Variability and accuracy of two FDG quantification techniques: A combined phantom and pre-clinical study

Objectives Non-invasive tracking of injected cells has a variety of applications, such as the development of stem cell therapies. Radiolabeled cells imaged with PET/CT rely on the accuracy of quantifying small volumes (vol) and concentrations (conc). Two methods of PET/CT processing were investigated for variability and accuracy. These methods were then applied to a large-animal model of FDG labeled cell injection data.

Methods Two phantoms were constructed, consisting of 7 syringes of FDG, either varying vols or concs, and 1 standard. Five minute 3D PET images were acquired per phantom over 1 bed position with a GE Discovery VCT. The data was processed using either GE Xeleris or GE Advanced Windows Workstation. The same processing was applied to data from a large-animal pre-clinical study in which 6 separate 0.1ml injections of FDG labeled cells were made into the myocardium of a swine and scanned with a standard.

Results Across different vols, both Xeleris and AW were similar (Table). Data became less accurate for both techniques as vols fell below ROI diameters of 2 cm. Across different concs, Xeleris outperformed AW and maintained high accuracy until the lowest conc, while the accuracy of AW steadily decreased. For the large-animal data, both Xeleris and AW were comparable but AW estimated slightly higher vols than Xeleris.

Conclusions Both techniques performed well across a variety of small vols and concs. Xeleris was more accurate at smaller concs, while both techniques suffered from volume averaging when below the FWHM of the PET camera. The latter can be addressed using a recovery coefficient algorithm. The vol-based technique used by AW suffered from incomplete mixing in the sample preparations. The large-animal data confirmed that both techniques were reliable and proper utilization of each is critical to their use in pre-clinical and clinical applications requiring accurate quantification of radiolabeled data

• Download figure
• Open in new tab
• Download powerpoint