Unfortunately, segmenting and co-registering
is not enough...
Knowing the composition of a PET voxel (from the co-registered
segmented MRI), and its value (from the PET image) is not enough to
know the real tracer concentration in that voxel. PVE activity “spills in” from adjacent structures
into that voxel, so that its value will be higher if “hot” structures
are close, or lower when surrounded by “cold “ structures. In turn,
also activity from that voxel will “spill over” adjacent structures,
altering their apparent activity. Each voxel final value, as measured in PET,
will depend from the composition of all the voxel in the area, and from
the concentrations of the tissues that compose these voxels. In the figure below, we simulate in 1D two
structures with two different concentrations (blue line). A “hot”
structure (voxels 9 to 19) and a “cold” structure (voxel 20 to 29),
surrounded by metabolically inert tissues. The resulting PET curve will
be due to these combined effects modulated by the resolution of the
scanner.