Table 1 Comparison of the performance of the proposed FKD-CSS model and different AI models and ophthalmologists on internal cross-validation

From: Advanced and interpretable corneal staining assessment through fine grained knowledge distillation

Comparison

Pearson

AUC

AI Methods

ResNeSt50

0.834

0.827

ResNeXt50

0.858

0.847

DenseNet121

0.842

0.825

Ophthalmologists

Junior A

0.755

0.631

Junior B

0.665

0.600

Mid-level A

0.821

0.714

Mid-level B

0.861

0.749

Senior A

0.837

0.739

Senior B

0.858

0.761

Proposed

FKD-CSS

0.898

0.881

  1. The best results are highlighted in bold a and the second-best results are underlined.
  2. All p-value of Pearson test are statistically significant (p-value < 0.001).
  3. FKD fine-grained knowledge distillation, CSS corneal staining score, AUC area under the curve.
Back to article page