Primary Tumor and Inter-Organ Augmentations for Supervised Lymph Node Colon Adenocarcinoma Metastasis Detection

The scarcity of labeled data is a major bottleneck for developing accurate\nand robust deep learning-based models for histopathology applications. The\nproblem is notably prominent for the task of metastasis detection in lymph\nnodes, due to the tissue's low tumor-to-non-tumor ratio, resulting in labor-\nand time-intensive annotation processes for the pathologists. This work\nexplores alternatives on how to augment the training data for colon carcinoma\nmetastasis detection when there is limited or no representation of the target\ndomain. Through an exhaustive study of cross-validated experiments with limited\ntraining data availability, we evaluate both an inter-organ approach utilizing\nalready available data for other tissues, and an intra-organ approach,\nutilizing the primary tumor. Both these approaches result in little to no extra\nannotation effort. Our results show that these data augmentation strategies can\nbe an efficient way of increasing accuracy on metastasis detection, but\nfore-most increase robustness.\n