Reconstruction for musculoskeletal tomosynthesis: A comparative study using image quality assessment in image and projection domain

Y. M. Levakhina, B. Kratz, R. L. Duschka, F. Vogt, J. Barkhausen, T. M. Buzug

Abstract

Digital tomosynthesis (DT) is an X-ray limited angle tomographic technique for producing an arbitrary number of cross-sectional slices of an object. Due to the limited tomographic nature and a violation of the tomographic sufficiency conditions, an accurate image reconstruction in DT is a challenging task. In this work, the Simultaneous Algebraic Reconstruction Technique (SART) for the musculoskeletal tomosynthesis imaging has been investigated. The purpose of this study is twofold. First, to determine SART optimal parameters, which deliver images with less artifacts. Second, to propose an automatic no-reference image quality measure in order to compare reconstructed volumes when varying reconstruction parameters. Several no-reference metrics were tested in both, the image and the projection domain. Measurements have been carried out on two hands from body donors using the Siemens Mammomat Inspiration device. All tested metrics show good agreement in their results. Based on the comparison between metric results, an optimal combination of reconstruction parameters has been selected.

OriginalspracheEnglisch
Titel 2011 IEEE Nuclear Science Symposium Conference Record
Seitenumfang3
Herausgeber (Verlag)IEEE
Erscheinungsdatum01.01.2011
Seiten2569-2571
Aufsatznummer3 6152693
ISBN (Print)978-1-4673-0118-3 , 978-1-4673-0119-0
ISBN (elektronisch)978-1-4673-0120-6
DOIs
PublikationsstatusVeröffentlicht - 01.01.2011
Veranstaltung2011 IEEE Nuclear Science Symposium and Medical Imaging Conference
- Valencia, Spanien
Dauer: 23.10.201129.10.2011
Konferenznummer: 88938

Fingerprint

Untersuchen Sie die Forschungsthemen von „Reconstruction for musculoskeletal tomosynthesis: A comparative study using image quality assessment in image and projection domain“. Zusammen bilden sie einen einzigartigen Fingerprint.

Zitieren