Abstract:
The effect of neuroanatomical structure size, shape, and position
versus spatial tomographic resolution on quantitation in positron
computed tomography was investigated. For neuroanatomical structures,
voxel sizes in excess of 3 ml exceeded the volume of most structures
examined. When the voxel size exceeded structure volume, calculated
recovery coefficient (fraction of the true isotope concentration
measured in the image) fell to less than or equal to 42%. Partial volume
effects in the plane of section analyzed by computer simulation produced
errors that were largest for small, thin, irregularly shaped structures
whose averaged pixel values were most different from neighboring
structures. Smallest errors occurred in large, circular structures
surrounded by regions of similar pixel values. Computer simulation of
regional cerebral asymmetries of pixel values demonstrated that the
measurement of these asymmetries was often predominated (enhanced or
obliterated) by partial volume effects related to structure size and
shape. Large, circular, and widely separated regional asymmetries were
more easily detected at a given spatial resolution than small, thin,
adjacent regions. Recommendations for error reduction and possible
correction factors are provided and discussed