Table 4 Performance of SE-CrossT models based on WSIs, US and multimodal according to validation and test cohorts
SE-CrossT | ACC | AUC | SENS | SPEC | PPV | NPV |
|---|---|---|---|---|---|---|
Validation cohort (n = 49) | ||||||
 WSIs | 0.857 (0.783–0.890) | 0.810 (0.777–0.819) | 0.730 (0.678–0.750) | 0.788 (0.726–0.818) | 0.751 (0.680–0.779) | 0.751 (0.712–0.781) |
 US | 0.878 (0.811–0.880) | 0.837 (0.806–0.855) | 0.883 (0.819–0.901) | 0.882 (0.815–0.890) | 0.859 (0.803–0.887) | 0.879 (0.816–0.884) |
 Multimodal | 0.918 (0.838–0.924) | 0.851 (0.793–0.869) | 0.896 (0.833–0.918) | 0.888 (0.861–0.895) | 0.866 (0.819–0.891) | 0.885 (0.830–0.896) |
Test cohort 1 (n = 51) | ||||||
 WSIs | 0.776 (0.746–0.792) | 0.703 (0.649–0.745) | 0.727 (0.647–0.763) | 0.738 (0.689–0.759) | 0.724 (0.680–0.725) | 0.729 (0.680–0.737) |
 US | 0.823 (0.751–0.829) | 0.817 (0.775–0.836) | 0.859 (0.805–0.869) | 0.867 (0.821–0.882) | 0.833 (0.783–0.870) | 0.855 (0.810–0.865) |
 Multimodal | 0.882 (0.854–0.920) | 0.851 (0.791–0.856) | 0.878 (0.828–0.893) | 0.855 (0.826–0.868) | 0.840 (0.775–0.864) | 0.864 (0.789–0.884) |
