Time frame

Due to the statistical nature of the isotope decay process, we have to integrate the events (counts) over a certain time "frame", which has to be long enough to achieve acceptable count statistics. As the radioactive isotope is decaying and the radioligand is flushed from the organ of interest, the duration of PET "frames" need to extended to achieve sufficient counting statistics for each concentration measurement. If counting statistics is not sufficient, the resulting image will be noisy, and even biased concentration values, depending on the image reconstruction method and parameters.

In in vivo study the concentration of the radioligand in tissue is changing over time. If the time frame is long compared to the rate of change in tissue concentration, we will lose some of the kinetic information in the data, which may lead to biased results, depending on the analysis method (Raylman et al., 1993). Graphical analysis, SUV, and ratio methods generally rely on the late phase data where the concentrations change slowly, and on area-under-curve (AUC) which can be correctly calculated, independent on the frame length, as demonstrated with a simulation of long time frames in Patlak analysis. Most affected are methods that rely on the initial phase of the PET study when the blood and tissue concentrations change rapidly; this applies to perfusion measurements and compartmental models, as demonstrated with a simulation of long time frames in myocardial radiowater study.

New PET scanners store the data in list mode files (listing each event time and detectors that observed it), which enables us to re-reconstruct the images with different time frames, and thus testing the effect of the frame lengths. The optimal framing should be determined after the first pilot studies. If new image reconstructions are requested later for a large set of studies, the data recovery and reconstructions will be very time-consuming. Long axis field of view PET scanners have markedly improved sensitivity, which may enable shorter time frames.

See also:


Feng D, Li X, Siu W-C. Optimal sampling schedule design for positron emission tomography data acquisition. Control Eng Practice 1997; 5(12): 1759-1766. doi: 10.1016/S0967-0661(97)10032-6.

Graham MM. Impact of PET image sequencing on the accuracy of fluorodeoxyglucose model parameters. In: Quantification of Brain Function. Tracer Kinetics and Image Analysis in Brain PET. Elsevier, 1993, ISBN: 0-444-89859-X. pp 171-177.

Häggström I, Axelsson J, Schmidtlein CR, Karlsson M, Garpebring A, Johansson L, Sörensen J, Larsson A. A Monte Carlo study of the dependence of early frame sampling on uncertainty and bias in pharmacokinetic parameters from dynamic PET. J Nucl Med Technol. 2015; 43(1): 53-60. doi: 10.2967/jnmt.114.141754.

Li X, Feng D, Wong K. A general algorithm for optimal sampling schedule design in nuclear medicine imaging. Comput Methods Programs Biomed. 2001; 65(1): 45-59. doi: 10.1016/s0169-2607(00)00114-0.


Updated at: 2023-06-16
Created at: 2014-01-30
Written by: Vesa Oikonen