Abstract

Women with breast cancer, whether screen detected or symptomatic, have both mammography and ultrasound for initial imaging assessment. Unlike X-ray or magnetic resonance, which produce an image of the whole breast, ultrasound provides comparatively limited 2D or 3D views located around the lesions. Combining different modalities is an essential task for accurate diagnosis and simulating ultrasound images based on whole breast data could be a way toward correlating different information about the same lesion. Very few studies have dealt with such a simulation framework since the breast undergoes large scale deformation between the prone position of magnetic resonance imaging and the largely supine or lateral position of ultrasound. We present a framework for the realistic simulation of 3D ultrasound images based on prone magnetic resonance images from which a supine position is generated using a biomechanical model. The simulation parameters are derived from a real clinical infrastructure and from transducers that are used for routine scans, leading to highly realistic ultrasound images of any region of the breast. © 2015 IEEE.