Pathology AI for a Federated Quality Assurance Program: Ovarian Cancer Pilot

As type-specific treatments are being developed for patients with epithelial ovarian cancer, it has become important to accurately diagnose the distinct cancer types. Our vision is to establish an international network for AI-based, privacy-protected pathology quality assurance. As proof of concept, we propose to develop and deploy a machine learning-based ovarian cancer histopathology classifier.

We plan to use privacy-preserving synthetic data generation to train an ovarian cancer classifier, using pathology images collected at our three institutions. We will specifically identify the privacy threats associated with the necessary data-sharing and then develop technical and socioethical guidelines that can be deployed around the world. Our long-term vision is to establish a learning systems network with point-of-care diagnostic applications that could also inform similar advances in other cancers.