The goal of this research is to show the feasibility of detecting density variations in green-state powder metallurgy (P/M) compacts from surface voltage measurements. By monitoring a steady electric current flow through the sample and recording the voltages over the surface, valu-able information is gathered leading to the prediction of the structural health of the compacts. Unlike prior research that concentrated on the detection of surface-breaking and subsurface de-fects, the results presented in this thesis target the density prediction throughout the volume of the sample. The detection of density variations is achieved by establishing a correlation between the conductivity and their respective density. The data obtained from the surface measurements is used as part of an inversion algorithm, calculating the conductivity distribution, and subse-quently the density within the compact. In a first step, the relationship between conductivity and density of green-state P/M com-pacts was investigated. Tests were conducted for a number of parts of various powder mixtures. In all cases a clear correlation between conductivity and density could be established, indicating that measurements of electric conductivity could indeed be exploited in an effort to render valid information about the density of the sample under test. We found a linear correlation for non-lubricated parts and a non-linear behavior for lubricated samples. Specifically, it was found that the conductivity increases with increasing density only up to a maximum value obtained at ap-proximately 6.9g/cm3. Interestingly, any additional density increase leads to a reduction of the conductivity. This behavior was confirmed to be inherent in all powder mixtures with lubricants. The thesis research is able to provide a physical model and a mathematical formulation describ-ing this counter-intuitive phenomenon. A finite element solver in conjunction with an inversion algorithm was then implemented to study arbitrarily shaped part geometries. Based on the principles of electric impedance imag-ing, the developed algorithm faithfully reconstructs the density distribution from surface voltage measurements. The feasibility of the instrumentation approach for both simple and complex parts can be demonstrated using a new sensor concept and measurement arrangement. Measurements were performed on both geometrically simple and complex parts.