Three-D data sets aid intraoperative navigation

October 5, 2005

Speakers at the CARS meeting effectively countered skepticism regarding surgical adoption of interactive 3D imaging tools by highlighting its practical uses in neurosurgery. Further use of 3D imaging in the operating room is far from inevitable if surgeons remain unconvinced of the need for such computerized anatomic reconstructions.

Speakers at the CARS meeting effectively countered skepticism regarding surgical adoption of interactive 3D imaging tools by highlighting its practical uses in neurosurgery. Further use of 3D imaging in the operating room is far from inevitable if surgeons remain unconvinced of the need for such computerized anatomic reconstructions.

Dr. Steffen Rosahl, vice chair of neurosurgery at Albert-Ludwigs University in Freiburg, Germany, outlined three key principles of making 3D imaging data sets a core part of intracranial neuronavigation.

The first is to uphold the motto "less is more," he said. While software packages may be capable of reconstructing an entire brain in exquisite detail, information irrelevant to the task in hand should be omitted from guiding images. Volumetric views should be translucent, enabling neurosurgeons to view their approach and the surgical target in a single 3D image. Finally, relevant functional information should be highlighted.

While such advice might be regarded as common sense, few commercially available 3D navigation packages include all these features, though research is progressing, Rosahl said.

"Most of the systems use tri-axial images and rely on surface renderings. What you want to incorporate in these systems is volumetric imaging. You want to get an image that matches not only the surface structures but also the deeper structures," he said.

Neurosurgical planning can also be greatly aided through the use of diffusion tensor imaging data to identify white matter tracts, said Dr. Christopher Nimsky, an associate professor of neurosurgery at the University of Erlangen-Nuremberg in Germany. Accurate identification of the pyramidal tract, for example, can ensure that patients undergoing brain tumor resections do not lose motor function.

Different options are available for displaying the tensor information. Glyph representations that provide a 3D color-coded voxel map (Figure 1) can be helpful in delineating the tumor's border, but they are unlikely to be used in surgical navigation systems, according to Nimsky. Fiber tract representations offer a more practical option. These can be generated by running a tracking algorithm on "seed" regions selected from the glyph representation.

Inclusion of DTI data within a 3D navigation system should allow surgeons to monitor their proximity to major white matter tracts. Practitioners should be aware, however, that these structures can move during surgery (Figure 2).

"Possible brain shift of major white matter tracts during surgery has to be taken into account when major tumor parts are resected," Nimsky said.

Neurosurgeons at Erlangen perform intraoperative and preoperative MR imaging on a 1.5T scanner to obtain an accurate representation of fiber tract positioning. Future work will include assessment of appropriate surgical safety margins, given the uncertainties in fiber tract representation on 3D images. The lower the quality of imaging data acquired, the greater the safety margin should be, he said.

In another presentation, Dr. Dirk Lindner from the department of neurosurgery at the University of Leipzig, Germany, demonstrated how intraoperative 3D ultrasound coupled with preoperative MRI can control for brain shift.

Investigators developed a freehand ultrasound navigation system by linking together a PC with a video grabber card, an optical tracking device, and a standard ultrasound device. Surgeons then used this setup to acquire 3D ultrasound data sets from 23 brain tumor patients after craniotomy and at several additional points throughout their operations. Each ultrasound data set was matched with preoperative MRI results to measure brain shift. Patients also underwent postoperative MRI within 24 hours of surgery to provide data on surgical outcome.

Intraoperative data acquisition, which took around five minutes, was performed from three to seven times for each patient. Fusion between the ultrasound and MRI data sets worked in all cases but one. Measured brain shifts ranged from 2 mm to 25 mm, with a mean shift of 5 mm.

Surgeons participating in the study were more accustomed to using MRI than ultrasound, and they underwent a considerable learning curve, according to Lindner.

Resection control was deemed good in 17 cases. Tumor remnants were found mostly in glioblastomas, where ultrasound failed to visualize tumor borders sufficiently well.

"The main problem is the differentiation between tumor and edema, especially in high-grade tumors. We think maybe we can solve this with the use of contrast enhancement, to give better visualization of vascular structures," Lindner said.