The Logan plot is a powerful algorithm used to generate binding-potential images from dynamic positron emission tomography (PET) images in neuroreceptor studies. However, it requires arterial blood sampling and metabolite correction to provide an input function, and clinically it is preferable that this need for arterial blood sampling be obviated. Estimation of the input function with metabolite correction using an intersectional searching algorithm (ISA) has been proposed. The ISA seeks the input function from the intersection between the planes spanned by measured radioactivity curves in tissue and their cumulative integrals in data space. However, the ISA is sensitive to noise included in measured curves, and it often fails to estimate the input function. In this paper, we propose a robust estimation of the cumulative integral of the plasma time-activity curve (pTAC) using ISA (robust EPISA) to overcome noise issues. The EPISA reduces noise in the measured PET data using averaging and clustering that gathers radioactivity curves with similar kinetic parameters. We confirmed that a little noise made the estimation of the input function extremely difficult in the simulation. The robust EPISA was validated by application to eight real dynamic [11C]TMSX PET data sets used to visualize adenosine A2A receptors and four real dynamic [11C]PIB PET data sets used to visualize amyloid-beta plaque. Peripherally, the latter showed faster metabolism than the former. The clustering operation improved the signal-to-noise ratio for the PET data sufficiently to estimate the input function, and the calculated neuroreceptor images had a quality equivalent to that using measured pTACs after metabolite correction. Our proposed method noninvasively yields an alternative input function for Logan plots, allowing the Logan plot to be more useful in neuroreceptor studies.
ASJC Scopus subject areas