Optimising use of adaptive filtering and anatomical prior for SPECT reconstruction of clinical data

This paper reports results demonstrated from phantom and clinical studies, showing that the iterative minimum cross-entropy (MXE) technique, using adaptive filtering and anatomical information, improves the quality (noise and contrast) of SPECT reconstruction compared with maximum likelihood-expectation maximisation (ML-EM) technique. Suitable parameters for using adaptive filtering in the MXE reconstruction of clinical data were first determined. This demonstrated that combined anatomical and physiological information is acceptable for adaptive filtering in MXE reconstruction of clinical SPECT data. Using data obtained from brain images of 15 healthy volunteers, the two techniques are compared by statistical parametric mapping (SPM) analysis. SPM produces significant differences (p<0.001) in ventricle region, which correlates well with findings in region of interest (ROI) analysis.