MiRID: Multi-Modal Image Registration Using Modality-Independent and Rotation-Invariant Descriptor

Thuvanan Borvornvitchotikarn,Werasak Kurutach

doi:10.3390/sym12122078

Thuvanan Borvornvitchotikarn, Werasak Kurutach

Open Access

https://doi.org/10.3390/sym12122078

Copy DOI

Journal: Symmetry	Publication Date: Dec 15, 2020
Citations: 1	License type: CC BY 4.0

Affiliation: Mahanakorn University of Technology

Abstract

Axiomatically, symmetry is a fundamental property of mathematical functions defining similarity measures, where similarity measures are important tools in many areas of computer science, including machine learning and image processing. In this paper, we investigate a new technique to measure the similarity between two images, a fixed image and a moving image, in multi-modal image registration (MIR). MIR in medical image processing is essential and useful in diagnosis and therapy guidance, but still a very challenging task due to the lack of robustness against the rotational variance in the image transformation process. Our investigation leads to a novel, local self-similarity descriptor, called the modality-independent and rotation-invariant descriptor (miRID). By relying on the mean of the intensity values, an miRID is simply computable and can effectively handle the complicated intensity relationship between multi-modal images. Moreover, it can also overcome the problem of rotational variance by sorting the numerical values, each of which is the absolute difference between each pixel’s intensity and the mean of all pixel intensities within a patch of the image. The experimental result shows that our method outperforms others in both multi-modal rigid and non-rigid image registrations.

Highlights

Over the last decade, medical imaging has been an important tool in clinical practice and many biomedical studies [1]
Borvornvitchotikarn and Kurutach [26] proposed a robust self-similarity descriptor (RSSD), which enhances modality-independent local binary pattern (miLBP) in terms of rotational variance and improves the accuracy of modal image registration (MIR). Both the miLBP and RSSD methods assess similarity based on the use of the central pixel and, any image artifact in that pixel can affect the performance of the descriptors
In order to demonstrate the efficiency of the proposed method, we employed simulated longitudinal relaxation time (T1), transverse relaxation time (T2) and proton density (PD) brain magnetic resonance (MR) images with an image size of 181 × 217 × 181 voxels, 3% noise and 40% intensity non-uniformity from the BrainWeb dataset [30]

Summary

Introduction

Medical imaging has been an important tool in clinical practice and many biomedical studies [1]. Heinrich et al [20] proposed a method called the modality independent neighbourhood descriptor (MIND) as a local self-similarity measure based on neighborhood information. Borvornvitchotikarn and Kurutach [26] proposed a robust self-similarity descriptor (RSSD), which enhances miLBP in terms of rotational variance and improves the accuracy of MIR. Both the miLBP and RSSD methods assess similarity based on the use of the central pixel and, any image artifact in that pixel can affect the performance of the descriptors.

Registration Framework

Experimental Results and Discussion

Conclusions