Abstract #2387

The repeatability of deep learning-based segmentation of the prostate on T2-weighted MR images

Mohammed R. S. Sunoqrot¹, Sandra Kucharczak², Magdalena Grajek², Kirsten M. Selnæs^1,3, Tone F. Bathen^1,3, and Mattijs Elschot^1,3

¹Department of Circulation and Medical Imaging, NTNU, Norwegian University of Science and Technolog, Trondheim, Norway, ²Department of quantum electronics, Adam Mickiewicz University, Poznań, Poland, ³Department of Radiology and Nuclear Medicine, St. Olavs Hospital, Trondheim University Hospital, Trondheim, Norway

Inter- and intra-observer variability are current limitations of radiological reading of multiparametric MR images of the prostate. Deep learning (DL)-based segmentation has proven to provide good performance, but little is known about the repeatability of these methods. In this work, we investigated the intra-patient repeatability of shape features for DL segmentation methods of the prostate on T2-weighted MR images and compared it to manual segmentations. We found that the repeatability of the investigated methods is excellent for most of the investigated shape features.

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

Join Here