3D-Guided Backward Iterative Structural Refinement for Histology-to-MRI Registration

Qitai Sun, Xiao Li, Han Zhang, Chenjie Jia, Xiaowei He, Tuo Zhang, Yudan Ren

Abstract: Histology-to-MRI volume registration is an important preprocessing procedure in brain structure quantification, brain structure mapping, and pathological research of brain diseases. However, there are still many challenges preventing precise registration, especially the complex deformation of slices during sectioning, distinct modality of images and sparse slices intervals. To improve registration accuracy, a 3D-guided backward iterative structural refinement registration method is proposed in this paper. Inspired by the thought of the back-propagation algorithm, this method indirectly obtains the structural information during each registration iteration, continuously finetunes the registration results, and successively achieves improvement in registration accuracy. According to several validation and contrast experiments, this method is demonstrated to be robust and effective. This method can be widely applied in solving histology-to-MRI registration problem.