Semi-automatic Registration for Image Guided Surgery

N.Glossop¹ PhD, R.Hu² MD, G.Dix² MD, Y.Behairy² MD

¹Traxtal Technologies, Houston, Texas, USA
²Foothills Hospital, Spine program, Division of Orthopaedics and Neurosurgery, University of Calgary, Calgary, AB, Canada

© Copyright 1999 by Neil Glossop

From a poster presented at CAOS '99 , 4th meeting: March 17-19, 1999 in Davos, Switzerland

INTRODUCTION

Problem

• Registration is one of the most difficult and time consuming parts of using IGS equipment
• Anatomy-based registration requires large exposure and distinct anatomy to register (eg spine)
• It takes time to do a good registration
• Paired point often supplemented with surface fitting takes additional time
• C-arm and ultrasound based techniques are difficult to implement and suffer from poor images

Our Solution

• Apply a radiolucent, dynamic reference body (DRB) to the anatomy before CT scanning
• Apply radio-opaque fiducials to the carrier at known positions relative to the origin of the DRB
• Use a passive dynamic reference body and Polaris® (NDI, Waterloo, Canada) hybrid camera
• The instant the camera sees the DRB, the position of then fiducials are known
• Correlate fiducial positions in camera space with the image space coordinates in the CT
• No need to touch the fiducials with a probe, which will reduce error further

METHODS AND MATERIALS

1. The radiolucent passive dynamic reference body (Traxtal Technologies, Houston, Texas) is made of autoclavable plastic and contains 7 fiducials.

2. The DRB is mounted on the anatomy near the operative site. In this case the iliac crest for an iliosacral lag screw. A locking screw (Traxtal) is used to secure the plastic DRB. The patient is scanned to include the DRB.

3a) The DRB is shown on the 3D reconstruction on the SNN Scout system (SNN, Mississagua, Ontario, Canada). A “translucent” window shows the registration fiducials in yellow through the trackers.

3b)The cadaver with attached DRB’s. One DRB was used at a time. The DRB mounted on the right iliac crest was used to place the wire in the left side and vice versa, simulating a unilateral disruption i.e. everything attached to the tracked side could be assumed to behave as a rigid body, including the sacrum.

4. The fiducial locations are identified on the preoperative scans as registration locations for paired point registration. Their locations are recorded in a “preregistration” file.

5. A transformation matrix is calculated using our own software and loaded into the SCOUT as if it were a valid registration. Since the positions of the fiducials are accurately known in the space of the dynamic reference, the real fiducial points are never matched to the image points with the workstation’s own software. Instead, the positions are matched using our own software based on the fiducial locations deduced from the CAD data of the frames.

6. The entry point is located with a tracked drill guide. Images such as those shown below are used to identify the ideal position and trajectory of the wire.

7. A stab incision is made and the drill guide placed on the bone. The position and angle
of the drill guide is refined by looking at the screen. A guide-wire is then drilled

These images were recorded intraoperatively immediately before we placed the wire. The yellow arrow shows the trajectory in axial and sagittal oblique reformats that we thought was the best position to place the lag screw to avoid critical structures.

Images recorded post-operatively. The yellow arrow superimposed over the white line shows where we actually placed the wire. This pair of images shows good agreement with the intraoperative images above.

CONCLUSIONS

1. There was excellent correlation between the intraoperative target and the postoperative positions of both the wires drilled into this cadaver.
2. All critical structures were avoided.
3. The semi-automatic registration system functioned well and is indicated wherever pins DRB mounting pins can be placed prior to scanning.
4. Automatic registration can be applied to pelvis, spine, long bone fractures etc. and ideal for minimally invasive procedures.
5. Technique can be used will all existing IGS systems with minimal software changes
6. The technique will also work with stereo x-ray registration techniques and does not require prescanning with pins in place.
7. The method is fast, simple and accurate.

ACKNOWLEDGEMENTS

We wish to thank David Crouch and Northern Digital Inc. (Waterloo, Canada) for assistance and inspiration for this project. We also wish to thank the SNN (Mississagua, Canada) for supplying the SCOUT image guided surgery workstation used in this project and the University of Calgary Gross Anatomy lab for facilities. Traxtal Technologies provided all of the instrumentation used in this study.