Document Type: Original Paper
Department of Medical Physics
Sefako Makgatho Health Sciences University
Introduction:Low-pass filters eliminate noise, and accordingly improve the quality of filtered back-projection (FBP) in the reconstruction of single photon emission computed tomography (SPECT) images. This study aimed at selection of an optimum low-pass filter for FBP reconstruction of SPECT images of small structures.
Material and Methods:Spheres A, B, and C (16 mm, 12 mm, and 11 mm in diameter, respectively) attached to capillary stems were filled with technetium-99m solution (activity concentration 300 kBq/mL). They were then mounted inside a Jaszczak Phantom forming a V-shaped structure. The phantom was then filled with distilled water. Two-dimensional (2D) projections were acquired on 128 128 pixels using a Siemens E-Cam dual-head gamma camera. The Parzen, Shepp-Logan, Low Pass Cosine filters (cut-off frequencies: 0.2-0.9), and Butterworth filter (order: 1-9; cut-off frequencies: 0.3-0.9) were employed during FBP reconstruction. The line command of ImageJ software was used to draw the point spread functions of acquired 2D transaxial central slices and for the measurement of their full-width at half-maximum (FWHM).
Results:The FWHM of 2D central image slices of spheres A, B, and C reconstructed using a Butterworth filter measured 20, 20, and 10 pixels, respectively. In comparison, the reconstructed images using the Parzen, Low Pass Cosine, and Shepp-Logan filters measured to 27, 25 and, 22 pixels for sphere A, 24, 22, 20 pixels for B, and 22, 20, 18 pixels for C, respectively.
Conclusion: The low-pass filters successfully suppressed noise during the FBP reconstruction of SPECT images of small structures. Accordingly, the Butterworth is a suitable choice.