Technical difficulties associated with excavation works in tectonized geological settings are frequent. They comprise instantaneous and/or delayed convergence, sudden collapse of gallery roof and/or walls, outpouring of fault-filling materials and water inflows. These phenomena have a negative impact on construction sites and their safety. In order to optimize project success, preliminary studies on the reliability of rock material found on site are needed. This implies in situ investigations (surface mapping, prospective drilling, waterflow survey, etc.) as well as laboratory investigations on rock samples (permeability determination, moisture and water content, mineralogy, petrography, geochemistry, mechanical deformation tests, etc.). A set of multiple parameters are then recorded which permit better insight on site conditions and probable behavior during excavation. Because rock formations are by nature heterogeneous, many uncertainties remain when extrapolating large-scale behavior of the rock mass from analyses of samples order of magnitudes smaller. Indirect large-scale field investigations (e.g. geophysical prospecting) could help to better constrain the relationships between lithologies at depth. At a much smaller scale, indirect analytical methods are becoming more widely used for material investigations. We discuss in this paper X-ray computed tomography (XRCT) and neutron tomography (NT), showing promising results for 3D petrographical investigations of the internal structure of opaque materials. Both techniques record contrasts inside a sample, which can be interpreted and quantified in terms of heterogeneity. This approach has the advantage of combining genetic parameters (physico-chemical rock composition) with geometric parameters resulting from alteration or deformation processes (texture and structure). A critical analysis of such 3D analyses together with the results of mechanical tests could improve predictions of short- and long-term behavior of a rock unit. Indirect methods have the advantage of being non-destructive. However, as it is the case with large- scale geophysical surveying, XRCT and NT are affected by several error factors inherent to the interaction of a radiation modality (X-ray or neutron beam) with the atomic structure of the investigated materials. Recorded signals are therefore in particular cases not artifact-free and need to be corrected in a subsequent stage of data processing.