Thesis: In vitro labelling of cells with β+-emitting radionuclides combined with nuclear medicine imaging is a potential method for in vivo cell trafficking analysis with PET imaging. The labeling-associated exposition of cells to high levels of activity still raises some concerns since it may result in cell death and therefore a loss of image quality. In addition, the administration of potentially damaged cells rises essential questions regarding the safety of such procedure. This research work was conducted with a view of better understand the issues underlying the radiolabelling procedure in order to optimize the current clinical practice. More precisely, this thesis focused on the calculation of the absorbed doses to cells during in vitro ¹⁸F-FDG radiolabelling and the correlation to the biological observed effects. As a first step, computing tools at the multi-cellular scale were developed and optimized. Based on a generic approach, we explored and compared several hybrid methods mixing Monte Carlo simulations, analytic approaches or molecular dynamics. Then, JURKAT and adipose mesenchymal stem cells (adMSCs) were radiolabelled with ¹⁸F-FDG and tested for clonogenic survival assay, cell cycle analysis and ᵧ-H2AX phosphorylation quantification. A multi-cellular dosimetry model describing the full experiment, from the incubation of cells with ¹⁸F-FDG, washing steps, to culture of cells for functional assays was developed. Dynamic changes in cell density, as well as experimentally determined activity uptake and retention with time were thus considered. Lastly, the mean cell absorbed dose was correlated with the three biological endpoints and results were compared with X-ray irradiation. The results helped to better understand the irradiation features associated to ¹⁸F-FDG labelling and the observed biological effects, thus providing a knowledge base in favour of harmonizing the labelling methods.