Evaluation of Deformable Image Registration for Three-Dimensional Temporal Subtraction of Chest Computed Tomography Images

Hindawi International Journal of Biomedical Imaging Volume 2017, Article ID 3457189, 11 pages https://doi.org/10.1155/2017/3457189 Research Article Evaluation of Deformable Image Registration for Three-Dimensional Temporal Subtraction of Chest Computed Tomography Images Ping Yan, Yoshie Kodera, and Kazuhiro Shimamoto Department of Radiological and Medical Laboratory Sciences, Graduate School of Medicine, Nagoya University, 1-1-20 Daiko-Minami, Higashi-ku, Nagoya 461-8673, Japan Correspondence should be addressed to Ping Yan; Received 30 May 2017; Accepted 13 September 2017; Published 12 October 2017 Academic Editor: A. K. Louis Copyright © 2017 Ping Yan et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Purpose. To perform lung image registration for reducing misregistration artifacts on three-dimensional (3D) temporal subtraction of chest computed tomography (CT) images, in order to enhance temporal changes in lung lesions and evaluate these changes after deformable image registration (DIR). Methods. In 10 cases, mutual information (MI) lung mask affine mapping combined with cross-correlation (CC) lung diffeomorphic mapping was used to implement lung volume registration. With advanced normalization tools (ANTs), we used greedy symmetric normalization (greedy SyN) as a transformation model, which involved MI-CC-SyN implementation. The resulting displacement fields were applied to warp the previous (moving) image, which was subsequently subtracted from the current (fixed) image to obtain the lung subtraction image. Results. The average minimum and maximum log-Jacobians were 0.31 and 3.74, respectively. When considering 3D landmark distance, the root-mean-square error changed from an average of 20.82 mm for 𝑃fixed to 𝑃moving to 0.5 mm for 𝑃warped to 𝑃fixed . Clear shadows were observed as enhanced lung nodules and lesions in subtraction images. The lesion shadows showed lesion shrinkage changes over time. Lesion tissue morphology was maintained after DIR. Conclusions. DIR (greedy SyN) effectively and accurately enhanced temporal changes in chest CT images and decreased misregistration artifacts in temporal subtraction images. 1. Introduction Deformable image registration (DIR) is a fundamental task in medical image processing and has many applications. One of the applications is the alignment of chest computed tomography (CT) images from the same subject and, in particular, of the lung and its internal structures for investigations, such as assessment of temporal structural changes or anatomical changes [1]. The applications of the information obtained from three-dimensional (3D) images have been rapidly increasing in the fields of diagnostics and surgical or radiotherapy planning. With the emergence of serial lowdose multidetector CT (MDCT) imaging for lung examination [2], the 3D temporal subtraction technique for chest CT has become a necessity for aiding in the diagnosis of conditions [3–5]. Temporal subtraction of serial volumes has the potential to efficiently identify areas with temporal changes, such as the lung, through approaches, such as visualization of tumor growth or shrinkage on subtraction images [6, 7]. However, because of the potential for misregistration between image volumes, direct subtraction typically does not achieve the desired results. Achieving accurate registration in repeated-interval chest CT scans is necessary to create temporal subtraction images for enhancing images of temporal changes in the lung, because successful registration will result in a perfect fit between lesions at two different time points. A subtraction image from alignment images is often used to evaluate the accuracy of a registration method. Dougherty et al. [6, 7] reported that serial images were aligned with at least a 0.95 correlation in five cases, and the subtraction image showed nodule growth involving an increase in size using the optical flow method. Various image registration algorithms have been proposed for lung images in the literature. Moreover, “Evaluation of Methods for Pulmonary Image Registration 2010” [8] directly compared various different registration algorithms 2 by applying each algorithm to the same set of pairs of thoracic CT images. In this study, we were interested in DIR involving diffeomorphic registration algorithms, which, by definition, preserve topology [9]. Preserving topology means that the structure in the deformed image maintains an adjacent relationship between the internal structures, connectivity is unchanged, tear or paste does not occur, no new structure appears, and the original structure does not disappear. A diffeomorphic transformation is defined as continuous mappings, with one-to-one correspondence between points in one image and points in the second image, and, for every position in one image, there is a signal corresponding position in the second image, with differentiability [10]. Diffeomorphic restriction is valid for a large number of problems in which the two images have the same structures and neighborhood relationships but the structures have different shapes [11]. However, diffeomorphic transformation is required for a geodesic connecting two images, 𝐹 and 𝑀, in the space of diffeomorphic transformations, and the computational and memory costs are significant because of the dense-in-time velocity field calculations and requisite reintegration of the diffeomorphisms after each iterative update [12]. Avants et al. [13] introduced the greedy symmetric normalization (SyN) method as a lower-cost strategy, and the details on mapping images 𝐹 to 𝑀 and 𝑀 to 𝐹 by using the diffeomorphism 0 and greedy optimization in diffeomorphic normalization have been previously described [9, 12, 13]. The study [9] describes and demonstrates two different diffeomorphic transformation methods for 3D image registration. The difference between a time-varying diffeomorphism (greedy SyN) and a diffeomorphism generated by exponential mapping has been described [14, 15], and comparison of both methods of lung image registration showed that the greedy SyN method was able to achieve top performance; however, the local lesion changes of the lung after diffeomorphic transformation were not reported. Diffeomorphic transformation may be used to identify areas where two image volumes differ topologically by analyzing the properties of the resulting transformation, such as when there is a problem of registration of an image with a large lesion to that of an image with a small lesion and when there is a problem of matching an image with a lesion to an image without a lesion [11]. In the present study, the aim was to perform lung volume DIR for the reduction in misregistration artifacts in temporal subtraction images to enhance temporal changes in lung lesions and evaluate the chang (...truncated)