Deep learning of deformable registration for breast DCE-MRI images

Conventional image registration approaches optimize an objective function for each pair of images, which can be time-consuming for a large dataset. In this work, we proposed a deep learning method for breast DCE-MRI images that eliminates the need for time consuming iterative methods, and directly generates the registered image with the deformation field. The model is trained by optimizing the similarity measurement between original images and distorted motion images without manual annotation information from doctors. Enhanced image is distorted into the original image through spatial transformation network to obtain the registration result. Our method can speed up medical image analysis, while facilitating novel directions in deep learning-based registration and its applications. The experimental results show that the model is effective and robust for breast images.