Browse Prior Art Database

A novel method to control noise texture in CT iterative reconstruction

IP.com Disclosure Number: IPCOM000245388D
Publication Date: 2016-Mar-04
Document File: 3 page(s) / 838K

Publishing Venue

The IP.com Prior Art Database

Related People

Jie Tang: INVENTOR

Abstract

Purpose: Iterative reconstruction (IR) is getting more and more attentions in CT for medical imaging in the last decade and has been provided in the product by multiple vendors. Comparing to the conventional filtered-backprojection (FBP) method, IR offers the advantage of higher spatial resolution and lower noise simultaneously in the reconstructed images. Accompanying with the advantage from IR, the noise texture in IR reconstructed images (fine grain appearance) is usually different from FBP, and this unfamiliar noise texture becomes a concern for some radiologists. Thus it is desirable to make the noise texture of IR similar to FBP. Methods: In IR, the projection data is usually not filtered, which enables IR to achieve higher spatial resolution than FBP, but also contributes to the different noise texture. In this work, a new method was proposed to control the noise texture – more specifically, to make the noise texture from IR similar to FBP. In this method, a low-pass filter was applied to the projection data to generate a second projection data set; then both the original projection and the filtered projection were used in IR with a modified objective function, in which the data fidelity term included both data sets. A weighting function was used to emphasize more on original data set in edge regions while more on filtered data set in flat regions. And the low-pass filter was the same as the window function in the FBP filtration kernel. Results: A human subject data was reconstructed using both FBP and the proposed IR method. The IR reconstructed image had sharper edges and lower noise than FBP; meanwhile, the noise texture was similar to the FBP. Conclusions: With the proposed method, IR using dual projection data sets can generate images with similar noise texture to FBP while keeping the high spatial resolution, low noise advantages.

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 52% of the total text.

Purpose: Iterative reconstruction (IR) is getting more and more attentions in CT for medical imaging in the last decade and has been provided in the product by multiple vendors. Comparing to the conventional filtered-backprojection (FBP) method, IR offers the advantage of higher spatial resolution and lower noise simultaneously in the reconstructed images. Accompanying with the advantage from IR, the noise texture in IR reconstructed images (fine grain appearance) is usually different from FBP, and this unfamiliar noise texture becomes a concern for some radiologists. Thus it is desirable to make the noise texture of IR similar to FBP.

Methods: In IR, the projection data is usually not filtered, which enables IR to achieve higher spatial resolution than FBP, but also contributes to the different noise texture. In this work, a new method was proposed to control the noise texture – more specifically, to make the noise texture from IR similar to FBP. In this method, a low-pass filter was applied to the projection data to generate a second projection data set; then both the original projection and the filtered projection were used in IR with a modified objective function, in which the data fidelity term included both data sets. A weighting function was used to emphasize more on original data set in edge regions while more on filtered data set in flat regions. And the low-pass filter was the same as the window function in the FBP filtration kernel.

Results: A human subject data was reconstructed using both FBP and the proposed IR method. The IR reconstructed image had sharper edges and lower noise than FBP; meanwhile, the noise texture was similar to the FBP.

Conclusions: With the proposed method, IR using dual projection data sets can generate images with similar noise texture to FBP while keeping the high spatial resolution, low noise advantages.

A novel method to control noise texture in CT iterative reconstruction

Background: Iterative reconstruction (IR) is getting more and more attentions in CT for medical imaging in the last decade and has been provided in the product by multiple vendors. Comparing to the conventional filtered-backprojection (FBP) method, IR has the advantage of higher spatial resolution and lower noise simultaneously in the reconstructed images. Accompanying with the advantage from IR, the noise texture in IR reconstructed images is usually different from FBP, and this unfamiliar noise texture becomes a concern for some radiologists. Figure 1 shows a sample image reconstructed from a commercial IR product; high frequency (fine grain) noise can be observed. It would be desirable to m...