Immersed boundary methods, such as the Finite Cell Method (FCM), have gained significant attention in structural mechanics for their ability to simulate complex geometries without the need for body-conforming meshes. By combining high-order finite elements with simple Cartesian grids, the FCM significantly streamlines the pre-processing pipeline for heterogeneous and microstructured materials. This talk presents an integrated image-based framework that bridges the gap between CT-scan data and high-fidelity simulation. We employ an L2-projection to transform voxel-based geometries into smooth level-set functions, which serve as the basis for capturing internal material interfaces. To accurately represent the weak discontinuities at these interfaces without refining the background mesh, a local enrichment of the ansatz space is implemented. Furthermore, we extend this framework to the simulation of brittle fracture by incorporating a phase-field modeling (PFM) approach. This combination allows for the robust prediction of crack initiation, propagation, and branching within complex microstructures. The result is an efficient "image-to-simulation" workflow that bypasses traditional meshing challenges while maintaining high accuracy in predicting both material response and structural failure. The framework's capabilities are demonstrated by analyzing cemented granular materials, with numerical results compared against experimental data.