Abstract
1591
Objectives [18F]Fluorothymidine (FLT) undergoes predominantly hepatic metabolism to form [18F]FLT-glucuronide and therefore shows high physiological uptake in the liver, with the result that tumors (metastases) in the liver are difficult to identify and quantify. The purpose of this study is to evaluate if lesion detectability can be improved with the use of parametric images based on differences in tracer kinetics between tumor tissue and normal liver. In addition, we also evaluate if these parametric images allow robust quantitation by assessing test-retest reliability of VOI measurements prior to treatment.
Methods In this multicenter, prospective observational study performed within the EU-IMI QuIC-ConCePT project, colorectal cancer (CRC) patients with at least 1 liver metastasis > 2 cm scheduled for chemotherapy underwent a 60 min dynamic liver FLT-PET scan twice within 1 week prior to the start of treatment and a follow-up scan after 2 cycles. Venous blood samples were collected at 6 time points during the FLT scan to measure radioactivity concentration and metabolite fraction. An image-derived input function (IDIF) was obtained from the descending aorta and corrected for metabolites to obtain a calibrated parent [18F]-FLT plasma input function. Different (parametric) images were reconstructed for visual and quantitative analysis: SUV from 50-60 min post injection (p.i.) normalized to body weight (SUVFLT), and Patlak derived Ki maps from 15-60 p.i based on IDIF with or without metabolite correction (Ki+ and Ki resp.). To assess lesion detectability, the number and location of the liver metastases > 2 cm were defined on the whole body FDG-PET/CT performed as part of the routine workup within the same week (reference standard). For each FLT image set, the number of hot spots corresponding to an FDG+ metastasis was recorded. Next, a volume of interest (VOI) was placed around each lesion on both FDG and FLT data using a 3D lesion growing tool in PMOD (version 3.6) encompassing the hot area. The maximum and mean voxel value as well as VOI volume were determined. A 2D circular VOI of 2 cm was placed in normal liver to calculate the mean normal liver value and tumor to background ratio’s (T/B). Test-retest reliability was evaluated using the coefficient of variation (CoV) and the intraclass correlation coefficient (ICC).
Results For this interim analysis, a total of 14 metastatic lesions (20% of accrual target) from 4 patients were analyzed for lesion detectability and test-retest analysis. None of the 14 FDG+ lesions were detectable on the SUVFLT-images, while only 1 lesion could not be visualized on Ki and Ki+ (example fig. 1). The average FDG tumor volume was significantly larger than the FLT tumor volume: 56 ± 90 mm³ on FDG vs 13.2 ±13.4 mm³ on Ki+ (p = 0.0464) and 13.8 ±13.5 on Ki (p = 0.0464). The mean volume ratio FDG/Ki+ was 8.38 ±14.59 and 8.23 ±14.94 for FDG/Ki. Lesion contrast was also better for FDG compared to Ki-FLT images reflected in higher T/B ratio’s: 4.31±1.31 on FDG vs 1.40±0.34 on Ki+ (p=0.001) and 1.62±0.35 on Ki (p=0.001). Repeatability was acceptable for the max and mean pixel value (CoV of 11±6% and 14±8% on Ki+ ; 13±7% and 15±7% on Ki). Repeatability of the FLT volume was weak (CoV of 40±20% on Ki+ and 30±19% on Ki). The ICC for Ki+ and Ki showed a moderate absolute agreement (ICC Ki+=0.506 and ICC Ki=0.594), with a standard error of measurement of 0.008 and 0.004, respectively.
Conclusions Preliminary data suggest that SUVFLT images 60 min p.i. are not suitable for detection liver metastasis. The use of parametric Patlak images improves lesion to liver contrast to levels above the detection threshold. Repeatability of VOI quantification seems acceptable for subsequent use in treatment response studies. $$graphic_EC722C03-4C78-4528-9F48-F55776FFA961$$