背景:图像配准在许多临床任务中是一个具有挑战性的问题,但是在过去的几年中,深度学习在这一领域取得了重大进展。通过监督变换估计,可以实现实时和鲁棒的配准。然而,使用此框架的注册质量取决于诸如位移场之类的地面实况标签的质量。
目的:提出一种简单可靠的方法,用于以完全无监督的方式基于图像结构相似性配准医学图像。
方法:我们提出了一种深度级联无监督可变形配准方法,用于在没有可靠临床数据标签的情况下对齐图像。我们的基本网络由位移估计模块(ResUnet)和变形模块(空间变换器层)组成。我们采用l2$l_2$范数正则化变形场,而不是传统的l1$l_1$范数正则化。此外,我们在训练阶段利用结构相似性(ssim)估计来增强变形图像和参考图像之间的结构一致性。
结果:实验结果表明,通过结合ssim损失,我们的级联方法不仅在CT图像上获得了更高的骰子得分0.9873,ssim得分0.9559,归一化互相关(NCC)得分0.9950和更低的相对平方差总和(SSD)误差0.0313,但也优于超声数据集上的比较方法。统计t$t$检验结果也证明了我们方法的这些改进具有统计学意义。
结论:在这项研究中,基于不同评估指标的有希望的结果表明,我们的模型在可变形图像配准(DIR)中简单有效。通过肝脏CT图像和心脏超声图像的实验验证了模型的泛化能力。
BACKGROUND: Image registration is a challenging problem in many clinical tasks, but deep learning has made significant progress in this area over the past few years. Real-time and robust registration has been made possible by supervised transformation estimation. However, the quality of registrations using this framework depends on the quality of ground truth labels such as displacement field.
OBJECTIVE: To propose a simple and reliable method for registering medical images based on image structure similarity in a completely unsupervised manner.
METHODS: We proposed a deep cascade unsupervised deformable registration approach to align images without reliable clinical data labels. Our basic network was composed of a displacement estimation module (ResUnet) and a deformation module (spatial transformer layers). We adopted l 2 $l_2$ -norm to regularize the deformation field instead of the traditional l 1 $l_1$ -norm regularization. Additionally, we utilized structural similarity (ssim) estimation during the training stage to enhance the structural consistency between the deformed images and the reference images.
RESULTS: Experiments results indicated that by incorporating ssim loss, our cascaded methods not only achieved higher dice score of 0.9873, ssim score of 0.9559, normalized cross-correlation (NCC) score of 0.9950, and lower relative sum of squared difference (SSD) error of 0.0313 on CT images, but also outperformed the comparative methods on ultrasound dataset. The statistical t $t$ -test results also proved that these improvements of our method have statistical significance.
CONCLUSIONS: In this study, the promising results based on diverse evaluation metrics have demonstrated that our model is simple and effective in deformable image registration (DIR). The generalization ability of the model was also verified through experiments on liver CT images and cardiac ultrasound images.