Today, the combination of Computer-Aided Design (CAD) and Additive Manufacturing (AM) offers the opportunity to create geometries of almost unlimited complexity. A striking example is lattice structures, which are architected structures providing exceptional mechanical properties (e.g., lightweight yet high mechanical stiffness) in a relatively straightforward manner. However, AM-induced defects are common in 3D-printed metallic lattices. Studies have shown that these defects can significantly affect the mechanical response of as-manufactured lattices compared to their as-designed counterparts. Investigating these defects requires a digital twin of the lattice, meaning a specimen-specific model that incorporates the actual geometry of the structure. The starting point for obtaining such a model is X-ray micro-Computed Tomography (μ-CT), a scanning method that provides 3D voxel-based images of the part's internal structure. Using such images as input, three main categories of methods can be identified to construct an accurate mechanical model. The oldest method involves generating a Finite Element (FE) model using the well-known marching cubes algorithm. A second approach relies on immersed methods, where the geometry is directly represented by voxels [Korshunova et al. 2021]. Finally, a third option involves template-fitting techniques [Willems et al. 2024]. This class of methods begins with the construction of a parameterized reference geometry, known as the template geometry, which captures the essential features of a specimen. This geometry is then deformed to match the scan data. In this work, a template-fitting method has been specifically developed for lattice structures [Bichet et al. 2025]. Using a virtual image correlation method, regularized by a thin membrane model, this approach allows for the construction of a multi-patch B-spline model to fit image data accurately, thereby capturing the geometric defects that occur during lattice fabrication. Then, a multi-patch B-spline volume mesh of the as-manufactured part is generated. Finally, this mesh is used to predict the behavior of the as-manufactured lattice structure with high accuracy and at a reduced computational cost by using isogeometric analysis.