Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

DE-NOISING GATED POSITRON EMISSION TOMOGRAPHY IMAGES

IP.com Disclosure Number: IPCOM000242266D
Publication Date: 2015-Jun-30
Document File: 4 page(s) / 392K

Publishing Venue

The IP.com Prior Art Database

Abstract

A technique to obtain a de-noised static positron emission tomography (PET) image from gated images is disclosed. The technique includes nonlocal means (NLM) to obtain high quality de-noised PET image from gated images. In order to de-noise a current pixel, several correlated pixel candidates are aggregated in neighboring gates, thereby, dramatically improving signal to noise ratio (SNR) of the final de-noised image. Also, a patch based metric is used within de-noising to identify correlated pixels, which ensures that artifacts do not occur due to large motion or noise.

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 49% of the total text.

DE-NOISING GATED POSITRON EMISSION TOMOGRAPHY IMAGES

BACKGROUND

 

The present disclosure relates generally to positron emission tomography (PET) and more particularly to a technique for obtaining a de-noised static PET image from gated PET images.

Movements such as those caused by respiration cardiac activity degrade image quality in positron emission tomography (PET) acquisitions. A common approach used for motion correction is Reconstruct-Register-Average (RRA). RRA involves gating PET data, reconstructing images obtained, registering each image to a reference gate and averaging the registered images. Registration technique can provide quantitatively more accurate PET images compared to an ungated cases, while using entire data. Therefore, RRA images have better counting statistics than single gates. However, in presence of count induced noise, and large motion, non-rigid registration may introduce artifacts in the gated images and finally degrade quality of an average image.

Conventionally, standard image or video de-noising methods, for example, linear or non-linear filtering, total variation based method, among others are used for de-noising PET data. One other conventional de-noising technique utilizes non-local means and refinements to obtain a high signal to noise ratio and improved image quality. However, these conventional techniques are challenged by noise or large gate motion.

It would be desirable to have an improved technique that provides a de-noised static PET image from gated images.

 

BRIEF DESCRIPTION OF DRAWINGS

Figure 1 is a block diagram depicting Non-Local (NL) de-noising means to obtain a de-noised static PET image from gated images.

Figure 2 depicts comparative images obtained with RRA, the first approach and the second approach of the technique using NLM.

DETAILED DESCRIPTION

A technique to obtain a de-noised static positron emission tomography (PET) image from gated images is disclosed. In contrast to conventional non-rigid registration (NRR) technique, according to the technique described herein, to de-noise a current pixel, several correlated pixel candidates are aggregated in neighboring gates, thereby, dramatically improving signal to noise ratio (SNR) of the final de-noised image. Also, a patch based metric is used within de-noising to identify correlated pixels, which ensures that artifacts do not occur due to large motion or noise. Formulation of non-local means (NLM) used for de-noising, according to the technique described herein, is given by equation 1:

                (1)

where w(x,y) = e-d(PIx,PIk)  is a distance between image I patches, PIx, PIy.  Figure 1 is a block diagram depicting non-local (NL) de-noising means to obtain a de-noised static PET image from gated images. Two approaches are described herein, for performing 4D de-noising on gated PET data, using a reference gate to define the domain for the final de-noised volume.

Figure 1

According to a first approach, for each pixel...