X-ray-based craniofacial visualization and surgery simulation

X-rays penetrate both soft and hard tissues, which record only the accumulated density value of tissues. It is therefore impossible to represent the geometry of human tissues accurately with a small number of X-rays. CT can acquire the geometry of human anatomy accurately. It, however, subjects the patient to a high dose of radiation which in many cases is undesirable and unhealthy. This chapter presents a novel craniofacial visualization technique with the developments from both computer graphics and computer vision. It is a low-radiation, low-cost alternative to CT-based system for the reconstruction of 3D cranium using only three X-rays. We paste lead markers on the subject's face which allow a 3D face model to be constructed using correlated vision. Then the surface of the cranium is obtained by subtracting the soft tissue depth from the face surface. Because of the penetrating nature of X-rays, existing computer vision techniques are not effective in matching the corresponding points for X-rays. We present a new matching algorithm to solve this problem by evolutionary programming. We also designed a supervised learning method to estimate the soft tissue stiffness parameters.