Helical cone-beam tomography

✍ Scribed by Henrik Turbell; Per-Erik Danielsson

• Book ID: 101259122
• Publisher: John Wiley and Sons
• Year: 2000
• Tongue: English
• Weight: 238 KB
• Volume: 11
• Category: Article
• ISSN: 0899-9457
• DOI: 10.1002/(sici)1098-1098(2000)11:1<91::aid-ima10>3.0.co;2-f

✦ Synopsis

Recently, several computed tomography (CT) machines with multirow detectors have been introduced on the market. Although the projections are obtained from a cone beam rather than a fan beam, multislice reconstruction algorithms in contemporary machines are firmly rooted in two-dimensional (2D) reconstruction of planar objects. Short-scan algorithms are dominating and these are preferably classified according to how redundant data are handled, using parallel rebinning, complementary rebinning, or Parker weighting. In the long run, truly 3D cone-beam algorithms are likely to take over, however. It has proved to be quite a challenge to design an algorithm that is both practical and exact under the constraints set by helical source path geometry. All attempts in this direction are readily seen to be derivatives of Grangeat's algorithm. Alternative research efforts have been building on the nonexact algorithm by Feldkamp et al. (J Opt Soc Am 1: 612-619, 1984). These algorithms strive for simplicity using 1D filtering only. In this overview, we pay special attention to the PI method, which has the unique property that all object points are illuminated by the source during exactly 180°, which generates complete and nonredundant data. Two new algorithms, PI-SLANT and PI-2D, are also presented with some experimental results.