Distinction between primary and metastatic mucinous ovarian carcinoma from histopathology images using deep learning

Accurately distinguishing between primary and gastrointestinal metastatic mucinous ovarian carcinoma (MOC) is crucial but remains highly challenging. We developed and validated the MOC Origin Prediction Model (MOCOPM), the deep learning model specifically designed to predict the origin of MOC from histopathology images. Cases of primary or gastrointestinal metastatic MOC were collected from three hospitals and divided into internal and external cohorts. Three neural networks were trained using the area under the receiver operating characteristic curve (AUROC) as the primary performance metric. After five-fold cross-validation on the internal cohort, the best-performing model was selected to construct MOCOPM, which was then externally validated. In total, 167 MOC patients were included. MOCOPM achieved an average AUROC of 0.91 internally and 0.96 externally. This model offers a promising tool to support clinical decision-making in MOC diagnosis.