Table 2 Performance of the MMF-PH

From: Development and validation of multimodal deep learning algorithms for detecting pulmonary hypertension

Dataset

Accuracy (95% CI)

Sensitivity (95% CI)

Specificity (95% CI)

Precision (95% CI)

NPV (95% CI)

F1 (95% CI)

AUROC (95% CI)

AUPRC (95% CI)

Retrospective test dataset

0.949 (0.929–0.967)

0.974 (0.959–0.987)

0.606 (0.441–0.769)

0.972 (0.957–0.985)

0.625 (0.459–0.800)

0.973 (0.961–0.983)

0.965 (0.940–0.983)

0.997 (0.995–0.999)

Prospective test dataset

0.933 (0.910–0.956)

0.949 (0.928–0.970)

0.783 (0.657–0.902)

0.976 (0.962–0.990)

0.621 (0.490–0.745)

0.962 (0.949–0.975)

0.939 (0.894–0.971)

0.991 (0.981–0.997)

External test dataset

0.722 (0.556–0.861)

0.895 (0.750–1.000)

0.529 (0.273–0.765)

0.680 (0.481–0.852)

0.818 (0.571–1.000)

0.773 (0.611–0.898)

0.814 (0.669–0.947)

0.867 (0.716–0.960)

  1. AUROC area under the receiver operating characteristic curve, AUPRC area under the precision-recall curve, CI confidence interval, MMF-PH multimodal fusion model for pulmonary hypertension, NPV negative predictive value.
Back to article page