Abstract : In digital mammography two approaches exist to estimate image quality. In the first approach, human observer assesses the lesion detection in mammograms. Unfortunately, such quality assessment is subject to interobserver variability, and requires a large amount of time and human resources. In the second approach, objective and human‐independent parameters relating to image spatial resolution and noise are used to evaluate digital detector performance; even if these parameters are objective, they are not directly related to lesion detection. A method leading to image quality assessment which is both human independent, and directly related to lesion detection is very important for the optimal use of mammographic units. This PhD thesis presents the steps towards such a method: the computation of realistic virtual images using an “X‐ray source/digital detector” model taking into account the physical parameters of the detector (spatial resolution and noise measurements) measured under clinical conditions. From results obtained in this work, we have contributed to establish the link between the physical characteristics of detectors and the clinical quality of the image for usual exposition conditions. Furthermore, we suggest the use of our model for the creation of virtual images, in order to rapidly determine the optimal conditions in mammography, which usually is a long and tedious experimental process. This is an essential aspect to be taken into account for radioprotection of patients, especially in the context of organized mass screening of breast cancer.