Résumé : Mammography is a specific type of breast imaging that uses low-dose X-rays to detect cancer in early stage. During the exam, the women breast is compressed between two plates until a nearly uniform breast thickness is obtained. This technique improves image quality and reduces dose but can also be the source of discomfort and sometimes pain for the patient. Therefore, alternative techniques allowing reduced breast compression is of potential interest. The aim of this work is to develop a 3D biomechanical Finite Element (FE) breast model in order to analyze various breast compression strategies and their impact on image quality and radiation dose. Large breast deformations are simulated using this FE model with ANSYS software. A particular attention is granted to the computation of the residual stress in the model due to gravity and boundary conditions (thorax anatomy, position of the patient inside the MRI machine). Previously developed biomechanical breast models use a simplified breast anatomy by modeling adipose and fibroglandular tissues only (Rajagopal et al. in Wiley Interdiscip Rev: Syst Biol Med 2:293--304, 2010). However, breast reconstruction surgery has proven the importance of suspensory ligaments and breast fasciae on breast mechanics (Lockwood in Plast Reconstr Surg 103:1411--1420, 1999). We are therefore consider using a more realistic breast anatomy by including skin, muscles, and suspensory ligaments. The breast tissues are modeled as neo-Hookean materials. A physical correct modeling of the breast requires the knowledge of the stress-free breast configuration. Here, this undeformed shape (i.e., without any residual stress) is computed using the prediction--correction iterative scheme proposed by Eiben et al. (Ann of Biomed Eng 44:154--173, 2016). The unloading procedure uses the breast configuration in prone and supine position in order to find a unique displacement vector field induced by gravitational forces. The 3D breast geometry is reconstructed from MRI images that are segmented (Yushkevich et al. in Neuroimage 31:1116--1128, 2006) to differentiate the four main tissue types. The breast volume is discretized with a hexa-dominant FE meshing tool as a unique volume. Finally, the model is evaluated by comparing the estimated breast deformations under gravity load with the experimental ones measured in three body positions: prone, supine, and oblique supine.