Abstract

Objectives. SPECT/slow-rotation low-output CT systems can produce streak artifacts in filtered backprojection (FBP) attenuation maps, impacting attenuation correction (AC) in myocardial perfusion imaging. This paper presents an adaptive Bayesian iterative transmission reconstruction (ABITR) algorithm for more accurate AC. Methods. In each iteration, ABITR calculated a three-dimensional prior containing the pixels with attenuation coefficients similar to water, then used it to encourage these pixels to the water value. ABITR was tested with a cardiac phantom and 4 normal patients acquired by a GE Millennium VG/Hawkeye system. Results. FBP AC and ABITR AC produced similar phantom results. For the patients, streak artifacts were observed in the FBP and ordered-subsets expectation-maximization (OSEM) maps but not in the ABITR maps, and ABITR AC produced more uniform images than FBP AC and OSEM AC. Conclusion. ABITR can improve the quality of the attenuation map, producing more uniform images for normal studies.