MRI Image Registration: Data-Driven Approach
Image registration is the challenging and important step to build computer-based diagnostic systems. One type of image modality is not able to provide all information for better diagnostic, and we combine information from multiple sources/image modalities
- PDF / 125,101 Bytes
- 6 Pages / 439.37 x 666.142 pts Page_size
- 97 Downloads / 194 Views
Abstract Image registration is the challenging and important step to build computer-based diagnostic systems. One type of image modality is not able to provide all information for better diagnostic, and we combine information from multiple sources/image modalities. In this work, we proposed a canonical correlation analysis (CCA)-based image registration approach. CCA provides us framework to combine information from multiple sources. In this work, we use the information presents in both images for image registration task. We perform multimodal registration on T1-weighted, T2-weighted, and FLAIR MRI images. We use public data sets to evaluate our algorithm. Our algorithm performs better with mutual information–based image registration approach. Keywords Image registration
CCA MRI
1 Introduction The goal of registration is to estimate the deformation between the images while taking the domain-specific information into consideration. A closer look at the problem statement intuitively reveals two methods of solving it. The first method operates directly on image intensity values, continuously transforming the entire
C. Hemasundara Rao (&) JNTUK, Guru Nanak Institutions Technical Campus, Ibrahimpatnam, Telegana State, India e-mail: [email protected] P.V. Naganjaneyulu ECE, MVR College of Engineering and Technology, Paritala, Krishna Dist, AP, India e-mail: [email protected] K. Satyaprasad ECE, JNTUK, University College of Engineering, Kakinada, AP, India e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2018 H.S. Saini et al. (eds.), Innovations in Electronics and Communication Engineering, Lecture Notes in Networks and Systems 7, https://doi.org/10.1007/978-981-10-3812-9_33
319
320
C. Hemasundara Rao et al.
image so as to align it with the other. The image is considered to be registered when desirable alignment is obtained for the respective transformation. These methods are called area-based methods [1–3]. The second method relies on a few salient points which are most prominent in both the images. The goal here is to detect the corresponding pairs of points/regions across the images from which the deformation is estimated. These are known as feature-based methods. Feature-based methods have gained popularity over the area-based methods as they are more robust to illumination changes, partial overlap between the images, occlusion, changes in background, and viewpoint. Despite these advantages, area-based methods are still preferred over feature-based methods in the medical domain due to two main factors: (1) its ability to handle local deformations, which is especially the case with human organs, and (2) dealing with information from different imaging sources. In general, the process of image registration involves finding the optimal geometric transformation that maximizes the correspondences across the images. This involves the following components (see Fig. 1). A transformation model defines a geometric transformation between the images. There are several classes of nonrigid tr
Data Loading...
