All slides were taken from different patients.
There was no independence in the sense that they do not belong to a single bigger dataset. At the time when the train dataset was released, the test dataset did not exist yet. It is more recent. But the samples came from the same sources. The only difference is the moment when the digitization happened. It is a time-consuming process that was implemented in different phases and by the time of the last digitization, the training dataset was already released. The distribution/proportion is not exactly the same but they are close enough. You already have the ground truth for each of the samples.
All the best
Eduardo