Mathematically, the ``true'' position of the source from an image
containing a specific number of counts may be statistically given as
the peak or mean of the Gaussian distribution, within ,
the standard deviation. When visually locating homologous points in
3-D however, the exact location of a corresponding point in a SPECT
image is not made up of a single source but of an activity
distribution. Locating homologous points exactly is therefore not only
dependent on the system's fidelity but also on variable information
density and the visual perception process
such that the practice of homologous point selection in a lower
resolution image is effectively the same as randomly sampling a
position on the system PSRF of the corresponding point in the SPECT
image. Upon successive trials at selecting a point on the SPECT image
which is exactly homologous to the same point in the MR data, it may
be found that the spatial distribution of the point chosen from the
SPECT data follows the normal probability distribution. This concept
is used to simulate the homology error in the point simulations.
Registration error, as it depends on finite homology error, is
dependent on the system's spatial resolution in this context.