Table 2 Performance of machine learning models for detecting refractory central serous chorioretinopathy cases in the external validation (Gangnam Severance Hospital dataset).

From: DeepPDT-Net: predicting the outcome of photodynamic therapy for chronic central serous chorioretinopathy using two-stage multimodal transfer learning

 

AUC

(95% CI)

Accuracy (%)

(95% CI)

Sensitivity (%)

(95% CI)

Specificity (%)

(95% CI)

P value*

(A) FP-ResNet50 model

0.813 (0.570‒0.952)

68.4

(43.5‒87.4)

100.0

(29.2‒100.0)

62.5

(35.4‒84.8)

0.196

(B) Clinical data-XGBoost model

0.854 (0.619‒0.972)

79.0

(54.4‒94.0)

100.0

(29.2‒100.0)

75.0

(47.6‒92.7)

0.568

(A + B) Combined model (DeepPDT-Net)

0.917 (0.697‒0.993)

84.2

(60.4‒96.6)

100.0

(29.2‒100.0)

81.3

(54.4‒96.0)

Ref.

  1. AUC Area under the receiver operating characteristic curve, CI Confidence interval.
  2. *The AUCs were compared using the DeLong's method.
Back to article page