Table 10 Ablation study of M3-VF

From: A three-tier AI solution for equitable glaucoma diagnosis across China’s hierarchical healthcare system

Ablation Module

Accuracy

Sensitivity

Specificity

AUC

Base

0.4742 ± 0.0377

0.4455 ± 0.0345

0.8191 ± 0.0129

0.7136 ± 0.0256

0.4907

0.4551

0.8242

0.7432

Base +CBAM

0.6134 ± 0.0212

0.5883 ± 0.0290

0.8686 ± 0.0079

0.8298 ± 0.0166

0.6324

0.6202

0.8758

0.8486

Base + transformer encoder

0.6694 ± 0.0186

0.6478 ± 0.0276

0.8881 ± 0.0073

0.8723 ± 0.0153

0.6947

0.6868

0.8978

0.8899

Base + transformer encoder+ CBAM (ours)

0.7670 ± 0.0372

0.7515 ± 0.0450

0.9220 ± 0.0126

0.9283 ± 0.0144

0.8255

0.8210

0.9414

0.9516

  1. The upright numbers represent the means and standard deviations of the five-fold cross-validation results. Italicized numbers represent the results of the best model during cross-validation. Values below 0.6 are underlined.
Back to article page