by
John W. Peifer, Sr. Research Scientist
Biomedical Interactive Technology Center, Georgia Institute of Technology
W. David Curtis, MD
Department of Medicine, Medical College of Georgia
Michael J. Sinclair, Sr. Research Engineer
Interactive Media Technology Center/OIP, Georgia Institute of Technology
Researchers from the Georgia Institute of Technology and the Medical College of Georgia (GIT/MCG) have developed an interactive computer simulation of Endoscopic Retrograde Cholangio-Pancreatography (ERCP). ERCP is a minimally invasive technique for evaluating and treating pathologic conditions of the biliary and pancreatic ducts. While ERCP provides the patient with substantial advantages over traditional methods, ERCP requires advanced skills and extensive experience to minimize the risk of complications. Computer simulation offers many advantages for efficiently and safely training physicians in ERCP. The GIT/MCG proof of concept simulation provides realistic training with both visual and force feedback while an endoscopist practices the ERCP procedure.
ERCP involves passing an endoscope with a flexible tip through the oral cavity, the esophagus, the stomach and into the first portion of the small intestine, the duodenum. The endoscope is maneuvered in the duodenum until the major duodenal papilla is identified. The papilla is a nodular protuberance on the medial wall of the duodenum through which the bile duct and pancreatic duct expel bile and pancreatic juice respectively into the duodenum to aid in the process of digestion. After the endoscope is positioned near the major duodenal papilla, a catheter is inserted through a channel of the endoscope and directed into the desired ductal system (biliary or pancreatic) by changing the position of the flexible endoscope tip in relation to the papilla. An "elevator" apparatus at the tip of the endoscope can also be utilized to adjust the angle at which the catheter enters the papilla. To guide the catheter into the papilla, the physician relies on visual feedback from the endoscopic video display and tactile resistance encountered when advancing the catheter. In diagnostic ERCP, contrast dye is injected through the catheter while fluoroscopic images are observed and roentgenograms are obtained. On the fluoroscopic display, the catheter tip can be located, but, unless contrast is injected, the anatomical structures are essentially invisible. In therapeutic ERCP, instruments are used that can enlarge the opening of the ducts (sphincterotomy), collect biopsy samples, remove stones, and place stents.
The major benefit of ERCP is that it allows patients to avoid more invasive surgical or radiological procedures. The therapeutic applications of ERCP (removal of bile duct stones, repair of bile duct injuries, stenting of obstructed bile ducts, etc.) significantly lower the risk of infection, speed recovery time, and reduce the cost of delivering care. Therapeutic ERCP continues to be among the most rapidly evolving disciplines in gastrointestinal endoscopy. New applications are being aggressively developed to further lower patient risk and the cost of care. The major risks associated with ERCP include (a) pancreatitis (inflammation of the pancreas gland), usually resulting from repeated injections of the pancreatic duct or improperly performed therapeutic techniques (b) cholangitis (infection in the bile ducts) which can result when contrast material is injected into an obstructed biliary system and cannot be removed by appropriate drainage procedures, and (c) perforation of the duodenum, which can occur during more advanced therapeutic biliary procedures. As trainees become more adept at performing ERCP, the risk of these complications is greatly reduced. Thus, safe repetitive practice on a computer simulator is highly advantageous.
The GIT/MCG ERCP simulator consists of a simulation interface into which an endoscope is inserted, a computer which controls and updates the virtual environment, a dial and button box for selecting simulation parameters, and a video monitor to display computer generated imagery. A simulation session begins as a real endoscope is inserted through the "mouth" of the simulated patient. The endoscope is guided into position using standard endoscopic techniques. A position tracking system reports the endoscope movements to a high performance Silicon Graphics computer which controls the interactions and updates the computer generated imagery on the monitor. In addition to the visual feedback displayed on the monitor, a computer controlled arrangement of servo motors provides force feedback to the endoscope and catheter held by the trainee.
The simulation interface for the ERCP simulator, shown in Figure 1, includes the physical components that the trainee sees and controls. These components approximate the physical environment for clinical endoscopy. During ERCP, the patient lies on his side on an examination table. After sedating the patient, the physician guides the endoscope through the patient's mouth and then navigates through endoscopic manipulation into position while observing the internal endoscopic view on a video monitor beside the exam table.
Figure 1. GIT/MCG Endoscopic Simulation Interface
In the GIT/MCG simulator, the trainee holds a real endoscope (donated by Olympus) that has already been passed through the mouth in a mannequin head. Generally hidden from view by a hospital gown or sheet, the end of the endoscope is attached to a movable track inside the simulated patient. As the scope is advanced by the trainee, the scope slides smoothly along this moving track. At the locations where the scope is attached to the moving track, it is supported by circular rings with bearings that allow the scope to freely rotate when the trainee applies torque. The end of the endoscope extends several inches beyond the final circular attachment, and it can move freely in space in response to the up/down and left/right controls on the endoscope. The final steps in ERCP are performed by advancing a catheter through the instrument channel in the endoscope, and positioning the catheter by adjusting an "elevator" control on the scope's handle. Small modifications were made to the Olympus scope to support the catheter movements and the control of the elevator. Instead of passing through the central instrument channel, the catheter is rerouted through a tube to a linear transducer that tracks how far the catheter has been advanced. Similar tubing connects a guide wire attached to the elevator control with a linear transducer measuring the elevator control position. These small modifications to the Olympus scope do not noticeably alter the weight or feel of the endoscope controls.
The ERCP simulation interface also includes a programmable dial and button box that allows the user to select from a set of display and training options. Training sessions can be recorded and then played back from a different perspective. Buttons can also select from a variety of anatomical models. In the future, this box will be programmed to present catastrophic, unusual, or classic cases that every student would be exposed to. Training sessions could also incorporate credentialing criteria that would measure performance and skill level.
The movable track supporting the simulator's endoscope follows a straight path, but the endoscope's movements are directly related to an anatomically correct, three-dimensional (3D) path through the virtual anatomy. The position and orientation of the scope in the virtual anatomy is determine by how far the endoscope has been advanced and rotated in the simulator interface. These motions are tracked by a Polhemus FastTrack system that is attached to the end of the endoscope (Figure 2) and transferred to the simulation computer which updates the display appropriately.
Figure 2. Simulation interface with position tracking on the end of the endoscope.
If the endoscope is advanced five centimeters into the mouth of the simulator, the virtual endoscope on the display advances five centimeters along the 3D gastrointestinal path in the virtual anatomy. The Polhemus is an off-the-shelf product that tracks and reports the position and orientation of a small sensor as it moves through a calibrated magnetic field. In the ERCP simulator, the Polhemus sensor is attached to the end of the endoscope and the magnetic field source is fixed near the moving track. Because metallic components can interfere with the magnetic tracking system, the ERCP simulation interface is primarily constructed from non-metallic components that will not distort the magnetic field. The Polhemus sensor on the endoscope tracks the movement of the endoscope through the virtual anatomy, but the catheter motions relative to the endoscope position are tracked separately by the linear transducers mentioned earlier. Through these tracking systems, the three-dimensional position and orientation of the endoscope and the catheter relative to the virtual anatomy are accurately represented in the simulation.
As the endoscope is maneuvered into position and the catheter is advanced during an actual ERCP procedure, the physician experiences a variety of force feedback information that contributes to his decision making process. The GIT/MCG simulation approximates force feedback in several ways. First, as the endoscope advances through the virtual anatomy, the up/down and left/right movements are constrained by the diameter of the current passage way. For example, through the esophagus, the endoscope can not be easily turned up/down or left/right because the endoscope is squeezing through this relatively narrow passageway. However, in the stomach, there is little or no resistance to the up/down and left/right controls as the endoscope moves freely in the cavity of the stomach. In the simulator, the force feedback associated with this restricted movement is approximated by reducing the range of up/down and left/right motion. Specifically, a flexible line connected to the end of the endoscope is pulled tight by a servo motor when the endoscope is moving through a tight passageway in the virtual anatomy, thus restricting its instantaneous radius of movement. The servo motor allows more slack in the line as the endoscope moves through less narrow passageways in the anatomy, and the line does not restrict the up/down and left/right movements at all as the endoscope passes through the stomach.
Force feedback is also produced as the catheter is advanced toward the major papilla. During ERCP, the catheter can be guided into the bile or pancreatic ducts if the catheter is positioned correctly before advancing toward the papilla. However, if the catheter misses the mark, it will distort the papilla and push back against the endoscope. The GIT/MCG simulator tracks the position of the catheter in relation to the surrounding anatomy, and if the catheter is positioned correctly, it will advance through the papilla into one of the ducts. If the catheter is advanced and it misses the papilla, the simulator produces surface deformation in the neighborhood of the papilla, and a compliant force feedback is applied to the end of the catheter by a servo motor in the simulation interface.
While force feedback is present and helpful, the physician relies primarily on visual feedback while navigating through the ERCP procedure. The GIT/MCG simulation provides computer generated visual feedback from the endoscope using texture mapped three-dimensional computer models that can be displayed and manipulated in real time. The visual simulation presents the correct endoscopic perspective and field of view. The simulated view of the three-dimensional anatomy transitions naturally in real time as the endoscope is guided into position, and the catheter is directed toward the papilla. The visual simulation is accomplished by moving a three-dimensional viewing window through the virtual anatomy in response to the instrument tracking information reported to the simulation computer. It is a real-time, interactive simulation that relies on rapid updates from the instrument tracking system and corresponding responses in the graphics display. The correct three-dimensional viewing window (defined by the endoscope's optical parameters) and rapid display updates are achieved through GL subroutine calls to the graphics subsystem on a high performance Silicon Graphics Onyx computer.
The virtual components of the visual simulation include three-dimensional computer models of the endoscope, the cannula, the endoscopic light (viewing frustum), and anatomical models of the gastrointestinal tract, gall bladder, pancreas, and the associated ducts merging into the major papilla. The standard display mode presents the endoscopic view, but a second viewing option presents an external view of the endoscope in relation to the surrounding anatomy. The second view, not available in real life, is presented to help training physicians to better understand the 3D geometry and positioning maneuvers. In the standard display mode, the only portions of the anatomy that are seen are the interior gastrointestinal tract, and the only instruments that are displayed are those that are advanced through the operating channel (such as the catheter). In the second, external view, the endoscope is displayed and the trainee can selectively display the surrounding anatomy with opaque or semitransparent properties. Another option in the external view is to display a light cone representing the current viewing direction. Both views as illustrated in Figure 3 can be displayed simultaneously on the computer console, and a button on the
Figure 3. Internal and external view from GIT/MCG endoscopic simulation.
simulation interface provides a toggle to switch between these views on the video monitor beside the patient interface.
A final step in ERCP is selective cannulation of the bile and pancreatic ducts. Just as in practice, this is controlled in the simulation by advancing a catheter into the endoscope and adjusting the elevator. Movements of the catheter and the elevator are measured by tracking components in the simulation interface and passed on to the simulation computer which continuously updates the display of the virtual catheter interacting with the virtual anatomy. The stomach, duodenum, and papilla are represented by three-dimensional computer models that are texture mapped with photographic images of the anatomy acquired during endoscopic examinations. The major duodenal papilla is represented by a compliant deformable polygonal surface that deforms as the catheter probes the anatomy in the vicinity of the papilla. Upon successful cannulation, the simulator changes the color of the virtual catheter to represent which duct has been entered; green for the bile duct and blue for the pancreatic duct.
It is the interactive three-dimensional aspect that best differentiates the GIT/MCG medical simulation approach from other endoscopic simulation efforts. Interactive simulation allows the user to manipulate three-dimensional computer models and see the response in real-time. Pushing or pulling the models produces immediate and appropriate model deformation. Furthermore, the GIT/MCG prototype includes force feedback that is important to provide realistic skills training. Other groups have produced endoscopic simulations based upon much less expensive but also less powerful computer platforms. The computer limitations have forced these other groups to rely on simplified graphics and a video-disk with two-dimensional graphic overlays to simulate endoscopic manipulation. These approaches provide useful examples of procedural steps, but they do not simulate the full three-dimensional aspect of the procedure. Computer cost and performance trends continue to move in the right direction for supporting increasingly affordable and realistic medical simulation. At the same time there is growing acceptance within the medical community for computer assisted training. The next important step in medical simulation is to convincingly demonstrate that this technology can reduce the learning curve in a clinical training program.
This research has been funded by a grant from the Georgia Tech / Medical College of Georgia Biomedical Research and Education Program.
For more information contact:
John W. Peifer - john.peifer@bitc.gatech.edu
Michael J. Sinclair -
michael.sinclair@oip.gatech.edu
Dr. W. David Curtis -
WCurtis.DEPTMED@mail.mcg.edu