This report presents my research activity in Computer Graphic, more exactly in the field of the 3D reconstruction from various sources, mainly from medical imaging, but not only. The modelling of the soft (or deformable) objects we propose is generally based on discrete models and, for some time, continuous. The integration of multi-sources information in these triangular or tetrahedral meshes allows to simulating their deformations while respecting their natural behaviour. The main purpose is to supply computing tools which allow to guide the medical staff in its diagnosis or during a treatment operation. After this latter, the simulation models may allow a treatment validation. The originality: the integration of the rheology in the discrete models, to offer an alternative of useful simulation of the deformations during the operation, a contrario to greedier but precise, classic finite elements methods, especially when non-linearity conditions are required. By physical models of living soft tissues, we imply as well deformable biological elastic tissues (i.e. skin, liver, brain, pulmon), deformable fluid constituents (air), even coupling of these structures (cerebral ventricles, bronchial trees). Two scientific communities historically worked on the use and/or development of physical models of biological tissues: Computer Graphics and Medical Computing. For each, necessary skills for physical models development araise domains of (1) medical images segmentation, (2) geometrical reconstruction and volume meshing, (3) biomechanical modelling, (4) in vivo measure of tissues characteristics, (5) numerical analysis. Each of these points meets itself in the researches presented here, which are mainly turned to the contribution of a geometrical model oriented approach to complement another discipline.