Abstract

The advantage of whole-body imaging for distribution studies is that it accounts for all activities. The problem, however, is that the classic approach to determining distribution from planar images does not accommodate overlapping structures. That approach assumes implicitly that the sampling region is a prismoid whose cross-section, parallel to the detector plane, is defined by a region of interest and whose sides are orthogonal to the detector plane.In the proposed organ-model approach, the region of interest is assumed explicitly to be the projected shadow of an organ or structure, whose general shape is known from anatomic generality, and whose size or specific shape variation is defined by the shadow or region of interest. If "j" is a pixel in the organ shadow "i", a fraction "Vij" of the volume of organ "i" is assumed to project orthogonally in "j". More than one organ shadow can overlap, in which case the volumes projecting in "j" are the sum of "Vij" over "i". The activity "Aj" (count rate density) in any location "j" is defined by the linear combination of volumes "Vij" and concentrations "Ci". For all the pixels in the image, this defines an overdetermined set of linear equations that can be solved by matrix inversion for "Ci", the organ concentrations.The organ-model method was tested on simulated and phantom data. It proved, on serial and repeat processing, to be robust (not subject to large errors due to small variations) if the images had sufficient contrast. This method was found to be superior to the classic approach in evaluating the same data, because in the classic approach, the border regions are too heavily weighted, and therefore, the size of the sampling region is critical. Furthermore, the expression of the results in concentrations is more relevant to dosimetry, the derivation of which is based on cumulative concentrations.Modeling organ shadows is a viable improvement on the use of regions of interest to quantify tracer distribution in planar imaging.

View details for Web of Science ID A1994MV41800025

View details for PubMedID 8306280