Analyse d’images par champs de Markov Application mammographie

Abstract

The mammography permits to visualize the breast as well as of possible lesions to his/her level. It is currently the best exam of tracking of the breast cancer. Indeed, the mammography permits to detect, at a woman whom has no symptom, of the tiny tumors non discerned to the palpation. It is also an exam of diagnosis that the physician or the gynecologist prescribes when he detects to the palpation an abnormal size or if the patient presents obstinate symptoms as an induration of the breast, an increase abnormal of volume, a nodule, a redness of skin, an out-flow abnormal of the nipple... The picture represents an inestimable wealth seen the diversity of the information that it conceals in it through these numerous contours and details. The medical picture is therefore a source of very important diagnosis for the physician. The pictures gotten from the numeric devices must be interpreted therefore correctly. Our project appears in this goal in order to present an unsupervised segmentation method which the formalism is relied on hidden Markov field. The originality of this method is, it takes into account structural information processed as flexible spatial. Our first results are very satisfactory.