Novel multi-parametric SUV/Patlak FDG-PET whole-body imaging framework for routine application to clinical oncology

Objectives Standardized uptake value (SUV), although established in clinical oncology, depends on post-injection (p.i.) scan times and patient metabolic state[1]. In this study, we investigate a clinically adoptable and quantitative whole-body (WB) PET framework, to replace conventional single-pass SUV with multi-pass dynamic imaging of same total duration, allowing generation of both SUV and Patlak images from the same PET scan.

Methods A series of 5 WB dynamic (0-90min) FDG-PET patient studies were conducted on the Siemens Biograph mCT TOF PET scanner, each consisting of 15 dynamic WB passes [2,3]. Then, a block of 4 WB passes (10^th to 13^th), was selected, corresponding in duration (4x45sec=3min/bed) and p.i. time (60-78min for 6 beds) to routine clinical single-pass protocols and summed together either in the sinogram or image domains to obtain respective SUV data. In parallel, the selected dynamic frames were also fit via the Patlak model and a validated population-based input function method to generate parametric WB images [4]. Finally, the sinogram- and image-based SUV methods were validated against reference routine single-pass SUV conducted a few days apart.

Results Contrast-to-noise ratio (CNR) agreements within 15% were found between the proposed and reference SUV methods for all patients, which is satisfactory given SUV repeatability issues [1]. Moreover, for all patients, sinogram-based SUV achieved only slightly better precision (~5%) than image-based SUV, while accuracy was similar, thus, summing of dynamic WB images 60mins p.i. can be adequate. Finally, Patlak tumor contrast was either similar or enhanced compared to SUV in all ROIs.

Conclusions We have proposed a readily-adoptable clinical transition from routine single-pass to multi-pass WB PET imaging, and demonstrated the potential to maintain SUV generation performance, while additionally providing quantitative WB parametric images.

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