Marginalisation over voxels wholly in the areas of no interest proceeds exactly as in the case of a flat intensity prior, since the multi-variate prior here does not include these parameters. These parameters therefore split into null, degenerate, and uninteresting. The degenerate and uninteresting parameters integrate in the same way as for the flat prior case and the null parameters integrate as part of in the previous section.
Specifically, this gives
(10) |