A method for model-free recovery correction in PET

Objectives : As is well known, limited spatial resolution leads to partial volume effects and consequently to limited signal recovery. Determination of the true activity concentration of a target structure is thus compromised even at target sizes much larger than the reconstructed spatial resolution. This leads to serioussize-dependent underestimates of true signal intensity in hot spotimaging. For quantitative PET in general and in the context of therapy assessment in particular it is therefore mandatory to perform an adequate recovery correction (RC). The goal of our work was to
developand to validate a model-free RC for hot spot imaging.
Methods : The algorithm proceeds in two steps. Step 1: automatic estimation of the true object volume V with an automatic thresholdbased method and the corresponding total activity A within V. Step 2:automatic determination of the background corrected activity fractionB, which is measured outside the object due to the partial volume effect. The recovery coefficient is then given by R = A/(A+B). For validation, we used a cylindrical phantom with six sphere inserts(volume: 2.5 - 27
ml) and performed measurements at three different target/background levels. RC was computed using the analytical convolution of the object function with the best fitting Gaussianpoint spread function and then compared to the results of the model-free approach.
Results : For the investigated target/background contrasts the model-free approach yields RC values which agree very well with the known (computable) RC values for the spheres (mean deviation +/-5%). The model-free nature of the approach allows, however, direct application to real patient data where the exact target shape is not known and model-based RC would fail.
Conclusions : The described method provides an easy and essentially automated way of doing quantitative RC in oncological hot spotimaging. It works well for sufficiently high contrast (>5) and a homogeneous background. The limits of the method have to be investigated further.