Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

Performance measures in the selection of reconstruction filters for SPECT imaging

Conference · · J. Nucl. Med.; (United States)
OSTI ID:6760163

Filter selection in SPECT image reconstruction poses an implicit tradeoff between image smoothness, image contrast and noise texture. It is known that the Ramachandran or ramp filter provides the greatest contrast (and resolution) in a reconstructed image at the expense of poor noise handling. In order to improve detection of either hot or cold lesions, practical experience dictates that some image smoothing must be provided. Thus, the question of the ''optimal'' filter selection has been the subject of this investigation. Using an extension of the approach as originally developed by Beck and Metz, the authors derived a figure-of-merit (FOM) that quantifies the performance of the reconstruction algorithm being tested. The FOM can be separated into two components: smoothness factor (SF) which is a measure of noise reduction and contrast factor (CF) which is a measure of the spatial resolution of the filter being tested. Each component is measured separately. The optimum filter should maintain the contrast of the Ramachandran filter (CF = 0.313) but improve the noise handling (SF = 8.86). A variety of commercially available filters were tested: rectangular, Hann, and Butterworth. Also examined were the order of the filter and the interpolation method (nearest-neighbor, linear, and Fourier methods with padding of zeros). The authors found that high order Butterworth filters performed best in maintaining contrast but reducing noise. Using a cutoff frequency of 0.200 (Nyquist = 0.500) and an order of 30, the filter provided a CF = 0.313 and a SF = 30.31. Linear interpolation was marginally better.

Research Organization:
Indiana Univ. School of Medicine, Indianapolis, IN
OSTI ID:
6760163
Report Number(s):
CONF-850611-
Journal Information:
J. Nucl. Med.; (United States), Journal Name: J. Nucl. Med.; (United States) Vol. 26:5; ISSN JNMEA
Country of Publication:
United States
Language:
English