Researches On Deep Learning Based Medical Image Registration

Image registration is one of the important parts in medical image processing and intelligent analysis.The accuracy of image registration will greatly affect the subsequent image processing and analysis.This paper focuses on the problem of medical image registration based on deep learning,and proposes the unsupervised deep learning methods based on model decoupling and regularization learning.Specifically,we first decompose the highly ill-conditioned inverse problem of image registration into two simpler sub-problems,to reduce the model complexity.Further,two light neural networks are constructed to approximate the solution of the two sub-problems and the training strategy of alternating iteration is used to solve the problem.Second,we propose a deep learning framework based on regularization learning,which approximates the regularization term in the form of neural network,and use data-driven method to learn more consistent with the regularization constraints of the actual data distribution,to avoid the registration errors caused by the deviation between the pre-set prior knowledge and the actual data distribution and then realize the higher accurate and dataadaptive registration algorithm.Finally,algorithms based on model decoupling and regularization learning are tested on image registration tasks on brain MRI images dataset LPBA40 and lung CT image dataset DIRLAB.The experimental results show that the proposed algorithms can obtain better result than traditional learning methods.