Premium
TU‐A‐213CD‐04: A Novel Technique for Estimating and Correcting Cross‐Scatter in Dual‐Source CBCT
Author(s) -
Giles W,
Bowsher J,
Li H,
Yin F
Publication year - 2012
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4735883
Subject(s) - imaging phantom , nyquist frequency , sampling (signal processing) , optics , projection (relational algebra) , nyquist–shannon sampling theorem , physics , contrast (vision) , mathematics , computer science , algorithm , computer vision , detector , filter (signal processing) , mathematical analysis
Purpose: Dual‐source CBCT may provide several advantages over current on‐board 3D localization techniques; however, dual‐source CBCT suffers greatly from cross‐scatter. A novel technique was developed that includes sampling the cross‐scatter only projections (CSPs) during data acquisition and using the CSPs for scatter correction. Methods: Cross‐scatter only projections (CSPs) were acquired by firing a single x‐ray tube every N degrees. N was determined based on estimating the angular Nyquist frequency of the cross‐scatter distribution for three different anatomic sites (head and neck, lung, pelvis) using an anthropomorphic phantom. The Nyquist frequency was estimated by calculating the divergence of angular frequency spectra of rotating phantoms from the noise band created with stationary phantoms. The sampling requirements were empirically verified by varying N and comparing interpolated CSPs with measured CSPs at every projection angle. To correct for cross‐scatter, the CSPs were low‐pass filtered, then interpolated across projection angle, and then subtracted from the projections. The effects of the correction on contrast and contrast‐to‐noise ratio (CNR) in reconstructed images were evaluated on an equal‐dose basis relative to acquisitions with no cross‐scatter present. Results: Approximately the same Nyquist frequency was estimated for all three anatomical sites, 0.035–0.042 cycles per degree, corresponding to somewhat less than 1 sample every 10 degrees. The percent contrast reduction, averaged over several contrast inserts, was 16.4% and 4.5% in the uncorrected and corrected images. CNR was reduced by 19.5% and 13.6%. Conclusions: For the phantom studied, the cross‐scatter Nyquist sampling requirement was insensitive to anatomic site. The scatter correction mitigated cross‐scatter and CNR degradations relative to cross‐scatter‐free acquisition, improving both contrast and CNR as compared to uncorrected images reconstructed from cross‐scatter‐inclusive dual‐source acquisition. This project was partially supported by a research grant from Varian Medical Systems