A Coarse-to-Fine Deformable Transformation Framework for Unsupervised Multi-Contrast MR Image Registration with Dual Consistency Constraint

Abstract

Multi-contrast magnetic resonance (MR) image registration is essential in the clinic to achieve fast and accurate imaging-based disease diagnosis and treatment planning. Nevertheless, the efficiency and performance of the existing registration algorithms can still be improved. In this paper, we propose a novel unsupervised learning-based framework to achieve accurate and efficient multi-contrast MR image registrations. Specifically, an end-to-end coarse-to-fine network architecture consisting of affine and deformable transformations is designed to get rid of both the multi-step iteration process and the complex image preprocessing operations. Furthermore, a dual consistency constraint and a new prior knowledge-based loss function are developed to enhance the registration performances. The proposed method has been evaluated on a clinical dataset that consists of 555 cases, with encouraging performances achieved. Compared to the commonly utilized registration methods, including Voxelmorph, SyN, and LDDMM, the proposed method achieves the best registration performance with a Dice score of 0.826 in identifying stroke lesions. More robust performance in low-signal areas is also observed. With regards to the registration speed, our method is about 17 times faster than the most competitive method of SyN when testing on a same CPU.

Multimodal Remote Sensing Image Registration Methods and Advancements: A Survey

“…This method is superior to other learning-based methods, achieved the highest success rate and Dice coefficient, and has significant robustness for poor quality images. Huang et al [109] proposed a novel unsupervised learning-based framework. This is an end-to-end network architecture composed of affine and deformable transformations.…”

Dual-Flow Transformation Network for Deformable Image Registration with Region Consistency Constraint

“…Learning-based deformable registration methods take advantage of GPU's powerful computing ability to inference the deformation flow displacements (DFFs) to achieve rapid registration [14,31]. Recently, there are many strategies [4,5,16,21,24,31,41,41] regarding deformable image registration. Mok et al [31] proposed a fast symmetric diffeomorphic image registration network (SYMNet), which maximizes the similarity of a pair of images to improve registration accuracy.…”

“…Mok et al [31] proposed a fast symmetric diffeomorphic image registration network (SYMNet), which maximizes the similarity of a pair of images to improve registration accuracy. Huang et al [21] proposed a coarse-to-fine network architecture consisting of affine and deformable transformations (ACNet) to achieve accurate registration with global similarity constraint. While these methods learn the all anatomies structure registration to maximize the global similarity of a pair of images, which easily leads to mis-alignment on regions of interest (ROIs) within images.…”

“…These methods achieve rapid deformable image registration, making possible the realtime registration in practical applications. However, these approaches [21,31,38,39,41] only learn global correspondence of a pair of images to obtain displacement flow fields (DFFs) and ignore the similarity of regions of interest (ROIs) within images. In addition, these methods only use a global similarity metric to evaluate the similarity of a pair of images, which cannot maximize the similarity of ROIs within a pair of images.…”

Toward Computing Cross-Modality Symmetric Non-Rigid Medical Image Registration

“…In recent years, machine learning (ML) and deep learning (DL) have vigorously gained traction in medical image registration. Huang et al [22] present an unsupervised learning-based framework, where the network consists of affine and deformable transformations. Hansen et al [23] propose a sparse key point-based geometric network that leverages discrete dense displacement maps facilitating the registration process.…”