Compton cameras are a promising alternative to collimated cameras in single particle emission computed tomography (SPECT). They meet the challenge of imaging sources having poly- chromatic spectrum and low photon yields. Currently used SPECT cameras rely on mechanical collimation, imposing a trade-off between efficiency and resolution. As the photon energy increases, the radiation can penetrate the collimator and thus degrade the resolution. A solution is then to increase the thickness of the septa which further reduces the detection efficiency. Based on a coincidence detection system and on the kinematics of Compton scattering, Compton cameras restrict to a half-cone the set of possible origins of a detected photon. They allow an increase in efficiency of one or two orders of magnitude, at the cost of more complex image reconstruction techniques. Specific tomographic algorithms capable to invert conical Radon transforms are then required to produce the image of the source. In this talk we focus on iterative reconstruction methods. We will first highlight the importance of choosing a data model well adapted to the measurement uncertainties. Then we will discuss some regularization strategies in statistical image reconstruction. In particular we will present a fast and convergent algorithm for the non-differentiable total variation prior. We will illustrate its performances on low-statistics data issued from a complex-shaped mono-energetic source. In addition, this example will allow us to compare the performances of collimated and Compton cameras. We will end with an example from proton-therapy, where treatment monitoring could be realized by imaging the secondary prompt-gamma radiation. In this application, extremely low photon counts are expected in a wide energy range. This work was done in collaboration with Yuemeng Feng (PhD student, CREATIS), Elie Bretin (Institut Camille Jordan, Lyon), Anne Etxebeste (post-doc, CREATIS), David Sarrut and Jean Michel Létang (CREATIS), Étienne Testa and Mattia Fontana (Institut de Physique Nucléaire de Lyon) and Denis Dauvergne (Laboratoire de Physique Subatomique et de Cosmologie, Grenoble).