Assessment of SPM in perfusion brain SPECT studies. A numerical simulation study using bootstrap resampling methods.

Area of Science:

• Neuroimaging
• Medical Physics
• Biostatistics

Background:

• Statistical Parametric Mapping (SPM) is widely used for analyzing functional brain imaging data (PET, fMRI, SPECT).
• Methodological assessments of SPM's performance in SPECT are limited.
• Understanding factors influencing SPM's sensitivity in SPECT is crucial for accurate clinical interpretation.

Purpose of the Study:

• To evaluate the performance of SPM in detecting regional cerebral blood flow (rCBF) changes in simulated brain SPECT studies.
• To investigate the impact of image reconstruction corrections (attenuation, scatter, collimator response) on SPM performance.
• To determine the relationship between group size, rCBF changes, and the effectiveness of SPM in SPECT.

Main Methods:

• Simulated brain SPECT studies using Monte Carlo techniques with a fan-beam collimator.
• Image reconstruction using OSEM algorithm with and without corrections for degradations.
• Statistical analysis with SPM2 (two-sample t-test) and bootstrap resampling for sample size determination.

Main Results:

• Image correction significantly reduces the required sample size for SPM analysis in SPECT.
• The reduction in sample size due to corrections is more pronounced for small brain regions and low rCBF changes.
• Differences in sample size requirements were observed between hypo- and hyperperfusion conditions, particularly without image corrections.

Conclusions:

• Image correction during SPECT reconstruction is essential for optimizing SPM analysis and reducing sample size requirements.
• SPM performance in SPECT is sensitive to region size, activation magnitude, and the inclusion of image corrections.
• These findings have implications for designing future SPECT studies and interpreting rCBF changes using SPM.