Browse Prior Art Database

MATCHING OF NON-OVERLAPPING MULTIMODAL CORONAL SLICES BY FAST DIFFERENCING OF BINARY SIGNATURE IN FOURIER DOMAIN

IP.com Disclosure Number: IPCOM000236783D
Publication Date: 2014-May-15
Document File: 8 page(s) / 155K

Publishing Venue

The IP.com Prior Art Database

Abstract

The invention proposes a technique that includes an algorithm to determine overlap region in a multimodal coronal scan to serve as an input to multimodal registration in magnetic resonance (MR) system. The algorithm morphs the multimodal registration as a 3D template matching, thereby, utilizing non-local binary signature on gradient images. Template matching is a highly computationally intensive process that includes a technique of computing difference of binary signature in a Fourier domain. Computational process is applicable to real time processing.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 46% of the total text.

MATCHING OF NON-OVERLAPPING MULTIMODAL CORONAL SLICES BY FAST DIFFERENCING OF BINARY SIGNATURE IN FOURIER DOMAIN

 

BRIEF ABSTRACT

The invention proposes a technique that includes an algorithm to determine overlap region in a multimodal coronal scan to serve as an input to multimodal registration in magnetic resonance (MR) system. The algorithm morphs the multimodal registration as a 3D template matching, thereby, utilizing non-local binary signature on gradient images. Template matching is a highly computationally intensive process that includes a technique of computing difference of binary signature in a Fourier domain. Computational process is applicable to real time processing.

KEYWORDS

 

Template matching, coronal slices, MR scan, CT scan, MR system

DETAILED DESCRIPTION

In order to identify matching anatomical portion of whole body, CT scan for a provided MR coronal scan requires registration. MR scan is anatomy specific, where anatomy includes neurological, neck, lung, abdomen, pelvic among others. Registration attempts to match a MR lung to a CT lung and CT abdomen, provides stretching effect resulting in an incorrect registration.

A conventional technique includes field of view correction by enhancing maximum similarity between target and reference volumes. However, the conventional technique is not efficient in matching field of view.

Another conventional technique includes a complete sum-of-squares difference (CSSD) metric graph matching or a histogram based matching approach for registration. However, the conventional technique does not provide descriptions of image as spatial location are not encoded.

           

Therefore, there is a need for a technique for efficient template matching by different metric in MR system.

The invention proposes a technique that includes an algorithm to determine overlap region in a multimodal coronal scan to serve as an input to multimodal registration in magnetic resonance (MR) system. The algorithm morphs the multimodal registration as a 3D template matching, thereby, utilizing non-local binary signature on gradient images. Template matching is a highly computationally intensive process that includes a technique of computing difference of binary signature in a Fourier domain. Computational process is applicable to real time processing.

The algorithm includes steps to determine overlap region in multimodal coronal scan in the MR system. For a corresponding MR and a CT volume, a coronal slice is identified in each volume, which describes same anatomy in anterior-posterior direction and is maximally descriptive. The mid MR and the mid CT coronal slices are sufficient for identification. FIGURE 1 depicts matching block for the coronal MRI image with coronal whole body CT. After acquisition of coronal slices, matching corresponding anatomies recast as a template search issue, where aim is to identify exact matching sub-portion in the whole body CT coronal slices for the given MR template or the c...