Affiliations: Department of Radiation Oncology, The Netherlands
Cancer Institute, Amsterdam, The Netherlands
Note: [] Corresponding author: Jan-Jakob Sonke, Department of Radiation
Oncology, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX Amsterdam,
The Netherlands. E-mail: [email protected]
Abstract: Cone-beam (CB) computed tomography (CT) is widely used in the field
of medical imaging for guidance. Inspired by Betram's directional interpolation
(BDI) methods, directional sinogram interpolation (DSI) was implemented to
generate more CB projections by optimized (iterative) double-orientation
estimation in sinogram space and directional interpolation. A new CBCT was
subsequently reconstructed with the Feldkamp algorithm using both the original
and interpolated CB projections. The proposed method was evaluated on both
phantom and clinical data, and image quality was assessed by correlation ratio
(CR) between the interpolated image and a gold standard obtained from full
measured projections. Additionally, streak artifact reduction and image blur
were assessed. In a CBCT reconstructed by 40 acquired projections over an arc
of 360 degree, streak artifacts dropped 20.7% and 6.7% in a thorax phantom,
when our method was compared to linear interpolation (LI) and BDI methods.
Meanwhile, image blur was assessed by a head-and-neck phantom, where image blur
of DSI was 20.1% and 24.3% less than LI and BDI. When our method was compared
to LI and DI methods, CR increased by 4.4% and 3.1%. Streak artifacts of
sparsely acquired CBCT were decreased by our method and image blur induced by
interpolation was constrained to below other interpolation methods.
