Browse by author
Lookup NU author(s): Dr Colin Kotre
Full text for this publication is not currently held within this repository. Alternative links are provided below where available.
One of the deleterious effects of scattered radiation in the digital radiograph is to add a slowly varying background to the image. This can reduce the ability of the observer to discern low contrast signals if the background gradient over a signal feature prevents the use of a small enough display window to make the signal visible. This paper presents an image processing scheme for suppressing the low spatial frequency effects of scattered radiation in digital radiography and demonstrates it on a range of clinical and phantom images. The approach relies on the approximate separation of high atomic number bony features from the low atomic number soft tissue background, and the use of forward convolution with a scatter kernel to produce an estimate of the scatter distribution arising from the soft tissue background. This is then scaled by an estimate of the soft tissue scatter fraction and subtracted from the original image to produce the final scatter-suppressed image. The implementation employs many approximations in order to make use of information that is readily available in the image headers of current x-ray imaging systems. The performance of the image processing scheme is demonstrated on phantom and clinical images. It is argued that clinical application of the approach could employ a user-controlled scatter subtraction step that would reduce any risk of misinterpretation of the processed image.
Author(s): Kotre CJ
Publication type: Article
Publication status: Published
Journal: Physics in Medicine and Biology
Year: 2016
Volume: 61
Issue: 1
Pages: 280-295
Print publication date: 01/01/2016
Online publication date: 09/12/2015
Acceptance date: 10/11/2015
ISSN (print): 0031-9155
ISSN (electronic): 1361-6560
Publisher: Institute of Physics Publishing Ltd.
URL: http://dx.doi.org/10.1088/0031-9155/61/1/280
DOI: 10.1088/0031-9155/61/1/280
Altmetrics provided by Altmetric
