Use of image derived input functions for [11C]PIB studies: Assessment of accuracy and test-retest variability

Introduction: Derivation of input functions directly from dynamic PET images obviates the need for arterial sampling and will enhance clinical applicability of quantitative brain studies. The aim of the present study was to assess accuracy and test-retest variability of [11C]PIB studies when using image derived input functions (IDIF) obtained using reconstruction-based partial volume correction (PVC) [1]. Methods: PET and arterial blood data from five repeat dynamic [11C]PIB scans, acquired using an ECAT EXACT HR+ scanner and an on-line blood sampler, were used in the present study. Test and retest scans were performed on the same day. Scans were reconstructed using standard (no PVC) ordered subset expectation maximization (OSEM, 2 iteration (i), 16 subsets (s)) and a PVC-OSEM reconstruction algorithm, which corrects for the spatial resolution of the scanner. PVC-OSEM scans were reconstructed using 4i, 16s and a Gaussian recovery kernel of 5.5 mm FWHM. For the region below the base of the skull, regions of interest (ROIs) were drawn semiautomatically over the four hottest pixels of the carotid arteries using a pseudo blood volume image, generated by summation of the early time frames (<60s). These ROIs were projected onto all dynamic frames, thereby generating corresponding IDIFs. Both IDIFs with and without calibration to manual samples taken during the scans were used for further analysis. Parametric images of volume of distribution (Vd), based on Logan analysis, were generated using both the on-line blood sampler input function (BSIF) and the various IDIFs. For each subject, 15 tissue ROIs were drawn and projected on all Vd images. The validity of each IDIF was assessed by correlating IDIF-based Vd values with the corresponding BSIF-based values. In addition, test-retest variability was investigated. Results: Image quality of BSIF- and IDIF-based Logan Vd images was similar. A summary of the quantitative comparison of IDIF- and BSIF-based Vd values is given in figure 1 and 2. Calibration of the IDIFs did not result in better correlation coefficients, but significantly reduced differences in Vd. For PVC-OSEM, calibration resulted in a slope of 1.00 (figure 1). No significant differences in test-retest variability were found between BSIF- and IDIF-based Vd values. Discussion and Conclusions: For [C11]PIB studies, an image derived input function is an accurate alternative to arterial blood sampling, provided that scans are reconstructed using a reconstruction-based partial volume correction method. Manual samples, however, are still needed for calibrating the IDIF, and for determining plasma/blood ratio and metabolites. These samples can probably taken from venous blood.