Future Artificial Intelligence-tailored Hospital (FAITH)
Current Artificial Intelligence (AI)-based diagnostic tools are developed using curated and clean datasets, therefore their applicability to the real-life clinical scenarios, when patient data is of limited quality and human life is at stake, is not guaranteed. This research project aims to revolutionize clinical AI realm by creating the next generation robust AI methods for imaging data quality control and automatic diagnosis. The very essence of our approach in automating the whole clinical imaging pipeline instead of performing only separate applications for a specific clinical situation as done currently. In this project, we focus on development of new methods for uncertainty estimation, anomaly detection, domain adaptation and semi-supervised learning applied to medical images at scale.