A novel loss function to reproduce texture features for deep learning-based MRI-to-CT synthesis.

Journal: Medical physics
Published Date: Dec 3, 2023

Abstract

BACKGROUND: Studies on computed tomography (CT) synthesis based on magnetic resonance imaging (MRI) have mainly focused on pixel-wise consistency, but the texture features of regions of interest (ROIs) have not received appropriate attention.

Authors

  • Siqi Yuan
    School of Arts and Communication, Beijing Normal University, Beijing, China.
  • Yuxiang Liu
    National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China; School of Physics and Technology, Wuhan University, Wuhan, China.
  • Ran Wei
  • Ji Zhu
    Department of Statistics, University of Michigan, Ann Arbor, Michigan.
  • Kuo Men
    State Key Laboratory of Advanced Materials for Smart Sensing, GRINM Group Co., Ltd., Beijing 100088, China.
  • Jianrong Dai
    National Cancer Center/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China.

Keywords

External Resources

Popular Topics
Recent Journals