Learning with less data via Weakly Labeled Patch Classification in Digital Pathology

In Digital Pathology (DP), labeled data is generally very scarce due to the\nrequirement that medical experts provide annotations. We address this issue by\nlearning transferable features from weakly labeled data, which are collected\nfrom various parts of the body and are organized by non-medical experts. In\nthis paper, we show that features learned from such weakly labeled datasets are\nindeed transferable and allow us to achieve highly competitive patch\nclassification results on the colorectal cancer (CRC) dataset [1] and the\nPatchCamelyon (PCam) dataset [2] while using an order of magnitude less labeled\ndata.\n