Quantifying the in vivo mechanical properties of soft tissues is essential for developing patient-specific models for precision medicine. The Virtual Fields Method (VFM) determines material parameters from full-field deformation measurements by minimizing the difference between internal and external virtual work, avoiding the iterative forward solutions required by traditional inverse finite element methods. However, parameter estimates may depend on the arbitrary choice of virtual fields, and in vivo deformation measurements are commonly obtained from a pre-stressed reference state. For general boundary conditions, the pre-stress state depends on the unknown material properties and must be determined at each optimization iteration. Moreover, deformation measurements may not be available for the entire tissue volume due to imaging limitations. To address these challenges, we present a VFM-based optimization framework to identify hyperelastic material parameters of soft tissues using full-field deformations measured from a pre-stressed reference state. A virtual field is defined for each parameter as the displacement map resulting 1 from a perturbation of that parameter. To accommodate limited clinical deformation data, we developed a model-reduction strategy that replaces unobserved regions with nodal force boundary conditions. The algorithm is verified using synthetic displacement data from finite element simulations of the optic nerve head response to intraocular and intracranial pressure lowering. To reduce computational cost, the pre-stress states and boundary forces are updated periodically rather than at every iteration. We investigate the effect of update frequency on convergence rate and robustness of the algorithm against Gaussian noise. The results demonstrate that the algorithm provides a potentially efficient tool for determining hyperelastic material parameters from in vivo deformation maps.