Accuracy and variability of PET activity concentration measurements for small volumes and short time frames

Objectives Commercial PET-CT systems are optimised for whole-body FDG-PET and there is little data on dynamic imaging performance. The aim of this study was to assess the accuracy and variability of activity concentration measurements for small objects, and for the short frame times and range of count rates encountered in dynamic acquisitions.

Methods Acquisitions of the NEMA image quality phantom were performed on a GE Discovery VCT PET-CT scanner. A protocol with frame durations of 10x1s, 10x2s, 10x5s, 10x10s, 5x20s, 5x30s, 5x60s and 5x300s was repeated four times as the total ¹⁸F activity decayed from 298 to 30MBq. Data were reconstructed using iterative and filtered back projection (FBP) algorithms; corrections for attenuation, scatter and dead time were performed. For the 6 spheres in the phantom (diameters 10-37mm), bias and variability were assessed for max (A_max) and mean (A_mean) activity concentration measurements.

Results For short frames (<10s) A_max measurements showed a positive bias with both FBP and iterative reconstructions. Bias increased as frame duration decreased, up to 67% for FBP and 148% for iterative when using 1s frames. Variance for repeated measurements also increased for short frame times and decreasing sphere volume. Greatest variance was for 1 s frames with the 10mm sphere giving CoV of 35.2% for A_max and 53.2% for A_mean. This compares to CoV of 2.9% for A_max and 2.7% for A_mean for the 300s frames. Dead-time effects were observed at higher activities causing bias of up to 21% at an activity of 276MBq. As expected partial volume effects increase as sphere size decreases to 10mm leading to biases of up to 68% and 72% for A_max and A_mean respectively.

Conclusions We have shown that substantial bias and variability exists in PET uptake measurements for small objects acquired over short time frames and at high count rates. These may lead to errors in calculation of kinetic parameters in dynamic PET studies.