We present a computational framework for modeling liver regrowth, integrating four key components: 1. a synthetic vascular tree generation method, formulated as a nonlinear optimization problem, capable of producing multiple non-intersecting trees within non-convex organ domains; 2. a multiscale perfusion model that couples synthetic vascular trees with a multi-compartment homogenized flow approach, incorporating homogenization to determine effective parameters; 3. a poroelastic finite growth model that acts on all compartments and the vascular tree structure; 4. an evolution equation for the local volumetric growth factor, driven by the homogenized flow rate through the microcirculation as a proxy for hyperperfusion. We validate the model against longitudinal µCT-based three-dimensional reconstructions of murine liver geometries and vascular architectures obtained after partial hepatectomy. In the standard 70% PHx protocol, the median and left lateral lobes are ligated and resected, removing approximately 70% of total liver mass. The experimental datasets enable quantitative comparison of organ-scale volume recovery, spatial growth patterns, and major vessel remodeling, complemented - where available - by hemodynamic indicators to evaluate perfusion redistribution and hyperperfusion-driven growth. Using five independent subjects, we perform cross-validation to assess predictive robustness with respect to inter-subject variability. Across animals, the model achieves consistent quantitative agreement with observations, demonstrating predictive rather than merely descriptive capability for liver regeneration.