Evaluation of a Fit Model for Time Activity Curves of Dynamic FET PET Acquisitions

Ziel/Aim:

Several studies have demonstrated that the shape of the time-activity-curve (TAC) of O-(2-[F-18]fluoroethyl)-L-tyrosine (FET) uptake in cerebral gliomas during the first hour p.i. depends on histological grade. Extraction of the parameters best describing the FET TACs in brain tumours is a matter of controversy. This study evaluates an explicit empirical model able to correctly describe the different TACs of FET with sufficient closeness.

Methodik/Methods:

The function class log(u)= log(A)+0.5 log(t)-κ √t successfully reproduces all typically observed TACs of FET, where u = u(t) is the FET TAC, A is the amplitude, and? describes the curve shape. The model was fitted to the average TACs, i.e. averaged over volume of interest and voxel-wise TACs. We evaluated the model in 36 PET data sets from low-grade (LG) and high-grade (HG) gliomas and with different dynamic framing schemes: standard scheme consisting of 16 time frames (5 × 1 min; 5 × 3 min; 6 × 5 min) and shorter schemes consisting of 4 time frames (4 × 5 min) starting at 20 min p.i. Results were compared to fitting with a linear model and diagnostic performance for glioma classification was evaluated (by ROC and LOOCV).

Ergebnisse/Results:

Relative fit parameter uncertainties for averaged TACs where smaller than 0.1% in all cases. The adjusted R² values for fits with the empirical model where always larger than 0.45 and larger than 0.95 in 80% of the cases. Relative fit parameter uncertainties for voxel-wise TACs where smaller than 0.4 in 90% of all cases. Adjusted R² values were larger than 0.7 in 50% of all cases. We observed a strong correlation between best-fit parameters from voxel-wise fits and averaged TAC fits. Diagnostic performance for glioma classification was best for the empirical model with 16 frames.

Schlussfolgerungen/Conclusions:

The empirical model precisely reproduces the uptake kinetics of both HG and LG gliomas in case of TAC averaged over the tumor volume and a contralateral region of normal brain tissue and allows for reliable glioma classification.