From Data-driven to Knowledge-driven AI

Recent advances in AI have led to major breakthroughs in many scientific fields, including medical imaging. Much of the current success can be attributed to the development of deep learning techniques, which autonomously learn from large amount of data. However, purely data-driven methods may have certain issues in theory and in practice. For example, the model can be biased without appropriate data prior, difficult to generalize in cases of distribution shift, fragile when confronted with adversarial attacks, and opaque to users with its non-intuitive parameters in millions.

KDAI

In the Knowledge-Driven AI (KDAI) Lab, we strive to strengthen the current data-driven AI by integrating fundamental knowledge of applied natural sciences. We will conduct research on knowledge-driven AI, and demonstrate its value in two applied science fields: medical imaging and chemical engineering. At the same time, our research can be used more broadly, as it studies the fundamental methodologies of bringing knowledge into all the key components of AI: data collection, algorithm design, user interaction, and practical deployment. We believe that knowledge-driven AI will improve upon purely data-driven AI, providing understanding and trust on AI and propelling future development of applied sciences.

The KDAI Lab is part of the TU Delft AI Labs Programme, and is directed by Dr. Qian Tao (Department of Imaging Physics) and Dr. Artur Schweidtmann (Department of Chemical Engineering).

Research

We study generalizability and explainability of AI in medical imaging, for AI to be better understood, trusted, and appreciated in clinical applications. Research projects cover medical image segmentation, registration, and classification.

We welcome PhD and master students who are enthusiastic about AI and medical imaging to join us.  For more information, please contact: Q.Tao@tudelft.nl.

Biography

Qian Tao received her BSc degree (with Distinction) in Electrical Engineering, and her MSc degree (with Distinction) in Biomedical Engineering, both from Fudan University, Shanghai, China. She received her PhD degree from University of Twente, the Netherlands, with her PhD thesis entitled "Face Verification for Mobile Personal Devices", which presented a first-generation biometric authentication system on a mobile device. Her Ph.D work contributed to the 3Dface project of European Commission for the prototype biometric passport. Since 2009 Qian Tao has been working at the Division of Image Processing, Department of Radiology, Leiden University Medical Center, and her multidisciplinary research lines included cardiac MRI analysis and clinical applications, image-guided treatment of cardiac arrhythmia, and artificial intelligence in Radiology. Her research interest includes medical image analysis, machine learning, and their state-of-the-art integration in theory and in practice.