Fig. 2: Diagnostic performances of G-NECNet and pathologists.
From: A deep learning model for the diagnosis of gastric neuroendocrine carcinoma
a On the Internal test cohort, the G-NECNet achieved an AUROC of 0.993, outperforming all three pathologists with an AUROC ranging from 0.677 to 0.903 (the p-value was marked in the figure, Delong’s test, two-sided). b On the External-cohort, the G-NECNet achieved an AUROC of 0.985, outperforming all pathologists with an AUROC ranging from 0.712 to 0.882 (the p-value was marked in the figure, Delong’s test, two-sided). c On the Consultation-cohort, the G-NECNet achieved an AUROC of 1.000, outperforming all pathologists with an AUROC ranging from 0.770 to 0.876 (the p-value was marked in the figure, Delong’s test, two-sided).
