Analysis of single-subject data sets with a low number of PET scans

We present a procedure for the statistical analysis of single-subject PET data sets with a low number of scans (six total, two scans for each of three conditions) and describe the results of applying this method of analysis on the PET data of 10 normal subjects, performing a motor task, with six scans per subject. Development of this procedure, which combines well-established methods of statistical analysis in functional neuroimaging (a pooled estimate of variance and spatial-extent-based analysis), was motivated by the need for a statistical, individual analysis of PET data sets with a low number of scans from stroke patients in which the location and extent of an infarct varied from subject-to-subject. The results of this spatial-extent-based, single-subject analysis of PET data (including coregistration of significant PET activation to every individual's MRI image and transformation into Talairach space) showed activation in all major cortical motor areas, while no activation was detected in small structures such as the putamen and thalamus. Individual variability in the pattern of statistically significant regions of activation across 10 normal subjects, performing the same motor task, was observed. The results obtained in this study from individual analysis of a low number of PET scans of normal subjects are consistent with previous results of group PET analysis of normal subjects performing similar motor tasks, and these current results indicate that this procedure can help in the examination of PET data sets demanding a single-subject analysis (e.g., involving stroke).