当前位置： SCI文献检索 > PHYSICS IN MEDICINE AND BIOLOGY期刊下所有文献 > Evaluation of robustness of maximum likelihood cone-beam CT reconstruction with total variation regularization.

Evaluation of robustness of maximum likelihood cone-beam CT reconstruction with total variation regularization.

Abstract：

:The objective of this paper is to evaluate an iterative maximum likelihood (ML) cone-beam computed tomography (CBCT) reconstruction with total variation (TV) regularization with respect to the robustness of the algorithm due to data inconsistencies. Three different and (for clinical application) typical classes of errors are considered for simulated phantom and measured projection data: quantum noise, defect detector pixels and projection matrix errors. To quantify those errors we apply error measures like mean square error, signal-to-noise ratio, contrast-to-noise ratio and streak indicator. These measures are derived from linear signal theory and generalized and applied for nonlinear signal reconstruction. For quality check, we focus on resolution and CT-number linearity based on a Catphan phantom. All comparisons are made versus the clinical standard, the filtered backprojection algorithm (FBP). In our results, we confirm and substantially extend previous results on iterative reconstruction such as massive undersampling of the number of projections. Errors of projection matrix parameters of up to 1° projection angle deviations are still in the tolerance level. Single defect pixels exhibit ring artifacts for each method. However using defect pixel compensation, allows up to 40% of defect pixels for passing the standard clinical quality check. Further, the iterative algorithm is extraordinarily robust in the low photon regime (down to 0.05 mAs) when compared to FPB, allowing for extremely low-dose image acquisitions, a substantial issue when considering daily CBCT imaging for position correction in radiotherapy. We conclude that the ML method studied herein is robust under clinical quality assurance conditions. Consequently, low-dose regime imaging, especially for daily patient localization in radiation therapy is possible without change of the current hardware of the imaging system.

journal_name

Phys Med Biol

journal_title

Physics in medicine and biology

authors

Stsepankou D,Arns A,Ng SK,Zygmanski P,Hesser J

doi

10.1088/0031-9155/57/19/5955

subject

Has Abstract

pub_date

2012-10-07 00:00:00

pages

5955-70

issue

eissn

0031-9155

issn

1361-6560

journal_volume

pub_type

杂志文章

在线工具

Evaluation of robustness of maximum likelihood cone-beam CT reconstruction with total variation regularization.