< X-ray mammography for breast cancer diagnosis Dr Maud Poissonnier maud.poissonnier@oriel.oxford.ac.uk Medical Vision Lab University of Oxford Breast cancer is the most common form of cancer amongst women in the Western world affecting approximately 10% during the course of their lives. The earlier the diagnosis, the better the prognosis. X-ray mammography and cytopathology, ultrasound and Contrast Enhanced Magnetic Resonance Imaging (CE-MRI) are currently the main imaging modalities for breast cancer diagnosis and therapy monitoring. X-ray mammography is the first imaging method used to inform clinical decision making in a screening environment. As an application of image analysis, mammography is however extremely tough: the images have poor signal to noise ratio, there is only weak control of the imaging process, images are densely textured and blurred by scattered radiation, and the signs that are clinically significant are extremely subtle. Highnam and Brady have shown how a representation of the non-fatty compressed breast tissue, the hint representation, can be derived from a single mammogram. Firstly, we will show that its importance is two fold: normalisation (it removes all those changes in the image that are due only to the particular imaging conditions, e.g., the film speed or exposure time), and tissue segmentation. Second, we will show how we extended Highnam and Brady's model, which assumes that there is no calcification, by carrying a joint mammographic tissue segmentation and microcalcification detection.