Clinical Whole-body PET Patlak imaging 60-90min post-injection employing a population-based input function

Objectives In this study, we propose a streamlined whole-body (WB) dynamic PET acquisition protocol relying on dynamic data of only 60-90min post-injection (p.i.), thus matching the respective scan period of SUV static PET studies, and a population-based input function (pIF).

Methods We utilized the Patlak kinetic model, thus, only requiring the time integral of the IF and the later section of the time-activity curves (TACs), and combined it with a population-based IF (pIF) method, therefore alleviating the need to scan from injection time [1]. For the pIF, a series of 0-90min WB dynamic clinical PET scans is currently in progress on 2 Siemens Biograph mCT TOF PET scanners. So far, IFs have been derived from 5 patients studies after drawing carefully selected left-ventricle regions-of-interest (ROIs) of dynamic PET images (image-derived IFs, IDIFs), to minimize partial volume effect. Then, for each of the 5 patients, the set of 4 IDIFs corresponding to the other patients were considered (leave-one-out cross validation) [2]. Subsequently, the IDIFs of each set were scaled and averaged to produce a pIF for each patient. Finally, each pIF was scaled to match the later samples (60min p.i.) of the respective patient IDIF. For validation purposes, the WB Patalk images utilizing either the pIFs or IDIFs, as calculated above, were generated utilizing only the last 6 WB passes (60-90min p.i.) for each patient and then compared with each other.

Results The differences in tumor-to-background (TBR) contrast between pIF and IDIF generated Patlak WB images were less than 5% in all ROIs of all 5 patient cases evaluated. This is an ongoing study and the accuracy of the method is expected to increase with higher number of patient studies.

Conclusions Clinically feasible WB Patlak imaging based only on 60-90min p.i. dynamic data can produce reliable WB parametric images when a cross-validated pIF method is employed.

Research Support This work was supported by the Swiss National Science Foundation under Grant SNSF 31003A-149957 and by Siemens Medical Solutions.