Fluorine-18 3'-deoxy-3'-fluorothymidine (18FLT) is a tissue proliferation marker which has been suggested as a new tumour-specific imaging tracer in positron emission tomography (PET). The objectives of this study were to investigate the pharmacokinetics of 18FLT in patients with colorectal cancer, defining methodologies for the quantitative analysis of the in vivo 18FLT uptake and subsequently assessing the accuracy of semi-quantitative measures. Dynamic acquisitions over a single field of view of interest identified by computed tomography were carried out for up to 60 min following injection of 18FLT (360 +/- 25 MBq). Dynamic arterial blood sampling was carried out in order to provide a blood input function. Simultaneous venous samples were also taken in order to investigate their potential utilisation in deriving a hybrid input function. Arterial and venous blood samples at 5, 15, 30, 60 and 90 min p.i. were used for metabolite analysis. Eleven patients with primary and/or metastatic colorectal cancer were studied on a lesion by lesion basis (n = 21). All acquired images were reconstructed using ordered subsets expectation maximisation and segmented attenuation correction. Time-activity curves were derived by image region of interest (ROI) analysis and image-based input functions were obtained using abdominal or thoracic aorta ROIs. Standardised uptake values (SUVs) were calculated to provide semi-quantitative indices of uptake, while non-linear regression (NLR) methodology in association with a three-compartment model and Patlak analysis were carried out to derive the net influx constant Ki. The metabolite analysis revealed two radioactive metabolites, with the parent compound representing approximately 80% of the total radioactivity in the 30-min plasma sample. In the case of NLR, better fits were obtained with a 3k model (i.e. k4 = 0) for both lesion and bone marrow time-activity curves. For the same lesions, a high correlation was observed between the Ki derived from either Patlak analysis or NLR(3k) and the corresponding SUVs. Our results also suggest that the quantitative behaviour of 18FLT in vivo (up to 60 min p.i.) may be characterised using a 3k model or Patlak analysis in combination with image-derived input functions. The good correlation found between the SUVs (at 60 min) and Ki values supports the use of semi-quantitative indices to assess the proliferation rate of colorectal cancer lesions in vivo with 18FLT.