Premium
A new stationary grid, with grid lines aligned to pixel lines with submicron‐order precision, to suppress grid artifacts
Author(s) -
Kawashima Hiroki,
Ichikawa Katsuhiro,
Iida Yasuko
Publication year - 2021
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 - 10.1002/mp.15099
Subject(s) - grid , pixel , artifact (error) , detector , line (geometry) , frequency grid , optics , computer science , physics , artificial intelligence , mathematics , geometry
Purpose We have developed a new stationary grid named a pixel‐aligned grid (PA grid), in which the grid lines are aligned to the pixel lines with submicron‐order precision. Further, we have evaluated its performance relative to that of a conventional grid combined with grid‐line removal (GLR) processing. Methods A flat‐panel detector system of an indirect type, with a pixel pitch of 150 μm, was employed. Four PA grids having a grid ratio of 6:1 associated with abdominal bedside radiography, with the grid‐line pitch (GP) varied around the target value of 150 μm, were produced. Blank images were obtained with four PA grids for measuring the period and amplitude of the grid artifact. In performance evaluation, acrylic and anthropomorphic abdominal phantom images were used with the PA grid, a conventional grid (40 lines/cm, grid ratio 6:1), and no grids. The grid artifacts were evaluated by power spectrum (PS) analysis. Also, the signal‐to‐noise ratio (SNR) improvement factor ( K SNR ) was measured. Results Grid artifacts were hardly recognizable with PA grids with GP errors of 0.3 μm and 0.6 μm because of the prolonged grid artifact periods. The measured artifact amplitudes of these PA grids were less than 0.6%. Furthermore, the PA grids did not produce notable frequency peaks in PS. In contrast, the conventional grid without GLR processing produced two conspicuous peaks. With GLR processing, notable reductions in PS were observed around the two peak frequencies, which caused blurring in bone structures. For the acrylic thickness of 20 cm, the K SNR s for the PA grid were around 1.4, suggesting some SNR improvement in abdominal bedside radiography. Conclusion The present study has demonstrated that PA grids with their grid‐line pitches close to the pixel‐line pitch within errors of 0.6 μm produce grid artifact‐free images without any signal losses. Thus, the proposed PA grid will prove to be effective and useful in various clinical applications.