PERFORM Centre, Concordia University, Montreal, Canada.
Department of Electrical and Computer Engineering, Concordia University, Montreal, Canada.
Int J Comput Assist Radiol Surg. 2021 Apr;16(4):555-565. doi: 10.1007/s11548-021-02323-2. Epub 2021 Mar 8.
Accurate multimodal registration of intraoperative ultrasound (US) and preoperative computed tomography (CT) is a challenging problem. Construction of public datasets of US and CT images can accelerate the development of such image registration techniques. This can help ensure the accuracy and safety of spinal surgeries using image-guided surgery systems where an image registration is employed. In addition, we present two algorithms to register US and CT images.
We present three different datasets of vertebrae with corresponding CT, US, and simulated US images. For each of the two latter datasets, we also provide 16 landmark pairs of matching structures between the CT and US images and performed fiducial registration to acquire a silver standard for assessing image registration. Besides, we proposed two patch-based rigid image registration algorithms, one based on normalized cross-correlation (NCC) and the other based on correlation ratio (CR) to register misaligned CT and US images.
The CT and corresponding US images of the proposed database were pre-processed and misaligned with different error intervals, resulting in 6000 registration problems solved using both NCC and CR methods. Our results show that the methods were successful in aligning the pre-processed CT and US images by decreasing the warping index.
The database provides a resource for evaluating image registration techniques. The simulated data have two applications. First, they provide the gold standard ground-truth which is difficult to obtain with ex vivo and in vivo data for validating US-CT registration methods. Second, the simulated US images can be used to validate real-time US simulation methods. Besides, the proposed image registration techniques can be useful for developing methods in clinical application.
术中超声（US）与术前计算机断层扫描（CT）的精确多模态配准是一个具有挑战性的问题。构建 US 和 CT 图像的公共数据集可以加速此类图像配准技术的发展。这有助于确保使用图像引导手术系统进行脊柱手术的准确性和安全性，其中使用图像配准。此外，我们提出了两种将 US 和 CT 图像配准的算法。
我们提出了三种不同的数据集，包含相应的 CT、US 和模拟 US 图像。对于后两个数据集中的每一个，我们还提供了 CT 和 US 图像之间匹配结构的 16 对地标对，并进行了基准配准，以获得评估图像配准的银标准。此外，我们提出了两种基于补丁的刚性图像配准算法，一种基于归一化互相关（NCC），另一种基于相关比（CR），以配准未对准的 CT 和 US 图像。
所提出的数据库的 CT 和相应的 US 图像经过预处理，并以不同的误差间隔未对准，使用 NCC 和 CR 方法解决了 6000 个配准问题。我们的结果表明，这些方法通过降低变形指数成功地对齐了预处理的 CT 和 US 图像。
该数据库为评估图像配准技术提供了资源。模拟数据有两个应用。首先，它们提供了黄金标准的真实值，这对于验证 US-CT 配准方法来说很难从离体和体内数据中获得。其次，模拟 US 图像可用于验证实时 US 模拟方法。此外，所提出的图像配准技术可用于开发临床应用中的方法。
