Creation of the Common Normal Database for Brain Perfusion SPECT: A Multicenter Phantom Study

Objectives Statistical image analysis of brain perfusion SPECT have provided improvement of diagnostic accuracy. However, normal database (NDB) for statistical image analysis is required to be created for each institution and SPECT scanner without scatter correction (SC) and attenuation correction (AC). The objectives of this study was evaluation of acquisition and processing protocol for creating common NDB in brain perfusion SPECT by multicenter phantom study.

Methods Our proposed acquisition and processing protocol were usual acquisition protocol at each center and additional SC (triple energy window method) and AC (CT based attenuation correction). A multicenter phantom study was performed in 5 centers on 6 imaging systems with our protocol and brain phantom with the following conditions: (1) The gray/white matter I-123 activity ratio in the brain phantom was 4. (2) The brain phantom was covered with artificial adult male skull (1300HU), female skull (850HU), or acrylic cover. The cut-off frequency of Butterworth filter was performed so that the spatial resolution was unified 17.9 mm FWHM in SPECT images. The ratios of gray and white matter counts were measured from SPECT images and compared against the actual activity ratio. To evaluate the variability of NDB, the mean, standard deviation (SD) and relative SD were calculated by using neurologic statistical image analysis software (NEUROSTAT).

Results Error of count ratios of gray and white matter in without SC, AC and with SC, AC were 23.7 ± 6.32% and 4.60 ± 2.81%, respectively. After unification of spatial resolution in SPECT images, the variability of NDB were decreased as compared with before, which were the equivalent of without SC and AC.

Conclusions Our proposed acquisition and processing protocol showed the potential of creation of the common NDB from the SPECT images with SC, AC and unification of spatial resolution.