Premium
WE‐G‐18A‐08: Axial Cone Beam DBPF Reconstruction with Three‐Dimensional Weighting and Butterfly Filtering
Author(s) -
Tang S,
Wang W,
Tang X
Publication year - 2014
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 - 10.1118/1.4889519
Subject(s) - streak , weighting , projection (relational algebra) , computer science , algorithm , computer vision , iterative reconstruction , artificial intelligence , truncation (statistics) , artifact (error) , mathematics , optics , physics , machine learning , acoustics
Purpose: With the major benefit in dealing with data truncation for ROI reconstruction, the algorithm of differentiated backprojection followed by Hilbert filtering (DBPF) is originally derived for image reconstruction from parallel‐ or fan‐beam data. To extend its application for axial CB scan, we proposed the integration of the DBPF algorithm with 3‐D weighting. In this work, we further propose the incorporation of Butterfly filtering into the 3‐D weighted axial CB‐DBPF algorithm and conduct an evaluation to verify its performance. Methods: Given an axial scan, tomographic images are reconstructed by the DBPF algorithm with 3‐D weighting, in which streak artifacts exist along the direction of Hilbert filtering. Recognizing this orientation‐specific behavior, a pair of orthogonal Butterfly filtering is applied on the reconstructed images with the horizontal and vertical Hilbert filtering correspondingly. In addition, the Butterfly filtering can also be utilized for streak artifact suppression in the scenarios wherein only partial scan data with an angular range as small as 270° are available. Results: Preliminary data show that, with the correspondingly applied Butterfly filtering, the streak artifacts existing in the images reconstructed by the 3‐D weighted DBPF algorithm can be suppressed to an unnoticeable level. Moreover, the Butterfly filtering also works at the scenarios of partial scan, though the 3‐D weighting scheme may have to be dropped because of no sufficient projection data are available. Conclusion: As an algorithmic step, the incorporation of Butterfly filtering enables the DBPF algorithm for CB image reconstruction from data acquired along either a full or partial axial scan.