Comparison of different references for brain perfusion SPECT quantification in clinical routine

Full text: We used 40 brain perfusion SPECT studies from the INM, UCL database to investigate the performance of several references (denominators) in the calculation of perfusion ratios with single photon emission tomography (S PET) within a routine clinical service. According to clinical diagnosis and previous SPECT findings 4 groups were identified composed of: 10 controls (C, 23 to 84 y old); 10 myalgic-encephalomyelitis / chronic fatigue syndrome (ME/CFS, 22 to 61 y old); 10 major depression (MD, 24 to 68 y old); and 10 temporal lobe epilepsy (TLE, 19 to 39 y old). Routine protocols for processing were used and the analysis was blind to group classification. Brain perfusion ratios were calculated using 7 different references: hemi cerebellum with higher counts (Cer), total counts in a 4 pixel slice through the basal ganglia slice (BG), average counts per pixel in the visual cortex (VC), average counts per pixel in the white matter (WM), total acquired counts (TAC), total reconstructed counts (TRC) and maximum counts per pixel in the entire study (MAXX). Unpaired test to compare different diagnostic groups, coefficient of variation (CV) to assess the reliability to each references followed by ANOVA were the statistical test used. The lowest mean CV's were found with VC (4.8%) and TRC (5.1%), with all the others significantly higher (p<0.0001). The range of CV's for Cer was the lowest (3.7% to 5.9%). Consistent differentiation between diagnostic groups and controls was only obtained with Cer. In conclusion, it appears that for clinical routine services Cer is the most reliable reference, exception made for all diseases affecting the cerebellum. In these cases TRC or VC should be preferred. (authors)