When a normalized subject image is compared against the Brain Norm, a z-score value is calculated in each image pixel. This operation requires the standard deviation of the normalized values across the controls data sets. If the number of control samples is low, the statistical power may not be sufficient for calculating reliable standard deviations in each individual pixel. In this case Variance Pooling on the 5. Variance Pooling tab should be enabled.
With this setting the user needs to define a spatial domain, within which the pixel standard deviations are averaged to derive a pooled standard deviation which will be used for all pixels. A VOI, a Mask, or the whole Normalized sample can be employed for variance pooling.
Note: After changing the Variance Pooling method the database has to be recalculated with the Norm Calculation button.