A smart brain MR image completion method guided by synthetic-CT-based multimodal registration

We propose a novel method for brain MR image completion in the case of unknown deformation in MR images. An efficient technique for MR image completion can keep patients from requiring a secondary shot, thus limiting radiation exposure. To address this challenge, we propose an MR image completion method guided by multimodal registration and simultaneously propose a new multimodal registration method, i.e., synthetic-CT-based multimodal registration (IC-G-MR). The majority of existing multimodal registration methods cannot perform well under the conditions mentioned above. The significance of our method can be summarized as follows: (1) Through a combination of transfer learning and indirect multimodal registration, the proposed IC-G-MR method can be used to solve the problem of brain MR image completion based on corresponding normal CT images. (2) Due to the transfer learning technique, our method can generate synthetic brain CT images of good quality. (3) Due to the novel multimodal registration method, the proposed IC-G-MR method performs effectively in registration for large-scale elastically deformed MR images. The experimental results indicate that our proposed IC-G-MR method can effectively complete MR images of the brain.