Stereo matching of skull landmarks

Stereotaxis is a widely available neurosurgical technique that uses a frame attached to a patient's head as a reference system for precision treatment. Radiosurgery builds on this concept by coupling stereotaxic localization with a highly collimated radiation source. Because the use of an external frame of conventional stereotaxis precludes the use of radiosurgical therapy in many circumstances, we have developed a novel frameless method for determining tumor localization. Our stereotaxic system can determine skull orientation and position by recognizing at least three corresponding landmarks on Computerized Tomography (CT) and two x-ray projections. A principal goal of the research has been to obtain the position and orientation of the skull by recognizing landmarks such as the sella turcica or the mastoid from x-ray images.; The position of a skull in an operation space can be determined by matching three non-collinear features between x-ray images and the CT model. A skull coordinate system can be built from the landmarks. Using CT data the position of a tumor in space is known with respect to the skull coordinate system. We can utilize knowledge about skull features to recognize them in x-ray images: the relative positions and generic shape of these landmarks like the sella and mastoids are known from CT. Moreover we can measure the size of these landmarks from CT images, and predict their 2D shape on x-ray images. After locating edges from x-ray images, we can recognize landmarks using their shape, size and contrast prediction generated from CT data. With a calibrated x-ray imaging system and feature matches, the position and orientation of the skull is determined. Upon implementing such an algorithm the positioning error has been found to be approximately 2mm in experiments.