Some studies (e.g., urothelial) contain both whole slide images (WSIs) and tissue microarrays(?). They're treated identically when being allocated to the training, validation, and test sets, so by chance the WSIs could all be allocated to the test set, even though they contain many more ROIs than the microarray specimens. We should stabilize the way they're allocated so that WSIs take priority during specimen allocation.
Some studies (e.g., urothelial) contain both whole slide images (WSIs) and tissue microarrays(?). They're treated identically when being allocated to the training, validation, and test sets, so by chance the WSIs could all be allocated to the test set, even though they contain many more ROIs than the microarray specimens. We should stabilize the way they're allocated so that WSIs take priority during specimen allocation.