Near InfraRed (NIR) frequency-domain techniques are promising clinical imaging tools, due to their ability in providing metabolic information about healthy and cancerous abnormalities (tumours, cysts, haematomas, etc.) embedded in thick biological highly diffusive tissues. Their main application is breast cancer screening. Detection, localisation and characterisation of heterogeneities are made possible by analysing the diffraction effects undergone by the intensity-modulated NIR signal (called a Diffuse Photon Density Wave (DPDW)) when propagating through the turbid medium. Imaging with DPDWs involves developing an algorithm that allows, on the basis of NIR data, the reconstruction of structural and optical properties of objects embedded in turbid media. This thesis proposes a reconstruction algorithm based on concepts from optical diffraction tomography. Its great advantage over other algorithms is to combine the two crucial characteristics of being non-iterative and of allowing full 3D reconstruction of arbitrarily shaped heterogeneities. This model, valid within the first Born approximation, greatly extends the domain of validity of the existing K-space spectrum analysis model, which provides reliable information only in the heterogeneity centre plane. The model developed in this thesis involves plane wave illumination, but an adaptation to spherical wave illumination is possible as well. The performance of our reconstruction algorithm is illustrated numerically, on the basis of a catalogue of forward data. These data are simulated using an analytical forward DPDW solution for plane wave illumination, derived from the existing spherical wave solution adapted from the Mie scattering theory. Knowledge of these solutions allows one to evaluate numerically the influence of typical structural and optical breast-like parameters on the propagation of DPDWs. This study provides important information about requirements to be fulfilled for optimizing a frequency-domain set-up within the framework of breast cancer screening. Experimental highly-sensitive frequency-domain NIR measurements performed on breast-like phantoms complete this study. By comparing experimental data to data simulated with the Mie spherical wave solution, we demonstrate the possibility of detecting very small optical heterogeneities (5-mm diameter), that induce distortions in DPDWs of 0.5% in amplitude and 0.14° in phase. These are, to our knowledge, the most sensitive frequency-domain measurements ever reported in thick highly diffusive media. Additional information about phantoms structure is provided by x-ray and computed tomography data. The presence of small air-bubble-like artefacts is detected. These objects are shown to have a significant disturbing effect on frequency-domain data. Their impact on the propagation of DPDWs is a crucial study in NIR imaging, since artefacts are also found in biological tissues (cutaneous or tissue lesions, scars, beauty spots, tattoos, etc.).