In this talk, I will present an overview of some of our recent attempts to improve the safety of machine learning models in medical imaging. We focus on model robustness and reliability in the context of dataset mismatch between the development and deployment stage. Our aim is to provide safeguards such as automatic quality control, failure detection, and uncertainty estimation which are necessary for safe clinical use. We also discuss the use of causal reasoning to identify potential biases already at the design stage of model development. We will conclude with an outlook on how counterfactual image generation could help with explainability of predictive models.
Ben Glocker is a Reader in Machine Learning for Imaging co-leading the Biomedical Image Analysis Group. He leads the HeartFlow -Imperial Research Team and work as scientific adviser for Kheiron Medical Technologies. His research is at the intersection of medical image analysis and artificial intelligence aiming to build computational tools for improving image-based detection and diagnosis of disease.
Zoom Participation. See announcement.
About MLxMED: Medicine is complex and data-driven and discovery and decision making are increasingly enabled by machine learning. Machine learning has the potential to support, enable and improve medical discovery and clinical decision making in areas such as medical imaging, cancer diagnostics, precision medicine, clinical trials, and electronic health records. This seminar series focuses on new algorithms, real-world deployment, and future trends in machine learning in medicine. Hosted by the Department of Biomedical Informatics – University of Pittsburgh, UPMC, and CMU