Fig. 4: Performance of image-based deep learning model, clinical model, readers, and the OvcaFinder. | Nature Communications

Fig. 4: Performance of image-based deep learning model, clinical model, readers, and the OvcaFinder.

From: Development and validation of an interpretable model integrating multimodal information for improving ovarian cancer diagnosis

Fig. 4

We reported the observed values (measure of centre) and 95% confidence intervals (error bars) of the AUCs, accuracies, specificities, sensitivities, PPVs, and NPVs of five readers (R1-R5), the image-based deep learning model, the clinical model, and OvcaFinder in the internal (a, n = 129) and external (b, n = 387) test datasets. OvcaFinder improved the readers’ AUCs, accuracies, specificities, and PPVs, while maintaining equivalent sensitivities and NPVs. AUC Area under the receiver operating characteristic curve, PPV Positive predictive value, NPV Negative predictive value.

Back to article page