Résumé : The literature concerning quantification of costovertebral joint (CVJ) motion under in vivo conditions is scarce. Most papers concerning this topic are related to ex vivo loading conditions. In vivo protocols are available from the literature to determine rib and vertebra kinematics but new developments are needed to improve data processing concerning CVJ behaviour obtained from discrete breathing positions showing limiting ranges-of-motion and sensitive to noise. Data from previous work were used to implement a method analyzing mean helical axis (MHA) and pivot point parameters of the CVJ complexes. Several levels of noises were estimated within Monte-Carlo simulations to optimize MHA results. MHA parameters were then used to transform and define a CVJ-specific local coordinate system. This study proposes an improvement for CVJ kinematics processing and description from in vivo data obtained from computed tomography. This methodology emphasizes the possibility to work with variability of MHA parameters using Monte-Carlo procedures on anatomical landmark coordinates and to define a local coordinate system from this particular joint behaviour. Results from the CVJ joint model are closer to a hinge joint (secondary motions inferior to 3°) when anatomical frames are expressed from MHA orientation. MHA orientation and position data obtained from the proposed method are relevant according to angular dispersion obtained (from 7.5° to 13.9°) and therefore relevant to define behaviour of CVJ.