Heidelberg 2022 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
ST: Fachverband Strahlen- und Medizinphysik
ST 2: Poster Session
ST 2.3: Poster
Montag, 21. März 2022, 16:15–17:30, P
Image registration of CT and MRI scans using deep learning — •Alexander Ratke, Hannah Timm, Johannes Wintz, Göksu Ünlü, and Bernhard Spaan — TU Dortmund University, Dortmund, Germany
In radiation therapy, precise localisation of tumour and risk structures is important for therapy planning. Medical imaging methods, such as computed tomography (CT) and magnetic resonance imaging (MRI), offer different advantages due to the respective physical process, which can be combined by image fusion.
In this project, CT scans and T1- and T2-weighted MRI scans of different body regions are used. For their registration, two processing steps are performed. The preparatory part includes equalising the formats of the images, which is required for a neural-network-based registration. Deep learning is then used to filter structures of an image and to match them to a second image. The results of the registration of three-dimensional CT and MRI scans of the skull and the thorax will be presented as well as studies of quality and uncertainty, performed with a Shepp-Logan phantom.