Quantitative SPECT in radiation dosimetry.

Accurate and precise radiation dosimetry is critical for the successful therapeutic application of systemically administered radionuclides, including, of course, radionuclides in the form of radiolabeled antibody. This requires determination, based on discrete serial measurements, of the time-dependent concentrations and/or total amounts of radioactivity in situ in order to calculate source region cumulated activities. Based on extensive studies (with clinically realistic numbers of counts and accuracies of the order of 10%) in simple geometric phantoms, in complex anthropomorphic phantoms, in animal models, and in humans, quantitative rotating scintillation camera-based single-photon emission computed tomography (SPECT) now appears to be a practical approach to such measurements. The basis of the quantitative imaging capability of a three-dimensional imaging modality such as SPECT is the elimination in the reconstructed image of counts emanating from activity surrounding the source region. Subject to considerations such as the reconstruction algorithm, attenuation and scatter corrections, and, most importantly, statistical uncertainty, the counts in a pixel in a reconstructed image are therefore directly proportional to the actual counts emanating from the corresponding voxel in situ. Among intrinsic, pre-processing, and post-processing attenuation corrections, post-processing algorithms, the most widely used approach in current commercial SPECT systems, have proven adequate in uniformly attenuating parts of the body (eg, abdomen, pelvis), subject to accurate delineation of the body contour. Although a number of sophisticated scatter correction methods have been developed, the lack of explicit scatter correction has, in practice, not been a major impediment to reasonably accurate quantitative SPECT imaging, despite scattered radiation representing up to 50% of the counts in a large source region (eg, liver). Because of its mathematical propagation in the image reconstruction process, statistical uncertainty (ie, "noise") in SPECT is far greater than would be expected if it were distributed according to Poisson statistics, as in planar imaging. The low "single slice" sensitivity of rotating scintillation camera-based SPECT is therefore the principal limitation of practical quantitative SPECT. Accordingly, absolute quantitation of count-limited clinical images has been accomplished using a judiciously selected "non-ramp" filter function. In summary, reasonable quantitative SPECT imaging is now feasible clinically, even without sophisticated scatter corrections, at least in uniformly attenuating parts of the body.(ABSTRACT TRUNCATED AT 400 WORDS)