Hydraulic fracturing is a technique often used in the oil and gas industry to enhance the production of wells in low-permeability reservoirs. It consists of fluid injection at a sufficiently large injection rate to create and propagate tensile fractures inside the rock formation. The growth of a hydraulic fracture is a highly non-linear moving boundary hydro-mechanical problem. Many physical processes get tightly involved and influence the fracture growth, such as the elastic deformation of the material, fracture surface creation, viscous fluid flow inside the propagating fracture, and fluid leak-off into the solid principally. The linear hydraulic fracture mechanics (LHFM) theory describes well this coupled process, with adequate comparisons between predictions and observations in the linear elastic medium. However, deviations from LHFM predictions have been reported at the laboratory and field scales in some cases, notably a larger fluid pressure and a shorter fracture length. These deviations are often associated with the non-linear nature of rock fracture deformation. However, the impact of these material non-linearities on the coupled problem of hydraulic fracture propagation has not been fully understood. In this context, this thesis investigates numerically and experimentally the effect of the quasi-brittle nature of rocks on hydraulic fracture growth. We first solve the propagation of a hydraulic fracture using a power-law fracture length-dependent fracture energy to model macroscopically the effect of an enlarging process zone. The use of Gauss-Chebyshev quadrature and barycentric Lagrange interpolation techniques reduces the hydraulic fracture propagation problem to a set of non-linear ordinary differential equations. We find that this increasing apparent toughness leads to a slower fracture growth compared with constant toughness LHFM solutions and a faster transition to the toughness dominated regime. We show that the solution can be approximated by the self-similar constant toughness solutions with the instantaneous toughness value. We then explore the impact of the quasi-brittle nature of rocks by using a cohesive zone model in combination with elastohydrodynamic lubrication to simulate HF growth. We notably account for fluid cavitation at the fracture tip and the effect of fracture roughness on flow. The fracture growth finally converges toward the LHFM predictions but deviate from LHFM predictions when the cohesive zone cannot be neglected. These deviations indicate an increase of energy dissipation. The additional energy dissipation compared with LHFM mainly comes from the viscous fluid flow in the rough fracture process zone and is larger for a larger ratio between the in-situ stress and the material peak cohesive stress. We finally discuss HF experiments performed on two impermeable rocks under different confinement and injection conditions. We notably develop an inversion technique to reconstruct the fracture front geometry using diffracted waves recorded via an array of 32 sources and 32 receivers. We find that transmitted waves attenuate before the arrival of the fracture front estimated from diffracted waves, indicating the existence of a fracture process zone of centimeters extent. This fracture process zone is found to be comparable with the fracture extent, thus implying an important artifact of laboratory HF experiments performed in quasi-brittle rocks on finite-size centimeter-decimeter samples.