~ 5.086)

3.575 (3.476 ~ 3.675)

0.649 (0.627 ~ 0.671)

   

 Pointwise-mean

2.706 (2.462 ~ 2.997)

2.016 (1.822 ~ 2.241)

0.850 (0.810 ~ 0.878)

   

 MS

2.847 (2.574 ~ 3.202)

2.099 (1.903 ~ 2.335)

0.834 (0.793 ~ 0.864)

   

ROI + Longitudinal

 Pointwise

4.391 (4.268 ~ 4.477)

3.098 (3.012 ~ 3.183)

0.674 (0.655 ~ 0.693)

   

 Pointwise-mean

2.432 (2.168 ~ 2.845)

1.625 (1.504 ~ 1.781)

0.843 (0.786 ~ 0.878)

   

 MS

2.470 (2.226 ~ 2.864)

1.666 (1.543 ~ 1.821)

0.839 (0.780 ~ 0.872)

   
  1. RMSE root mean square error, MAE mean absolute error, R2  R-squared, CFP color fundus photographs, ROI region of interest, MS mean sensitivity, OD optic disc, OC optic cup.
Back to article page