Video based surface alignment technique for 3D patient positioning

We present a technique to align a 3D surface model of a patient to video images of the patient on the treatment table. The technique can be used to quantify patient positioning errors in 3D. Three video cameras were mounted to the ceiling of the treatment room and geometrically calibrated in linear accelerator beam coordinate system. The calibration essentially determines how the cameras view the radiation isocenter. After therapists position the patient for treatment, video images record the setup. Camera calibration results are used to project the 3D skin surface of a treatment planning CT scan into the set of 2D video images. This method determines the 6 parameters of a rigid transformation that best aligns the skin surface from CT with the set of 2D video images. The correlation of pixel values of corresponding points in images of a pair of cameras is used as a measure of the quality of the alignment. Point correspondences are determined by projecting the 3D skin surface onto the patient images. Once optimized the 6 parameters are a measurement of the patient positioning error. Multiple pairs of cameras can be used to improve the alignment and increase the visible surface of the patient. Preliminary tests using this technique for breast, abdomen and head and neck treatments demonstrate its ability to match the skin surface well to the video images. This technique could be used to position patients before each treatment and reduce errors in patient positioning.