Deep Learning Approaches Using Convolutional Neural Networks to Generate Synthetic CT from Spinal MRI for Radiotherapy Planning
Yang Zhang1,2, Farouk Nouizi1, Ning Lang3, Xiaoying Xing3, Yongye Chen3, Qizheng Wang3, Enlong Zhang3, Huishu Yuan3, Ke Nie2, and Min-Ying Su1
1Department of Radiological Sciences, University of California, Irvine, CA, United States, 2Department of Radiation Oncology, Rutgers-Cancer Institute of New Jersey, Rutgers-Robert Wood Johnson Medical School, New Brunswick, NJ, United States, 3Department of Radiology, Peking University Third Hospital, Beijing, China
A total of 27 patients receiving both spinal CT and MR for evaluation of back pain were identified for analysis. MR images and CT image were co-registered first, and the CT was used as ground truth for training a deep learning algorithm using MR images to generate synthetic CT. In this study, we implemented cycleGAN to generate these synthetic CT images from their corresponding MR slices. Five-fold cross validation was used to evaluate the performance of the trained model. Compared to the original images, the Mean Average Error was 27.63±11.51, and the Peak Signal-to Noise Ratio was 19.44±5.72.
This abstract and the presentation materials are available to members only;
a login is required.