In the very-high-cycle fatigue (VHCF) regime, fatigue cracks in high strength metallic components typically initiate from the interior of the material, often at local defects such as non-metallic inclusions or pores. Accurate numerical prediction of crack initiation and early crack growth therefore requires a realistic representation of these internal defects. X-ray computed tomography (XCT) provides detailed three-dimensional voxel data of defect geometries. However, converting this data into high-quality finite element (FE) meshes is a time-consuming and poorly automatable process. Moreover, common mesh-preparation steps such as smoothing or geometric simplification inevitably alter the original “ground truth” represented by the XCT data. To overcome these limitations, the Finite Cell Method (FCM) is investigated as an alternative numerical approach. In FCM, the computational mesh does not need to conform to the physical geometry. Instead, the distinction between physical and fictitious domains is introduced via a scalar indicator function that modifies the global stiffness matrix. This enables the direct use of complex defect geometries without explicit meshing of their boundaries. As a first step toward defect-based VHCF simulations, convergence studies are performed using a fundamental benchmark problem: an infinite plate with a central hole (Kirsch problem). The ability of FCM to accurately capture stress concentrations is assessed, and the influence of the polynomial approximation order and cell refinement strategies on solution accuracy is analysed. The discretization approach is then applied to an additively manufactured specimen with a synthetic defect designed for VHCF fatigue experiments.