Emanuel Ström: Medical Imaging using a Learned Physics Model

Abstract

With the emergence of increasingly powerful high resolution imaging systems for medical imaging, there is also a growing need to accelerate the hand crafted physics models that are used in the imaging process. This talk is about carefully replacing specific components of the physics model with a neural network.

We will touch on topics like deep learning (obviously), inverse problems, optimisation and continuum formulations of medical imaging methods (lots of Hilbert spaces and integral operators and stuff here).