Coronary microcirculatory flow is governed by a number of determinants including network anatomy, systemic afterload and the mechanical interaction with the myocardium throughout the cardiac cycle. The range of spatial scales and multi-physics nature of coronary perfusion highlights a need for a multi-scale framework that captures the relevant details at each level of the network. Using a data driven structural based approach we have developed a set of novel methods to reconstruct three-dimensional coronary vasculature from images, comprised of two distinct sets of data – fluorescent microsphere beads and anatomically vasculature. Current state-of-the-art radii estimation methods and a model-based method were applied on whole-organ porcine data that was automatically reconstructed. Using this anatomical model blood flow simulations were performed to obtain blood flow rate, calculated using the conductance and pressure drop across each branch in the tree throughout the whole vasculature. The results from this blood flow simulation were compared to regional microsphere perfusion distribution. At each branch in the arterial tree, the number of microspheres that passed through it in the vasculature was counted. Shortest path tracing also revealed the main source coronaries for every branch in the arterial tree and the myocardial perfusion regions of these main coronaries. The regional microsphere distribution (number of microspheres counted going through a branch, normalized by the total number of microspheres counted in the entire perfused region of myocardium) was compared to the regional flow (flow rate in a branch normalized by the total flow rate in the root artery). To analyze the discrepancy observed between the microsphere perfusion and flow model, we studied the distribution of microspheres at bifurcations. Using this data we have constructed a model of this error parameterised by the branching angle and daughter-to-parent ratio of the vessel cross-sectional area is shown to more accurately predict the biased distribution of the microspheres. Incorporating the disproportionate microsphere distribution model at bifurcating segments with the largest branch angles, ranging from 100- 120 degrees, reduced the error from 24 to 7 percent error in between the fraction of flow predicted by the model and the fraction of microspheres.