Table 1 Performance comparison between G-NECNet and pathologists
From: A deep learning model for the diagnosis of gastric neuroendocrine carcinoma
Method | Internal test cohort | Â | External cohort | Â | Consultation cohort | Â | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Â | Sensitivity | difference | P | Specificity | difference | P | Sensitivity | difference | P | Specificity | difference | P | Sensitivity | difference | P | Specificity | difference | P |
G-NECNet | 0.952 | NA | NA | 0.986 | NA | NA | 0.923 | NA | NA | 0.942 | NA | NA | 0.952 | NA | NA | 1.000 | NA | NA |
Junior pathologist | 0.405 | −0.547 | <0.001 | 0.948 | −0.038 | <0.001 | 0.600 | −0.323 | <0.001 | 0.824 | −0.118 | <0.001 | 0.619 | −0.333 | 0.039 | 0.935 | −0.065 | 0.031 |
Senior pathologist | 0.655 | −0.297 | <0.001 | 0.995 | 0.009 | 0.227 | 0.538 | −0.385 | <0.001 | 0.939 | −0.003 | 1.000 | 0.714 | −0.238 | 0.125 | 1.000 | 0.000 | 1.000 |
Expert pathologist | 0.821 | −0.131 | 0.019 | 0.985 | −0.001 | 1.000 | 0.800 | −0.123 | 0.077 | 0.964 | 0.022 | 0.307 | 0.762 | −0.190 | 0.125 | 0.989 | −0.011 | 1.000 |
