A Monte Carlo analysis technique applied to cerebral perfusion SPECT scans

Abstract

Voxel-based techniques are used to provide objective analyses of SPECT scans. The aim of this study was to develop a voxel-based analysis technique that uses a Monte Carlo method of statistical inference and assess this analysis technique for application to cerebral perfusion SPECT scans. Assessment of the validity of this non-parametric, Monte Carlo method of statistical inference has been performed for a range of study designs, image characteristics and analysis parameters using phantom SPECT and Gaussian images. The Monte Carlo method of statistical inference and the voxel-based analysis technique were clinically evaluated for the analysis of individual cerebral perfusion SPECT scans using control subject data. In addition, a comparison has been performed with an existing analysis package that uses a theoretical parametric method of statistical inference (statistical parametric mapping). The Monte Carlo method was found to provide accurate statistical inference for phantom SPECT and Gaussian images independent of degrees of freedom, acquired counts, image smoothness and voxel significance level threshold. The clinical evaluation of the analysis of individual cerebral perfusion SPECT scans demonstrated satisfactory statistical inference and characterization of perfusion deficits. An analysis method incorporating a Monte Carlo method of statistical inference has been successfully applied for the analysis of cerebral perfusion SPECT scans.