In car side impacts, the shoulder is directly exposed to injuries. These lesions are rarely fatal but produce long term impairments. To develop protection devices, numerical tools such as finite element (FE) body models are used. If the first FE body models represented average males, geometrical personalization methods are now available and enable taking into accountthe large inter-individual variability. Actually, an average model cannotbe used for predicting injuries for a subject whose anthropometry differssignificantly from that of a 50th percentile model. It is not common to find validated shoulder FE models and these last models represent 50th percentile male and did not allow the prediction of fractures. Thus, the purpose of this study is to assess an upgraded version of the shoulder part of the 50th percentile male HUMOS model and to assess the ability of geometrical personalization methods to improve shoulder injury risk prediction. Concerning the isolated shoulder bone models, i.e. clavicle and humerus models, their responses under dynamic loading were close to the experimental ones. The upgraded shoulder model, when submitted to lateral impacts at different speeds and with different directions, produced high validation scores allowing us to consider the quality of its validation as acceptable. However, at higher velocity, the 50th upgraded shoulder model did not produce realistic injurious consequences. The geometrical personalization process appears to be efficient as upgraded personalized model responses were closer to experimental results than upgraded model responses. The clavicle thickness appeared as one of the geometrical parameters that has the most importance in the geometrical personalization. Actually, bone thickness is linked with age and adjusting the geometrical parameters of cortical bone,allows for the effects of aging to be taken into account. Thus, if the upgraded model representing an average model is validated against non-injurious tests, only models seem able to produce realistic injuries when submitted to shoulder side impacts. This first step towards shoulder injury prediction by means of a FE model is promising since the upgraded personalizedshoulder model can simulate realistic shoulder injuries. Furthermore, this study puts forward the necessity of geometrical personalization. For thecovering abstract see ITRD E141762.
Samenvatting