Noninvasive quantification of (18)F-FLT human brain PET for the assessment of tumour proliferation in patients with high-grade glioma

PURPOSE: Compartmental modelling of 3′-deoxy-3′-[(18)F]-fluorothymidine ((18)F-FLT) PET-derived kinetics provides a method for noninvasive assessment of the proliferation rate of gliomas. Such analyses, however, require an input function generally derived by serial blood sampling and counting. In the current study, (18)F-FLT kinetic parameters obtained from image-derived input functions were compared with those from input functions derived from arterialized blood samples. METHODS: Based on the analysis of 11 patients with glioma (WHO grade II–IV) a procedure for the automated extraction of an input function from (18)F-FLT brain PET data was derived. The time–activity curve of the volume of interest with the maximum difference in (18)F-FLT uptake during the first 5 min after injection and the period from 60 to 90 min was corrected for partial-volume effects and in vivo metabolism of (18)F-FLT. For each patient a two-compartment kinetic model was applied to the tumour tissue using the image-derived input function. The resulting kinetic rate constants K(1) (transport across the blood–brain barrier) and K(i) (metabolic rate constant or net influx constant) were compared with those obtained from the same data using the input function derived from blood samples. Additionally, the metabolic rate constant was correlated with the frequency of tumour cells stained with Ki-67, a widely used immunohistochemical marker of cell proliferation. RESULTS: The rate constants from kinetic modelling were comparable when the blood sample-derived input functions were replaced by the image-derived functions (K(1,img) and K(1,sample), r = 0.95, p < 10(−5); K(i,img) and K(i,sample), r = 0.86, p < 0.001). A paired t-test showed no significant differences in the parameters derived with the two methods (K(1,img) and K(1,sample), p = 0.20; K(i,img) and K(i,sample), p = 0.92). Furthermore, a significant correlation between K(i,img) and the percentage of Ki-67-positive cells was observed (r = 0.73, p = 0.01). CONCLUSION: Kinetic modelling of (18)F-FLT brain PET data using image-derived input functions extracted from human brain PET data with the practical procedure described here provides information about the proliferative activity of brain tumours which might have clinical relevance especially for monitoring of therapy response in future clinical trials.