Surgeons operate with real-time imaging

Providing physicians with real-time 3-D image data during surgery enables operational precision and instant adjustment.

Sep 1st, 2000
Th Icon Spotlight

By Andrew Wilson,Editor

Click here to enlarge image

Providing physicians with real-time 3-D image data during surgery enables operational precision and instant adjustment.

Software that provides physicians with three-dimensional (3-D) data reconstructed from computed-tomography (CT) and magnetic-resonance-imaging (MRI) scanners has proved invaluable in preoperative surgical planning. When used in conjunction with patient registration systems and high-performance workstations, this software allows surgeons to visualize how to virtually perform specific operations before surgery begins.

One of the drawbacks of these imaging systems, however, is that as surgery progresses, anatomical structures may change. In brain surgery, for example, the brain swells when the skull is opened, and the structure of the brain changes as tumor tissue is removed.


To compensate, interoperative imaging procedures are being developed at both the Massachusetts Institute of Technology (MIT; Cambridge, MA) and Harvard Medical School (Boston, MA). Because these procedures perform image capture, processing, and display in real time, as the operation proceeds, surgeons are presented with more accurate 3-D images of the brain. And should unexpected complications such as hemorrhaging occur, these can also be visualized and corrected in real time.

Click here to enlarge image

"Typically," says David Gering, graduate student in the MIT Artificial Intelligence Laboratory, "interventional imaging is performed with x-ray fluoroscopy or ultrasound. More recently, interventional CT or MRI systems have been used," he adds. "However, compared with these modalities, open MRI systems provide high soft-tissue contrast, protect both patient and surgeon from radiation exposure, and offer continuous access to the operative field."

To perform image capture, Gering and his colleagues use an open-configuration intraoperative MRI system (Signa SP) from General Electric Medical Systems (Milwaukee, WI). With this system, patients are imaged during interventional and surgical procedures. The physicians use not only direct visualization but also real-time generated images. To register the imaging plane between the patient and the 3-D images being reconstructed by the system, an optical tracking system from Image Guided Technologies (Boulder, CO) is used.

FIGURE 1. Running 3D Slicer software, jointly developed by researchers at the Massachusetts Institute of Technology and the Brigham and Women's Hospital, the GE Signa SP system processes MRI scan data to form 3-D surface models of the skin (top) and to highlight tumors (shown in green). Functional MRI data are segmented to build models of the brain integrated in a 3-D view along with three slices obtained via MRI images. The MRI slice data are also shown in 2-D views (bottom).
Click here to enlarge image

During a surgical procedure, light-emitting diodes (LEDs) attached to the surgical instruments or probes are tracked by three CCD cameras mounted on a rail above the operating scene. This bar-like apparatus establishes the position of a probe in relation to the LEDs located in the area around the patient's head and reports the probe's position and orientation to the Signa SP workstation console. To perform 3-D image reconstruction in real time, the Signa SP imaging workstation is connected using a TCP/IP-type communications interface to an Ultra 30 visualization workstation from Sun Microsystems (Mountain View, CA).

"Although existing software packages, such as Analyze from the Mayo Foundation (Rochester, MN), MEDx from Sensor Systems (Sterling VA), and MNI from the Montreal Neurological Institute of McGill University (Montreal, Quebec, Canada), present an extensive collection of tools for data analysis and planning, they do not support surgical guidance," says Gering. Although systems have been developed to facilitate surgical guidance tracking, they feature lean support for analysis and planning and are not designed to incorporate intraoperative updates. "Interleaving the results of several of the currently available systems can be cumbersome and time-consuming, making it impractical for clinical applications," says Gering.

Accordingly, image reconstruction is performed using 3D Slicer software developed jointly at the MIT Artificial Intelligence Laboratory and the Surgical Planning Laboratory at Brigham and Women's Hospital (Boston, MA). Available for both Sun Solaris and Windows operating systems, this software can be freely downloaded from the Web at

Running on the Ultra 30 workstation, the 3D Slicer software aligns MRI data sets, extracts structures such as vessels and tumors from the data, and generates 3-D surface models for viewing the segmented structures. In addition to providing 3-D visualization, the software can measure distances, angles, surface areas, and volumes of MRI scan data. (see Fig. 1).

"Integrating the 3D Slicer software with an open MRI scanner such as the GE Signa SP augments intraoperative imaging with a full array of preoperative data," says Gering. "Now, the same analysis previously reserved for preoperative data can be applied to exploring the anatomical changes as the surgery progresses," he adds. Surgical instruments are tracked and drive the location of reformatted slices. Real-time scans are visualized as slices in the same 3-D view along with the preoperative slices and surface models.


The 3D Slicer software is connected through a socket—a software object—to a server created on the Signa SP imaging workstation. Consequently, the 3-D software can send and receive TCP/IP messages by opening the socket and reading and writing data to and from the socket. This approach simplifies program development because programmers need only to manipulate the socket and can rely on the Solaris software to transport messages across the network correctly.

FIGURE 2. To perform 3-D image reconstruction, the MRI data are reconstructed on a Sun Ultra 30 workstation configured with two Sun Creator3D graphics accelerator cards. One card is used to drive a 20-in. monitor display in the control area of the surgical suite, and the second card outputs 3-D images in NTSC video format that are displayed on an LCD panel in the scanner gantry (top left).
Click here to enlarge image

"Whenever the probe's position or orientation is updated or a new image is scanned, the Signa SP imaging workstation server sends the new data to the Ultra 30 workstation," says Gering. "When the 3D Slicer opens a connection, the server spawns a new process (called a child) to perform the serving and the originating (parent) process goes to sleep (idle mode). In this way, if the server crashes, the 3D Slicer merely reconnects to the sleeping parent process and surgery continues uninterrupted," he says.

To speed 3-D visualization, the Ultra 30 workstation is configured with two Sun Creator3D graphics accelerator cards. One card is used to drive a 20-in. monitor display in the control area of the surgical suite; the other card outputs NTSC video and displays the 3-D view on a liquid-crystal display in the scanner gantry (see Fig. 2). "The 3D Slicer was developed on top of Sun's OpenGL graphics library using the Visualization Toolkit (VTK) from Kitware (Clifton Park, NY) for processing and the Tcl/Tk scripting language from Ajuba Solutions (Mountain View, CA) for the user interface," says Gering.

The VTK provides a set of objects written in C++ that can be chained together to form a data-processing pipeline. Pipelined processing maximizes interactivity because the output of each stage is stored in memory, and any update from the user-interface controls triggers a change at a minimal depth into the pipeline. "Several classes or categories of objects were added to the VTK by deriving them from existing documented classes, which results in well-defined inputs and outputs," says Gering.


While the 3D Slicer software runs under the Solaris operating system during a surgical procedure, the surgeon points at the intended craniotomy, for example, at various angles using the tracking pointer. As this pointer moves within the surgical field, its imaging data are rendered in a 3-D view, and the reformatted slice planes follow its position, sweeping through the volumes. The surgeon can verify and alter the planned procedure by visualizing it on the display relative to all the surface models of critical structures. Reformatted images can be automatically generated along this path with the click of one button.

"A single 2-D view cannot reveal all hazards, but when two slices are reformatted in the plane of the locator perpendicular to each other, multiple hazards in 3-D space can be seen simultaneously," says Gering. Moreover, Gering claims that the system is the first to implement an open MRI scanner with a full array of preoperative data.

"Now," he says, "the analysis previously reserved for preoperative data can be applied to exploring anatomical changes that occur as surgery progresses. To date, the system has been used in more than 30 neurosurgical cases at Brigham and Women's Hospital. It continues to be routinely used for between one and two operations weekly.

Using the 3-D medical imaging system developed jointly by Brigham and Women's Hospital and GE Medical Systems, surgeons can operate within the magnetic space of the Signa SP MRI scanner. This procedure allows more accurate images of the brain to be visualized as imaging data are captured, processed, and displayed in real time as the surgical operation proceeds.

Company Information

Ajuba Solutions
Mountain View, CA 94043

Brigham and Women's Hospital
Boston, MA 02115

GE Medical Systems
Milwaukee, WI 53201

Harvard Medical School
Boston, MA 02115

Image Guided Technologies
Boulder, CO 80301

Clifton Park, NY 12065

Massachusetts Institute of Technology
Cambridge, MA 02139

Mayo Clinic
Rochester, MN 55905

McConnell Brain Imaging Centre
Montreal Neurological Institute of McGill University
Montreal, Quebec H3A 2B4, Canada

Sensor Systems
Sterling VA, 20164

Sun Microsystems
Palo Alto, CA 94303

More in Life Sciences