MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

A New Method for Denoising and Enhancement of Megavoltage CT Based On Dictionary Learning

c yue1*, J Zhu2 , (1) Shandong Provincial Key Laboratory of Network Based Intelligent Computing, Jinan, Shizhong, (2) Shandong Cancer Hospital to Shandong University, Shandong Academy of Medica, Jinan City, shandong provice

Presentations

(Sunday, 7/29/2018) 3:00 PM - 6:00 PM

Room: Exhibit Hall

Purpose: Due to the Compton Effect of megavoltage CT (MVCT) during imaging, which results in a large amount of noise in the image and low soft tissue contrast, so MVCT images are only used for patient position correction currently. Registration, delineation, and adaptive radiotherapy are Restricted. This article aims to improve the soft tissue contrast by using a combination of Discriminative feature representation and 3D block matching algorithms

Methods: MVCT images of four patients were obtained, including Head and neck and abdomens and pelvic and chest image of a case. Four pairs of noisy MVCT images are denoised and enhanced respectively by using the method of Discriminative feature representation (DFR) and 3D block matching algorithms (BM3D)..

Results: MVCT images combined with post-processing by applying (DFR) and (BM3D) show a significant improvement in soft tissue contrast. In Head and neck images, skull becomes noticeable in this method. In abdomens and pelvic and chest images, there is a clear boundary between soft tissues. The CNRs of both regions of interest in the Head and neck patients improved from 4.40and 6.51to 6.51and11.16, respectively. The CNRs of both regions of interest in the abdomen patients improved from 0.98and 1.09 to 3.34 and 3.45, respectively. The CNRs of both regions of interest in the pelvic patients improved from 1.60and 2.51 to 2.38 and 3.83, respectively. The CNRs of both regions of interest in the chest patients improved from 0.96and 0.99 to 1.42 and 1.61, respectively. The result is far superior to the Discriminative feature representation.

Conclusion: Discriminative feature representation and 3D block matching algorithms can not only preserve the original details of the image but also improve the visual effect of the image while using MVCT for image guided radiotherapy and self- Adapt to radiotherapy to lay the foundation.

Keywords

Image Analysis, CT

Taxonomy

IM- CT: Theory

Contact Email