Abstract #2339

Quantifying Information Content of Multiparametric MRI Data for Automatic Tumor Segmentation using CNNs

Lars Bielak^1,2, Nicole Wiedenmann^2,3, Thomas Lottner¹, Hatice Bunea^2,3, Anca-Ligia Grosu^2,3, and Michael Bock^1,2

¹Dept.of Radiology, Medical Physics, Medical Center University of Freiburg, Faculty of Medicine, University of Freiburg, Freiburg, Germany, ²German Cancer Consortium (DKTK), Partner Site Freiburg, Freiburg, Germany, ³Department of Radiation Oncology, Medical Center University of Freiburg, Faculty of Medicine, University of Freiburg, Freiburg, Germany

Multimodality imaging with CT, PET, and MRI is the basis for precise tumor segmentation in radiation therapy. We analyze which MR imaging contrasts mainly improve the segmentation performance of a CNN by training multiple networks using different input channels. The predictive value of 7 different contrasts is compared for two tumor regions, gross tumor volume and lymph node metastasis, in head and neck tumor patients.

This abstract and the presentation materials are available to members only; a login is required.

Join Here