In proton therapy cancer treatments, interactions between the beam and the patient's body lead to nuclear reactions that produce secondary prompt-γ radiation. It was shown that some correlation exists between the emission point of those photons and the Bragg peak position. Ongoing work aims to design imaging systems able to monitor the deposited dose by exploiting prompt-γ radiation, in particular the Compton camera. A 3D image of the prompt-γ cartography can be obtained with this device. The Compton camera detects a photon in two steps. The photon is scattered (at least once) in a first detector, then it is absorbed in a second detector. The hits, recorded with positions and energies, define an event. A Compton cone is defined for each event and the incoming path of the initial ray lies on its surface. The image of the source is then calculated by tomographic reconstruction from projections calculated on conical surfaces. However, several factors influence the quality of the reconstructed images. Because of the small extent and of the limited sensitivity of the camera, many photons are not detected. The spatial and energy resolutions of the detectors bring about measurement errors. Moreover, the image artifacts can be induced by random coincidences, uncertainties on the order of the photon interactions in the camera, incomplete absorptions. Other secondary particles emitted during the irradiation and the absorption in the volume, are also significant noise factors. In this study, we simulated the irradiation of a volume with a proton pencil beam. We modeled a Compton camera with geometry, spatial and energy resolutions inspired from the one currently developed at IPNL. All the secondary particles except γ-photons are withdrawn from the simulation. The photons are recorded one by one, in order to avoid random coincidences in the camera. Once the data are acquired, we reconstructed the 3D image using a list-mode MLEM (LM-MLEM) method where energy measurement errors could be considered. We compare the images obtained with and without attenuation in the patient with the true image of the simulated prompt-γ distribution. In both cases, we reconstruct the images with data acquired respectively with a perfect detector and with a realistic detector. The effects of the different noise sources on the ability to accurately detect displacements of the Bragg peak are showed. Preliminary results obtained in prompt-γ imaging are encouraging and will be further analyzed. However more effort has to be put in the development of the reconstruction algorithms in order to better account for the sources of noise in the images.