Mutifractals based multimodal 3D image registration. (January 2019)

• Record Type: Journal Article
• Authors: Palanivel, Dhevendra Alagan; Natarajan, Sivakumaran; Gopalakrishnan, Sainarayanan
• Is Part Of: Biomedical signal processing and control. Volume 47(2019)
• Page Start: 126
• Page End: 136
• Publication Date: 2019-01
• Subjects: Multifractal -- Multimodal image registration -- Mutual information; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2018.08.015
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: Multifractal characterization of 3D multimodal images for registration. Evaluation of different multifractal measures for output image quality. Analysis of multifractal spectra to understand the accuracy performance. Extensive validation done using simulated images and real brain images. Abstract: Multimodal registration is a method to register the volumes of different modalities, for e.g., computed tomography (CT) and magnetic resonance (MR). Mutual information (MI) based methods are widely used for multimodal registration. The MI characterizes the statistical dependence between the voxel intensities of volumes. Robustness of the MI based registration is affected, when there is a low correspondence between the voxel intensities of volumes. This can be improved by integrating the geometric characteristics of volumes like complexity, singularity and irregularity with registration. A novel approach for 3D multimodal image registration based on the multifractal characterization of volumes is being proposed in this paper. The proposed method uses multifractal formalism to incorporate geometric characteristics into registration. Multifractal formalism involves determination of Holder exponent followed by computation of Hausdorff dimension. Holder exponents quantify the local regularity of the volumes and Hausdorff dimensions quantify the global regularity (multifractality) of the volumes. The performance of the proposed algorithm is evaluated using synthetic phantom images for different noise levels and 41 clinical 3D brain images of 7 different patients from a public domain database. The above-mentioned test platforms highlight the efficiency of the proposed method towards improving the robustness and accuracy of registration.