Meeting Banner
Abstract #1138

Comparison of tumor autosegmentation techniques from an undersampled dynamic radial bSSFP acquisition on a low-field MR-linac

Florian Friedrich1,2, C. Katharina Spindeldreier3, Juliane Hörner-Rieber3, Sebastian Klüter3, Peter Bachert1,2, Mark E. Ladd1, and Benjamin R. Knowles1
1Medical Physics in Radiology, German Cancer Research Center (DKFZ), Heidelberg, Germany, 2Department of Physics and Astronomy, Heidelberg University, Heidelberg, Germany, 3Department of Radiation Oncology,, University Hospital of Heidelberg, Heidelberg, Germany

MR-linac hybrid systems can dynamically image a tumor during radiotherapy to aid in a more precise delivery of the radiation dose. Motion tracking of the target is required and is currently performed by a deformable image registration on Cartesian bSSFP images. This study compares three different tracking methods (convolutional neuronal network, multi-template matching, and deformable image registration) to track a lung tumor in Cartesian images, where the performance of the three methods did not differ significantly. The convolutional neuronal network provided minimal decrease in tracking accuracy in a healthy volunteer when undersampled radial images were used to accelerate image acquisition.

This abstract and the presentation materials are available to 2020 meeting attendees and eLibrary customers only; a login is required.

Join Here