Scatter and attenuation correction for brain SPECT using attenuation distributions inferred from a head atlas
R Z Stodilka1, B J Kemp, F S Prato
1Department of Nuclear Medicine and Magnetic Resonance, St. Joseph's Health Center, London, Ontario, Canada.
Unlabelled:
Sequential transmission scanning (TS)/SPECT is impractical for neurologically impaired patients who are unable to keep their heads motionless for the extended duration of the combined scans. To provide an alternative to TS, we have developed a method of inferring-attenuation distributions (IADs), from SPECT data, using a head atlas and a registration program. The validity of replacing TS with IAD was tested in 10 patients with mild dementia.
Methods:
TS was conducted with each patient using a collimated 99mTc line source and fanbeam collimator; this was followed by hexamethyl propyleneamine oxime-SPECT. IAD was derived by deformably registering the brain component of a digital head atlas to a preliminary SPECT reconstruction and then applying the resulting spatial transformation to the full head atlas. SPECT data were reconstructed with scatter and attenuation correction. Relative regional cerebral blood flow was quantified in 12 threshold-guided anatomic regions of interest, with cerebellar normalization. SPECT reconstructions using IAD were compared with those using TS (which is the "gold standard") in terms of these regions of interest.
Results:
When we compared all regions of interest across the population, the correlation between IAD-guided and TS-guided SPECT scans was 0.92 (P < 0.0001), whereas the mean absolute difference between the scans was 7.5%. On average, IAD resulted in slight underestimation of relative regional cerebral blood flow; however, this underestimation was statistically significant for only the left frontal and left central sulcus regions (P = 0.001 and 0.002, respectively). Error analysis indicated that approximately 10.0% of the total error was caused by IAD scatter correction, 36.6% was caused by IAD attenuation correction, 27.0% was caused by discrepancies in region-of-interest demarcation from quantitative errors in IAD-guided reconstructions, and 26.5% was caused by patient motion throughout the imaging procedure.
Conclusion:
SPECT reconstructions guided by IAD are sufficiently accurate to identify regional cerebral blood flow deficits of 10%, which are typical in moderate and advanced dementia.
