Absolute activity quantitation from projections using an analytical approach: comparison with iterative methods in brain SPECT

Estimates of SPECT activity within certain deep brain structures could be useful for clinical tasks such as early prediction of Alzheimer's disease with Tc-99m or Parkinson's disease with I-123; however, such estimates are biased by poor spatial resolution and inaccurate scatter and attenuation corrections. We compared an analytical approach to more accurate quantitation (AA) to a slower iterative approach (IA). Monte Carlo simulated projections of 12 normal and 12 pathologic Tc-99m perfusion studies, as well as 12 normal and 12 pathologic I-123 neurotransmission studies, were generated using a digital brain phantom, and corrected for scatter by a multi-spectral fitting procedure. AA included attenuation correction by a modified Metz-Pan algorithm and activity estimation by a technique that incorporated nonstationary Metz filtering to compensate for variable collimator response (VCR). IA modeled attenuation and VCR in the projector/backprojector of an OSEM algorithm. Bias and standard deviation over the 24 patients were calculated with respect to the reference values in the corpus callosum, caudate nucleus and putamen. IA and AA yielded similar quantitation results in both Tc-99m and I-123 studies in all considered brain structures in normal and pathologic patients. The bias with respect to the reference activity distributions was less than 6% for Tc-99m studies, but greater than 30% for I-123 studies, due to partial volume effect in the striata. Analytical activity estimation from projections is a promising approach to quantitation in brain SPECT.