MultiRecon

MultiRecon aims at developing new image reconstruction techniques for multimodal medical imaging (PET/CT and PET/MRI) using machine learning. In multimodal imaging, current image reconstruction techniques reconstruct each modality independently. However, it is possible to exploit inter-modality information in order to “consolidate” the images to reduce noise and ultimately to reduce of the patient the dose. This information can be based on analytical models, but it can also be learned. In this project we explore machine and deep learning methods that can learn and exploit inter-modality information so that images can “talk to each other”.

MultiRecon is funded by the French National Research Agency (ANR) with grant number ANR-20-CE45-0020.