Automatic Prostate Tumor Segmentation: Does the Convolutional Neural Network Learn How the Tumor Looks, or What the Radiologist Sees?
Deepa Darshini Gunashekar1, Lars Bielak1,2, Benedict Oerther 3, Matthias Benndorf 3, Anca Grosu 2,3, Constantinos Zamboglou2,3, 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, Freiburg im Breisgau, Germany, 3Dept.of Radiology, Medical Center University of Freiburg, Faculty of Medicine, University of Freiburg, Freiburg, Germany
A convolutional neural network was implemented to automatically segment tumors in multi-parametric MRI data. The influence of the variability in the ground truth data was evaluated for automated prostate tumor segmentation. Therefore, the agreement between the predictions of the CNN was measured with co-registered whole mount histopathology images and the tumor contours drawn by an expert radio-oncologist. The results indicate that the network can discriminate tumor from healthy tissue rather than mimicking the radiologist.
This abstract and the presentation materials are available to members only;
a login is required.