The morphological features that pathologists use to differentiate neoplasms from normal tissue are nonspecificto tissue type. For example, if given a Ki67 stained biopsy of neuroendocrine or breast tumor, apathologist would be able to correctly identify morphologically abnormal cells in both samples but maystruggle to identify the origin of both samples. This is also true for other pathological malignancies such ascarcinomas, sarcomas, and leukemia. This implies that computer algorithms trained to recognize tumorfrom one site should be able to identify tumor from other sites with similar tumor subtypes. Here, we presentthe results of an experiment that supports this hypothesis. We train a deep learning system to distinguishtumor from non-tumor regions in Ki67 stained neuroendocrine tumor digital slides. Then, we test the same,unmodified, deep learning model to distinguish breast cancer from non-cancer regions. When applied to asample of 96 high power fields, our system achieved a cumulative pixel-wise accuracy of 86% across thesehigh-power fields. To our knowledge, our results are the first to formally demonstrate generalizedsegmentation of tumors from different sites of origin through image analysis. This paradigm has thepotential to help with the design of tumor identification algorithms as well as the composition of the datasetsthey draw from.
展开▼
