Although non-uniform attenuation correction (NUAC) can supply more accurate absolute quantification, it is not entirely clear whether NUAC provides clear-cut benefits in the routine clinical practice of brain SPET imaging. The aim of this study was to compare the effect of NUAC versus uniform attenuation correction (UAC) on volume of interest (VOI)-based semi-quantification of a large age- and gender-stratified brain perfusion normal database. Eighty-nine healthy volunteers (46 females and 43 males, aged 20-81 years) underwent standardised high-resolution single-photon emission tomography (SPET) with 925 MBq 99mTc-ethyl cysteinate dimer (ECD) on a Toshiba GCA-9300A camera with 153Gd or 99mTc transmission CT scanning. Emission images were reconstructed by filtered back-projection and scatter corrected using the triple-energy window correction method. Both non-uniform Chang attenuation correction (one iteration) and uniform Sorenson correction (attenuation coefficient 0.09 cm(-1)) were applied. Images were automatically reoriented to a stereotactic template on which 35 predefined VOIs were defined for semi-quantification (normalisation on total VOI counts). Small but significant differences between relative VOI uptake values for NUAC versus UAC in the infratentorial region were found. VOI standard deviations were significantly smaller for UAC, 4.5% (range 2.6-7.5), than for NUAC, 5.0% (2.3-9.0) (P<0.01). Higher filter roll-off values of the transmission reconstruction filter increased noise in the emission images and altered estimated cortical attenuation coefficients as well as uptake values. In conclusion, semi-quantification based upon reconstruction with UAC results in very similar 99mTc-ECD uptake values in healthy volunteers to those obtained with NUAC, although values for the infratentorial region are slightly lower. NUAC produces a slight increase in inter-subject variability. Further study is necessary in various patient populations to establish the full clinical impact of NUAC in brain perfusion SPET.