In vivo lung counting, one of the preferred methods for monitoring people exposed to the risk of actinide inhalation, is nevertheless limited by the use of physical calibration phantoms which, for technical reasons, can only provide a rough representation of human tissue.
A new approach to in vivo measurements has been developed to take advantage of advances in medical imaging and computing; this consists of numerical phantoms based on tomographic images (CT) or magnetic resonance images (IRM) combined with Monte Carlo computing techniques. Under laboratory implementation of this innovative method using specific software called OEDIPE, the main thrust of this thesis was to provide answers to the following question: what do numerical phantoms and new techniques like OEDIPE contribute to the improvement in calibration of low-energy in vivo counting systems?
After a few developments of the OEDIPE interface, the numerical method was validated for systems composed of four germanium detectors, the most widespread configuration in radiobioassay laboratories (a good match was found, with less than 10% variation). This study represents the first step towards a person-specific numerical calibration of counting systems, which will improve assessment of the activity retained.
A second stage focusing on an exhaustive evaluation of uncertainties encountered in in vivo lung counting was possible thanks to the approach offered by the previously-validated OEDIPE software. It was shown that the uncertainties suggested by experiments in a previous study were underestimated, notably morphological differences between the physical phantom and the measured person. Some improvements in the measurement procedure were then proposed, particularly new biometric equations specific to French measurement configurations that allow a more sensible choice of the calibration phantom, directly assessing the thickness of the torso plate to be added to the Livermore phantom based on the weight and height of the measured person.
Lastly, the study underlined the interest of numerical phantoms and Monte Carlo simulation through actual contamination cases of lungs or wounds, which are impossible to study using traditional methods. In the case of contaminated wounds, this method was used to adjust the level of the retained activity in an actual injury on a hand and should improve the determination of source geometry, thereby refining the dose calculation.
Personalised calibration of counting systems (for morphological purposes or distribution of radionuclides in the body) appears possible thanks to this innovative method and represents an important step towards implementation of personalised dosimetry.