Once the corrections for the various sources of bias described in this chapter have been applied to PET data, images can be reconstructed which are free of artefacts and which reflect the regional concentration of radiotracer in the body. In most clinical applications of PET this is sufficient as the images are interpreted visually without reference to the absolute voxel values. Indeed, this degree of "quantification" is sufficient in some types of kinetic research study also. For example, methods which take a reference tissue as the input function for the kinetic model do not require the PET images to be calibrated in absolute units of tracer concentration. However, for most kinetic study protocols PET image values must be related to the tracer concentration in blood samples withdrawn during the study. These blood samples are normally counted in a well counter. Thus, it is essential in these studies to have an accurate calibration between the PET scanner and the well counter. This is usually achieved by scanning a phantom with uniform radioactivity concentration and then counting an aliquot taken from the phantom in the well counter. The phantom images are reconstructed using the same corrections as are applied in research studies and the voxel values directly compared with the counted aliquot to determine a calibration factor.
Note that the above procedure still does not necessarily provide a reading in absolute units of radioactivity concentration for the PET image voxels unless the radioactivity concentration in the aliquot is accurately known. Once again, this is not usually necessary for quantitative tracer kinetic studies. It is usually sufficient to have an accurate calibration factor that relates PET image values to well counter measurements. If absolute units of radioactivity concentration are required, a procedure is available which results in a measurement of the detection efficiency of the tomograph in air . When PET images are corrected for all the effects described above, the image values (in units of counts/sec/voxel) can be divided by the tomograph efficiency (in units of counts/sec/kBq) and then divided by the voxel volume to yield images calibrated in units of kBq/ml.
Was this article helpful?