Online head motion tracking applied to the patient registration problem.

Image-guided systems for surgical procedures in the region of the head require a method to correlate the diagnostic image data with the corresponding site of pathology in the patient. Considering that patient movement can occur, detection and correction of such movement errors during the acquisition of images is a basic prerequisite for accurate treatment. For this reason, we developed a new registration method based upon on-line tracking of the patient's head to solve the problem of registration in the presence of head motion. The method provides non-invasive active patient registration for correction of movements during imaging and continuous update of the patient's head position during surgery. The patient motion correction applies the rigid body model to register the images using feature correspondence. The new registration method is described, and results of experiments that were performed to evaluate its accuracy and reliability in a plastic skull model and in patients are presented. The error analysis resulted in a final target registration error of 0.90 mm +/- 0.16 mm using experimental model data and 1.58 mm +/- 0.26 mm using clinical patient data. In addition, the residual registration error is modeled as a function of the measured and predicted head motion in order to determine the error that is introduced by motion tracking during image data acquisition. Furthermore, the clinical application of the method is demonstrated for oto-, rhino-, and neurosurgical procedures in the region of the head.