The recent years have witnessed growing interest in the power of scientific visualization for simulation-based neuroscience. The research presented in this thesis develops methods to generate physically-realistic visualizations of neocortical models reconstructed by the Blue Brain Project, a pioneering endeavor that uses a simulation-based approach to build large-scale, biologically-detailed digital reconstructions of neocortical microcircuitry. The project has established a domain-specific framework to visualize hybrid representations of these neocortical models, which it uses to evaluate their structure, composition and dynamics. The current framework offers naive, optical emission-only models that oversimplify light-tissue interaction and ignore other computationally expensive phenomena such as absorption, scattering and fluorescence. The methods presented in this dissertation overcome these limitations, making it possible to visualize neocortical models on a physically-plausible basis, taking into account the intrinsic optical properties of cortical tissue and the spectroscopic characteristics of its fluorescent structures. This requires rigorous optical models that accurately simulate light interaction with cortical tissue, and accurate volumetric models of the tissue itself which reflect its optical properties during simulation. We present an efficient framework for creating high fidelity large-scale volumetric models of neocortical microcircuitry that govern the way in which light is distributed in cortical tissue. These models are reconstructed in two steps: first, we build piecewise watertight mesh models of neocortical neurons from their morphological skeletons. Then, we apply solid voxelization to the meshes to build volumetric models. We also present two novel optical models for simulating light interaction with fluorescent structures in low- and highly-scattering volumes. These optical models are integrated into a high level framework that simulates the imaging pipelines of transmitted light brightfield, widefield epi-fluorescence and light-sheet fluorescence microscopes. It is based on the principles of geometric optics and Monte Carlo ray tracing. Afterwards, we exploit this framework to render in silico realistic microscopic images resembling those created by the actual instruments. In conclusion, we generalize the concept of physically-based simulation of brain imaging modalities and review the computational models required to simulate fMRI.