Image-guided Surgery

Surgical spine marker

Figure 1: Fiducial marker for spinal procedures. Markers such as this one are implanted percutaneusly before spinal procedures. They are easily detected and tracked in X-ray projection images.

In collaboration with Prof. C.R. Maurer at Stanford University's Department of Neurosurgery and Prof. J. Denzler in the Department of Computer Science at the University of Jena (Jena, Germany) we are interested in developing methods for surgical target tracking without the commonly used invasive fiducial markers (Fig. 1).

CyberKnife System

Figure 2: CyberKnife™ System (Accuray Inc., Sunnyvale, CA) for image-guided frameless stereotactic radiosurgery. Photo courtesy of John R. Adler, Department of Neurosurgery, Stanford University.

In image-guided surgery, surgical instruments or other invasive means of therapy are tracked and targeted inside the patient using a typically three-dimensional image of the patient. For the procedure to be accurate, the position of the patient (more precisely: the target inside the patient) must be known at all times. Thus, the patient position in the OR must be determined initially, and the patient motion must be tracked during the procedure.

We have developed methods for fast and accurate registration of 2D X-ray projection images and pre-operative 3D CT images using Attenuation Fields, a specialization of Light Fields that are used in Computer Graphics for fast 3D rendering. We have also pioneered the use of region tracking algorithms from computer vision (e.g., Hyperplane Tracking) for motion backprojection. These methods are capable of true realtime performance, accurately tracking regions in two X-ray projections at more than 30 frames per second on a single Pentium 4 CPU.

The following graphic illustrates the general principle of motion backprojection, where a consistent 3D motion estimate is generated from the in-plane motion detected in multiple 2D projection images:

Schematic illustration of motion backprojection

X-ray with tracking region

Figure 3: X-ray image of the cervical spine with a typical tracking region (rectangle). Since this image originates from an actual patient, three implanted fiducial markers are clearly visible.

Figure 3 shows an X-ray image from an actual patient with a tracking region defined around the vertebra of interest. This figure also nicely shows the implanted fiducial markers still in use today.
  1. T. Rohlfing, J. Denzler, D. B. Russakoff, C. Gräßl, and C. R. Maurer, Jr., “Markerless real-time target region tracking: Application to frameless stereotactic radiosurgery,” in Proceedings of 9th Fall Workshop Vision, Modelling, and Visualization, November 16-18, 2004. Stanford, CA (B. Girod, M. Magnor, and H.-P. Seidel, eds.), (Berlin, Germany), pp. 5-12, Infix, Akademische Verlagsgesellschaft Berlin, Nov. 2004.
  2. T. Rohlfing, D. B. Russakoff, J. Denzler, and C. R. Maurer, Jr., “Progressive attenuation fields: Fast 2D-3D image registration without precomputation,” in Medical Image Computing and Computer-Assisted Intervention - MICCAI 2004. 7th International Conference, St. Malo, France, September 26-28, 2004, Proceedings, Part I (C. Barillot, D. P. Haynor, and P. Hellier, eds.), vol. 3216 of Lecture Notes in Computer Science, (Heidelberg), pp. 631-638, Springer-Verlag, 2004.
  3. D. B. Russakoff, C. Tomasi, T. Rohlfing, and C. R. Maurer, Jr., “Image similarity using mutual information of regions,” in Computer Vision - ECCV 2004: 8th European Conference on Computer Vision, Prague, Czech Republic, May 11-14, 2004. Proceedings, Part III, vol. 3023 of Lecture Notes in Computer Science, (Heidelberg), pp. 596-607, Springer-Verlag, 2004.
  4. T. Rohlfing, C. R. Maurer, Jr., D. Dean, and R. J. Maciunas, “Effect of changing patient position from supine to prone on the accuracy of a Brown-Roberts-Wells stereotactic head frame system,” Neurosurgery, vol. 52, pp. 610-618, Mar. 2003.

URL: http://www.stanford.edu/~rohlfing/research/igs/index.html
Last updated September 08 2011 16:24:19.
Torsten Rohlfing, Ph.D., torsten@synapse.sri.com
SRI International, Neuroscience Program
333 Ravenswood Avenue, Menlo Park, CA 94025-3493, USA