Non-invasive 3-D patient registration for image-guided skull base surgery

Image-guided, computer-assisted surgery systems require a common reference between the preoperative image data and the corresponding patient pathology. Therefore, the problem of an accurate and reliable 3-D patient registration and referencing has to be solved. Considering that during image data acquisition patient misalignment and movement can arise in three dimensions, an automatic image registration method must be based on a three-dimensional approach. The method presented in this paper provides accurate 3-D patient registration for correction of these movement errors during data acquisition and active patient referencing to update the position of the head during surgery. Experimental data concerning the accuracy of repeated positionings of our patient registration and reference system demonstrated a high degree of accuracy with mean spatial errors of 0.82 mm +/- 0.31 mm in a plastic skull and 1.56 +/- 0.76 mm in patients, respectively. Results suggest that non-invasive 3-D patient registration for image-guided surgery may be a precise and useful method for computer-assisted identification of anatomical structures. Copyright (C) 1996 Elsevier Science Ltd.