- Email: [email protected]
A surgical simulator for planning and performing repair of cleft lips
ARTICLE IN PRESS Journal of Cranio-Maxillofacial Surgery (2005) 33, 223–228 r 2005 European Association for Cranio-Maxillofacial Surgery doi:10.1016/j.jcms.2005.05.002, available online at http://www.sciencedirect.com
A surgical simulator for planning and performing repair of cleft lips Stephen SCHENDEL1,2, Kevin MONTGOMERY1, Andrea SOROKIN1, Giancarlo LIONETTI1 National Biocomputation Center, Stanford University, USA; 2Division of Plastic and Reconstructive Surgery, Stanford University, USA 1
The objective of this project was to develop a computer-based surgical simulation system for planning and performing cleft lip repair. This system allows the user to interact with a virtual patient to perform the traditional steps of cleft-lip repair (rotation-advancement technique). Materials and methods: The system interfaces to force-feedback (haptic) devices to track the user’s motion and provide feedback during the procedure, while performing real-time soft-tissue simulation. An 11-day-old unilateral cleft lip, alveolus and palate patient was previously CT scanned for ancillary diagnostic purposes using standard imaging protocols and 1 mm slices. Highresolution 3D meshes were automatically generated from this data using the ROVE software developed in-house. The resulting 3D meshes of bone and soft tissue were instilled with physical properties of soft tissues for purposes of simulation. Once these preprocessing steps were completed, the patient’s bone and soft tissue data are presented on the computer screen in stereo and the user can freely view, rotate, and otherwise interact with the patient’s data in real time. The user is prompted to select anatomical landmarks on the patient’s data for preoperative planning purposes, then their locations are compared against that of a ‘gold standard’ and a score, derived from their deviation from that standard and time required, is generated. The user can then move a haptic stylus and guide the motion of the virtual cutting tool. The soft tissues can thus be incised using this virtual cutting tool, moved using virtual forceps, and fused in order to perform any of the major procedures for cleft lip repair. Real-time soft tissue deformation of the mesh realistically simulates normal tissues and haptic-rate ð41 kHzÞ force-feedback is provided. The surgical result of the procedure can then be immediately visualized and the entire training process can be repeated at will. A short evaluation study was also performed. Two groups (non-medical and plastic surgery residents) of six persons each performed the anatomical marking task of the simulator four times. Results: Results showed that the plastic surgery residents scored consistently better than the persons without medical background. Every person’s score increased with practice, and the length of time needed to complete the 11 markings decreased. The data was compiled and showed which specific markers consistently took users the longest to identify as well as which locations were hardest to accurately mark. Conclusion: These findings suggest that the simulator is a valuable training tool, giving residents a way to practice anatomical identification for cleft lip surgery without the risks associated with training on a live patient. Educators can also use the simulator to examine which markers are consistently problematic, and modify their training to address these needs. r 2005 European Association for Cranio-Maxillofacial Surgery
SUMMARY.
Keywords: cleft lip; surgical simulation; soft-tissue deformation; haptics; education
The use of live animals though in surgical training has raised ethical concerns and is banned in some countries. Similarly, there exist availability problems and other issues with the use of human cadavers. Bench models have proven useful in the teaching and assessment of general surgical skills such as suturing and performing simple Z-plasties. However, they have not included the development of surgical decision making and the procurement of specific complex surgical skills. An example of this is the repair of a cleft lip, which is a complex procedure incorporating a complex rotation advancement flap and the identification of precise landmarks and advanced decision making (Anastakis et al., 1999). Computer simulation (Pieper et al., 1995; Delingette, 1998; Haasfeld et al., 1998; Cotin et al., 2000; Montgomery et al., 2001; Cutting, 2005) and the creation of virtual reality technologies (Zonneveld
INTRODUCTION The acquisition of surgical skills and proficiency is a sequential process. Competency requires the appropriate decision making based on knowledge of anatomy, pathology and surgical technique combined with technical proficiency. Technical proficiency is comprised of efficient time motion, tissue handling, knowledge of procedure, visual spatial ability and dexterity. Some of the skills are general, others are very specific. Traditionally under the apprenticeship model, much of this learning has occurred in the operating room under the supervision of an attending surgeon which has obvious limitations. Recently, successful teaching and assessment of technical skills outside the operating room have included human cadavers, live animals and bench models (Anastakis et al., 1999). 223
ARTICLE IN PRESS 224 Journal of Cranio-Maxillofacial Surgery
et al., 1989; O’Toole et al., 1999; Montgomery et al., 2000; Brown et al., 2001) has allowed this type of learning to occur in other areas such as the aviation and military arenas. In surgery, computer-based simulation holds the promise of easily providing different training scenarios (anatomical variations, pathologies, operating conditions), objective quantification of surgical performance, and acceleration of baseline skills, all without the risk to real patient outcomes. Endoscopic procedures have proven particularly amenable to this type of learning and assessment (Lamoral et al., 1990; Pokrandt et al., 1996; Bholat et al., 1999; Gorman et al., 1999; O’Toole et al., 1999). A computer simulation program was created designed to assess and teach the repair of cleft lip. The program is based on an actual case derived from patient-specific data. The student is allowed to choose the correct landmarks needed to perform the procedure. The three-dimensional model can be rotated at will in order to choose the correct points. A score is then immediately given based on the proximity of the chosen points to the actual points and the time taken to choose the points. The tissue can then be ‘cut’ by either using the mouse or a force feedback device. Virtual tools have been created that then allow pulling the tissue to the desired position.
Fig. 1 – Visualization of skin surface.
triangles using a quadratic slimming algorithm (Garland and Heckbert, 1997). The 3D meshes of the bone and soft tissues were instilled with physical properties for purposes of simulation. Additional mesh layers were added between the skin and bone meshes in order to simulate the volume of enclosed tissue. The skin stiffness parameters were derived from previous studies of the physical properties of soft tissues (Fig. 1).
METHODS Interaction Acquisition The patient was an 11-day-old male with a complete left unilateral cleft lip, alveolus and palate who has not undergone any previous surgery. The patient was scanned using standard imaging protocols on a GE LightSpeed helical-scan CT with 1 mm slice thickness and a pixel spacing of 0.3 mm. The image data were stored as DICOM images then transferred to a DICOM server in the Biocomputation Center. Once there, an automated process was begun to generate a three-dimensional computer model of the patient’s bone and soft tissue directly from the imaging data. Segmentation, mesh generation, and artefact removal were performed automatically using in-house software ROVE (Reconstruction of Volumetric Elements) in approximately 15 min. The resulting three-dimensional mesh of the patient’s skull comprised over 608,000 vertices with 1.2 million triangles and the face comprised over 350,000 thousand vertices with 700,000 triangles. Visualization Preoperative images using these ultra high-resolution meshes were generated (A/P, lateral, 2/3 frontal) of each object independently and also together, with the skin displayed semitransparent over the skull. For subsequent interaction and simulation, each mesh was reduced to roughly 50,000 vertices and 100,000
Once these preprocessing steps were completed, the patient’s bone and soft-tissue data were presented on the computer screen (Fig. 2) and the user can freely rotate and interact with the patient’s data in real time. The user is prompted to select 11 different anatomical landmarks on the skin surface (Fig. 3) and the system compares the location of their markers with that of a previously stored ‘gold standard’ derived from experts to produce a score based on the deviation between these markers and elapsed time, based on the following formula: ð100n ð4 average deviationÞÞ þ ð100 time to completeÞ so that both accuracy (primarily) and rapidity (secondarily) factor into the equation. An average deviation greater than four millimeters or at time over 100 s was judged unacceptable. After the marker placement task, the user could then move a haptic stylus (Phantom, Sensable Technologies) in order to guide the motion of a virtual cutting tool (Fig. 4). The soft tissues could be incised and moved in order to perform any of the major procedures for cleft lip repair (Fig. 5). During the procedure, the system performed collision detection (Quinlan, 1994; Sorkin, 2000) and calculated the forces that the user should feel while rendering those forces through the haptic device to provide forcefeedback to the user. Finally, the user could view the
ARTICLE IN PRESS A surgical simulator for cleft lip planning and repair 225
Fig. 5 – View of incisions performed.
Fig. 2 – Skin semitransparent over bone/cartilage.
Fig. 3 – Evaluation of marker placement.
with integrated collision detection and resolution that interfaces to a number of force-feedback (haptic) and non-haptic input/output devices. This system models an object as a collection of point masses connected by linear springs in a three-dimensional mesh structure. The behaviour of each tissue is modelled by modulating these stiffness coefficients and additional springs are placed between adjacent internal organs in order to propagate the effects of grasping connected components. Bone is modelled as a rigid object that can also be manipulated and also provides constraints on the soft-tissue geometry in space. Solution of the deformation equations is performed using a localized semi-static solver (a simplification of the traditional Euler method that ignores inertial and damping forces), which provides a significant increase in performance. In order to speed up the simulation further, it was chosen to solve the deformation equations using parallel processing through a multithreaded implementation on a multi-processor Sun Microsystems (Menlo Park, CA) E3500 (8 400 MHz UltraSparc) workstation. The simulation system is written in C þ þ using the OpenGL, GLUT [Robins], and GLUI [Rademacher] libraries for visualization and user interface. Crossplatform and multithreading capabilities are provided via the POSIX libraries.
Evaluation Fig. 4 – Real-time haptic feedback during cutting.
resulting outcome and choose whether to repeat the procedure. Simulation In order to accomplish the procedures listed above, a real-time, soft-tissue modelling engine was developed
It was important to quantitatively determine the validity and utility of the simulator for medical education. Six persons without any previous medical training, as well as six plastic surgery residents (Fig. 6), used the simulator. Both groups were given a brief lecture describing how to properly mark the lip of a cleft lip patient prior to surgery and were briefly introduced to the surgical simulator. Each person ran through the program four times, and it took each person approximately 30 min to do so.
ARTICLE IN PRESS 226 Journal of Cranio-Maxillofacial Surgery
400
Layperson Resident
350 Average Score
300 250 200 150 100 50 0 1
1 2 3 4 5 6
RESULTS
DISCUSSION From the data obtained, the surgical simulator appears to be a valid model of a unilateral cleft lip patient. The residents scored higher than the laypeople after receiving the same instruction, which
3 Trial Number
4
5
Fig. 7 – Average score vs. trial number.
Fig. 6
- Normal Nasal Dome - Cleft Nasal Dome - Philtral Base - Normal Side - Philtral Base - Cleft Side - Prolongation Line 3-4 to Vermilion - Alar Base - Cleft Side
7 - Cupid's Bow - Normal Side 8 - Middle Cupid's Bow 9 - Cleft Cupid's Bow 10 - Prolongation 6 to Vermilion 11 - End of White Roll - Cleft Side
12 10 Average Time (seconds)
After compiling the data, the average scores of the untrained laypeople and the residents were compared. The average simulator score of the six laypeople was 151, and the residents received an average score of 296. The residents’ average score was higher, which is what would be expected if the simulator represented a realistic model. The residents should score the best since they are currently in, or have just completed, the cleft lip and palate portion of their surgical training. Although the persons without medical training received instruction on how to correctly place the anatomical markings, they have never been trained in this area before and have no practical experience with this marking system. Therefore, it is logical that the residents performed better, since they comprise the group with the most recent exposure to the tasks tested by the simulator (Fig. 7). The residents’ scores were higher for each trial. Both groups improved as they performed more trials. Thus, the simulator increased the user’s score with practice. The residents more rapidly approached the highest score attainable on the simulator. The amount of time it took both groups decreased overall as they practiced with the simulator. The utility of the simulator as an educational tool was also examined. For each marker, the residents’ scores were averaged and recorded graphically (Fig. 8). This presented the data so that an instructor could quickly and easily see which markers were hardest for the residents to accurately identify.
2
8 6 4 2 0 Marker Number Fig. 8 – Residents’ average time vs. marker number.
implies that the residents were able to apply their recent experiences to the simulator. Both groups increased their accuracy and decreased their speed after practicing on the simulator. The simulator also provided a simplistic way to evaluate groups of people and examine trends among them. While performing the study, the residents approached the simulator differently than the laypersons. After running through the simulation once and receiving feedback, the residents seemed to relate this information back to their anatomical knowledge so they could understand where they deviated from the gold standard. The laypersons, however, simply tried to memorize where the gold standards were. Both the residents’ as well as the laypeople’s scores improved, although only the residents seemed to gain a better
ARTICLE IN PRESS A surgical simulator for cleft lip planning and repair 227
and objective evaluation without risk to real outcomes. This system for cleft lip repair starts with automatically reconstructed patient-specific data, allows for preoperative planning and visualization, and allows the user to perform the major steps of cleft lip repair. It provides objective metrics on performance and enables repeated trials until a sufficient level of proficiency is obtained. Future work will be aimed toward validating the teaching component of landmark identification and simplifying the virtual surgical correction. This is an important first step in the creation of cleft lip simulators for teaching based on patient-specific data.
References Fig. 9 – Closure of cleft and resulting outcome.
understanding of the correct placement of the anatomical markers. The laypersons often commented that they were purely memorizing the data, while the residents discussed the intricacies of the model (Fig. 9). Therefore, when this simulator is used for actual training purposes, it will be important to add additional models so that the users cannot simply memorize the gold standard locations. These additional models will also allow users to experience many cases and gain exposure to different anatomical variations. It will give surgeons a way to practice some of the intricacies of their upcoming operations. The surgical simulator will provide a way to continually challenge its users and force them to apply their anatomical knowledge which will ultimately benefit patient care. It was also evident from this study that not all landmarks in cleft-lip repair have the same significance to the surgeon. Several points took longer to identify. These were in order: the prolongation line from the philtral bases to the vermilion, prolongation of the cleft alar base to the vermilion and cupid’s bow on the cleft side. This last mark, cupid’s bow on the cleft side is probably the most significant mark for an acceptable repair and it is not unusual that one would take more time in choosing this point than others. The other two points, however, are prolongation points which require one to visualize the projection of a point or line to another anatomical structure. This type of visualization is more complex than just identifying a structure and thus it is apparent from here that this task thus requires more time in the analytical decision making process. This type of visualization seems to vary amenably to computer simulation learning.
CONCLUSIONS Computer-based surgical simulation holds the promise for broader training, repetitive skills acquisition,
Anastakis DJ, Regehr G, Reznick RK, et al.: Assessment of technical skills transfer from the bench training model to the human model. Amer J Surg 177: 167, 1999 Bholat O, Haluck R, Kutz R, Gorman P, Krummel T: Defining the role of haptic feedback in minimally invasive surgery. In: Medicine Meets Virtual Reality, 1999; pp. 62–66 Brown J, Montgomery K, Latombe J-C, Stephanides M: A microsurgery simulation system. In: Fourth International Conference on Medical Image Computing and ComputerAssisted Intervention (MICCAI 2001), October 2001 Cotin S, Delingette H, Ayache N: A hybrid elastic model allowing real-time cutting, deformations and forcefeedback for surgery training and simulation. Visual Comput 16: 437–452, 2000 Cutting C: A deformer-based surgical simulation program for cleft lip and palate surgery. In: Medicine Meets Virtual Reality (MMVR05), Long Beach, CA, January 2005 Delingette H: Towards realistic soft tissue modeling in medical simulation. In: Proceedings of the IEEE: Special Issue on Surgery Simulation, April 1998, pp. 512–523 Garland M, Heckbert P: Surface simplification using quadratic error metrics. In: ACM SIGGRAPH 97 Conference Proceedings, 1997, pp. 43–52 Gorman PJ, Lieser BS, Murray WB, Haluck RS, Krummel TM: Evaluation of skill acquisition using a force-feedback virtual reality-based surgical trainer. In: Medicine Meets Virtual Reality (MMVR). Newport Beach, CA: IOC Press, 1999 Hassfeld S, Zo¨ller J, Albert FK, Wirtz CR, Knauth M, Mu¨hling J: Preoperative planning and intraoperative navigation in skull base surgery. J Cranio-Maxillofac Surg 26: 220–225, 1998 Lamoral Y, Quirynen M, Peene P, Vanneste F, Lemahieu SF, Baert AL, van Steenberghe D: Computer tomography in the preoperative planning of oral endo-osseous implant surgery. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 153: 505–509, 1990 Montgomery K, Stephanides M, Schendel S: Development and application of a virtual environment for reconstructive surgery. J Computer-Aided Surg 5: 90–97, 2000 ISSN:1092-9088 Montgomery K, Heinrichs L, Bruyns C, Wildermuth S, Hasser C, Ozenne S, Bailey D: Surgical simulator for operative hysteroscopy and endometrial ablation. Computer-Aided Radiology and Surgery (CARS 2001), Berlin, Germany, June 27, 2001 O’Toole R, Playter R, Krummel T, Blank WC, Cornelius NH, Roberts WR, Bell WJ, Raibert M: Measuring and developing suturing technique with a virtual reality surgical simulator. J Amer Coll Surg 189: 114–127, 1999 Pieper S, Laub Jr D, Rosen J: A finite element facial model for simulating plastic surgery. Plastic Reconstr Surg 96: 1100–1105, 1995 Pokrandt P, Bohner P, Knolle JP, Hassfeld S, Raczkowsky J: A communication system supporting simultaneous planning and
ARTICLE IN PRESS 228 Journal of Cranio-Maxillofacial Surgery execution in cranio-maxillo-facial surgery. J Telemed Telecare 2: 3–7, 1996 Quinlan S: Efficient distance computation between nonconvex objects. In: Proceedings of the IEEE International Conference on Robotics and Automation, 1994, pp. 3324–3329 Sorkin S: Distance computation between deformable objects. Honors Thesis, Computer Science Department, Stanford University, June 2000 Zonneveld FW, Lobregt S, van der Meulen JCH, Vaandrager JM: Three-dimensional imaging in craniofacial surgery. World J Surg 13: 328–342, 1989
Dr. S. SCHENDEL National Biocomputation Center Stanford University Stanford, CA 94305 USA E-mail: [email protected]
Paper received 18 November 2003 Accepted 26 April 2005
A surgical simulator for planning and performing repair of cleft lips Stephen SCHENDEL1,2, Kevin MONTGOMERY1, Andrea SOROKIN1, Giancarlo LIONETTI1 National Biocomputation Center, Stanford University, USA; 2Division of Plastic and Reconstructive Surgery, Stanford University, USA 1
The objective of this project was to develop a computer-based surgical simulation system for planning and performing cleft lip repair. This system allows the user to interact with a virtual patient to perform the traditional steps of cleft-lip repair (rotation-advancement technique). Materials and methods: The system interfaces to force-feedback (haptic) devices to track the user’s motion and provide feedback during the procedure, while performing real-time soft-tissue simulation. An 11-day-old unilateral cleft lip, alveolus and palate patient was previously CT scanned for ancillary diagnostic purposes using standard imaging protocols and 1 mm slices. Highresolution 3D meshes were automatically generated from this data using the ROVE software developed in-house. The resulting 3D meshes of bone and soft tissue were instilled with physical properties of soft tissues for purposes of simulation. Once these preprocessing steps were completed, the patient’s bone and soft tissue data are presented on the computer screen in stereo and the user can freely view, rotate, and otherwise interact with the patient’s data in real time. The user is prompted to select anatomical landmarks on the patient’s data for preoperative planning purposes, then their locations are compared against that of a ‘gold standard’ and a score, derived from their deviation from that standard and time required, is generated. The user can then move a haptic stylus and guide the motion of the virtual cutting tool. The soft tissues can thus be incised using this virtual cutting tool, moved using virtual forceps, and fused in order to perform any of the major procedures for cleft lip repair. Real-time soft tissue deformation of the mesh realistically simulates normal tissues and haptic-rate ð41 kHzÞ force-feedback is provided. The surgical result of the procedure can then be immediately visualized and the entire training process can be repeated at will. A short evaluation study was also performed. Two groups (non-medical and plastic surgery residents) of six persons each performed the anatomical marking task of the simulator four times. Results: Results showed that the plastic surgery residents scored consistently better than the persons without medical background. Every person’s score increased with practice, and the length of time needed to complete the 11 markings decreased. The data was compiled and showed which specific markers consistently took users the longest to identify as well as which locations were hardest to accurately mark. Conclusion: These findings suggest that the simulator is a valuable training tool, giving residents a way to practice anatomical identification for cleft lip surgery without the risks associated with training on a live patient. Educators can also use the simulator to examine which markers are consistently problematic, and modify their training to address these needs. r 2005 European Association for Cranio-Maxillofacial Surgery
SUMMARY.
Keywords: cleft lip; surgical simulation; soft-tissue deformation; haptics; education
The use of live animals though in surgical training has raised ethical concerns and is banned in some countries. Similarly, there exist availability problems and other issues with the use of human cadavers. Bench models have proven useful in the teaching and assessment of general surgical skills such as suturing and performing simple Z-plasties. However, they have not included the development of surgical decision making and the procurement of specific complex surgical skills. An example of this is the repair of a cleft lip, which is a complex procedure incorporating a complex rotation advancement flap and the identification of precise landmarks and advanced decision making (Anastakis et al., 1999). Computer simulation (Pieper et al., 1995; Delingette, 1998; Haasfeld et al., 1998; Cotin et al., 2000; Montgomery et al., 2001; Cutting, 2005) and the creation of virtual reality technologies (Zonneveld
INTRODUCTION The acquisition of surgical skills and proficiency is a sequential process. Competency requires the appropriate decision making based on knowledge of anatomy, pathology and surgical technique combined with technical proficiency. Technical proficiency is comprised of efficient time motion, tissue handling, knowledge of procedure, visual spatial ability and dexterity. Some of the skills are general, others are very specific. Traditionally under the apprenticeship model, much of this learning has occurred in the operating room under the supervision of an attending surgeon which has obvious limitations. Recently, successful teaching and assessment of technical skills outside the operating room have included human cadavers, live animals and bench models (Anastakis et al., 1999). 223
ARTICLE IN PRESS 224 Journal of Cranio-Maxillofacial Surgery
et al., 1989; O’Toole et al., 1999; Montgomery et al., 2000; Brown et al., 2001) has allowed this type of learning to occur in other areas such as the aviation and military arenas. In surgery, computer-based simulation holds the promise of easily providing different training scenarios (anatomical variations, pathologies, operating conditions), objective quantification of surgical performance, and acceleration of baseline skills, all without the risk to real patient outcomes. Endoscopic procedures have proven particularly amenable to this type of learning and assessment (Lamoral et al., 1990; Pokrandt et al., 1996; Bholat et al., 1999; Gorman et al., 1999; O’Toole et al., 1999). A computer simulation program was created designed to assess and teach the repair of cleft lip. The program is based on an actual case derived from patient-specific data. The student is allowed to choose the correct landmarks needed to perform the procedure. The three-dimensional model can be rotated at will in order to choose the correct points. A score is then immediately given based on the proximity of the chosen points to the actual points and the time taken to choose the points. The tissue can then be ‘cut’ by either using the mouse or a force feedback device. Virtual tools have been created that then allow pulling the tissue to the desired position.
Fig. 1 – Visualization of skin surface.
triangles using a quadratic slimming algorithm (Garland and Heckbert, 1997). The 3D meshes of the bone and soft tissues were instilled with physical properties for purposes of simulation. Additional mesh layers were added between the skin and bone meshes in order to simulate the volume of enclosed tissue. The skin stiffness parameters were derived from previous studies of the physical properties of soft tissues (Fig. 1).
METHODS Interaction Acquisition The patient was an 11-day-old male with a complete left unilateral cleft lip, alveolus and palate who has not undergone any previous surgery. The patient was scanned using standard imaging protocols on a GE LightSpeed helical-scan CT with 1 mm slice thickness and a pixel spacing of 0.3 mm. The image data were stored as DICOM images then transferred to a DICOM server in the Biocomputation Center. Once there, an automated process was begun to generate a three-dimensional computer model of the patient’s bone and soft tissue directly from the imaging data. Segmentation, mesh generation, and artefact removal were performed automatically using in-house software ROVE (Reconstruction of Volumetric Elements) in approximately 15 min. The resulting three-dimensional mesh of the patient’s skull comprised over 608,000 vertices with 1.2 million triangles and the face comprised over 350,000 thousand vertices with 700,000 triangles. Visualization Preoperative images using these ultra high-resolution meshes were generated (A/P, lateral, 2/3 frontal) of each object independently and also together, with the skin displayed semitransparent over the skull. For subsequent interaction and simulation, each mesh was reduced to roughly 50,000 vertices and 100,000
Once these preprocessing steps were completed, the patient’s bone and soft-tissue data were presented on the computer screen (Fig. 2) and the user can freely rotate and interact with the patient’s data in real time. The user is prompted to select 11 different anatomical landmarks on the skin surface (Fig. 3) and the system compares the location of their markers with that of a previously stored ‘gold standard’ derived from experts to produce a score based on the deviation between these markers and elapsed time, based on the following formula: ð100n ð4 average deviationÞÞ þ ð100 time to completeÞ so that both accuracy (primarily) and rapidity (secondarily) factor into the equation. An average deviation greater than four millimeters or at time over 100 s was judged unacceptable. After the marker placement task, the user could then move a haptic stylus (Phantom, Sensable Technologies) in order to guide the motion of a virtual cutting tool (Fig. 4). The soft tissues could be incised and moved in order to perform any of the major procedures for cleft lip repair (Fig. 5). During the procedure, the system performed collision detection (Quinlan, 1994; Sorkin, 2000) and calculated the forces that the user should feel while rendering those forces through the haptic device to provide forcefeedback to the user. Finally, the user could view the
ARTICLE IN PRESS A surgical simulator for cleft lip planning and repair 225
Fig. 5 – View of incisions performed.
Fig. 2 – Skin semitransparent over bone/cartilage.
Fig. 3 – Evaluation of marker placement.
with integrated collision detection and resolution that interfaces to a number of force-feedback (haptic) and non-haptic input/output devices. This system models an object as a collection of point masses connected by linear springs in a three-dimensional mesh structure. The behaviour of each tissue is modelled by modulating these stiffness coefficients and additional springs are placed between adjacent internal organs in order to propagate the effects of grasping connected components. Bone is modelled as a rigid object that can also be manipulated and also provides constraints on the soft-tissue geometry in space. Solution of the deformation equations is performed using a localized semi-static solver (a simplification of the traditional Euler method that ignores inertial and damping forces), which provides a significant increase in performance. In order to speed up the simulation further, it was chosen to solve the deformation equations using parallel processing through a multithreaded implementation on a multi-processor Sun Microsystems (Menlo Park, CA) E3500 (8 400 MHz UltraSparc) workstation. The simulation system is written in C þ þ using the OpenGL, GLUT [Robins], and GLUI [Rademacher] libraries for visualization and user interface. Crossplatform and multithreading capabilities are provided via the POSIX libraries.
Evaluation Fig. 4 – Real-time haptic feedback during cutting.
resulting outcome and choose whether to repeat the procedure. Simulation In order to accomplish the procedures listed above, a real-time, soft-tissue modelling engine was developed
It was important to quantitatively determine the validity and utility of the simulator for medical education. Six persons without any previous medical training, as well as six plastic surgery residents (Fig. 6), used the simulator. Both groups were given a brief lecture describing how to properly mark the lip of a cleft lip patient prior to surgery and were briefly introduced to the surgical simulator. Each person ran through the program four times, and it took each person approximately 30 min to do so.
ARTICLE IN PRESS 226 Journal of Cranio-Maxillofacial Surgery
400
Layperson Resident
350 Average Score
300 250 200 150 100 50 0 1
1 2 3 4 5 6
RESULTS
DISCUSSION From the data obtained, the surgical simulator appears to be a valid model of a unilateral cleft lip patient. The residents scored higher than the laypeople after receiving the same instruction, which
3 Trial Number
4
5
Fig. 7 – Average score vs. trial number.
Fig. 6
- Normal Nasal Dome - Cleft Nasal Dome - Philtral Base - Normal Side - Philtral Base - Cleft Side - Prolongation Line 3-4 to Vermilion - Alar Base - Cleft Side
7 - Cupid's Bow - Normal Side 8 - Middle Cupid's Bow 9 - Cleft Cupid's Bow 10 - Prolongation 6 to Vermilion 11 - End of White Roll - Cleft Side
12 10 Average Time (seconds)
After compiling the data, the average scores of the untrained laypeople and the residents were compared. The average simulator score of the six laypeople was 151, and the residents received an average score of 296. The residents’ average score was higher, which is what would be expected if the simulator represented a realistic model. The residents should score the best since they are currently in, or have just completed, the cleft lip and palate portion of their surgical training. Although the persons without medical training received instruction on how to correctly place the anatomical markings, they have never been trained in this area before and have no practical experience with this marking system. Therefore, it is logical that the residents performed better, since they comprise the group with the most recent exposure to the tasks tested by the simulator (Fig. 7). The residents’ scores were higher for each trial. Both groups improved as they performed more trials. Thus, the simulator increased the user’s score with practice. The residents more rapidly approached the highest score attainable on the simulator. The amount of time it took both groups decreased overall as they practiced with the simulator. The utility of the simulator as an educational tool was also examined. For each marker, the residents’ scores were averaged and recorded graphically (Fig. 8). This presented the data so that an instructor could quickly and easily see which markers were hardest for the residents to accurately identify.
2
8 6 4 2 0 Marker Number Fig. 8 – Residents’ average time vs. marker number.
implies that the residents were able to apply their recent experiences to the simulator. Both groups increased their accuracy and decreased their speed after practicing on the simulator. The simulator also provided a simplistic way to evaluate groups of people and examine trends among them. While performing the study, the residents approached the simulator differently than the laypersons. After running through the simulation once and receiving feedback, the residents seemed to relate this information back to their anatomical knowledge so they could understand where they deviated from the gold standard. The laypersons, however, simply tried to memorize where the gold standards were. Both the residents’ as well as the laypeople’s scores improved, although only the residents seemed to gain a better
ARTICLE IN PRESS A surgical simulator for cleft lip planning and repair 227
and objective evaluation without risk to real outcomes. This system for cleft lip repair starts with automatically reconstructed patient-specific data, allows for preoperative planning and visualization, and allows the user to perform the major steps of cleft lip repair. It provides objective metrics on performance and enables repeated trials until a sufficient level of proficiency is obtained. Future work will be aimed toward validating the teaching component of landmark identification and simplifying the virtual surgical correction. This is an important first step in the creation of cleft lip simulators for teaching based on patient-specific data.
References Fig. 9 – Closure of cleft and resulting outcome.
understanding of the correct placement of the anatomical markers. The laypersons often commented that they were purely memorizing the data, while the residents discussed the intricacies of the model (Fig. 9). Therefore, when this simulator is used for actual training purposes, it will be important to add additional models so that the users cannot simply memorize the gold standard locations. These additional models will also allow users to experience many cases and gain exposure to different anatomical variations. It will give surgeons a way to practice some of the intricacies of their upcoming operations. The surgical simulator will provide a way to continually challenge its users and force them to apply their anatomical knowledge which will ultimately benefit patient care. It was also evident from this study that not all landmarks in cleft-lip repair have the same significance to the surgeon. Several points took longer to identify. These were in order: the prolongation line from the philtral bases to the vermilion, prolongation of the cleft alar base to the vermilion and cupid’s bow on the cleft side. This last mark, cupid’s bow on the cleft side is probably the most significant mark for an acceptable repair and it is not unusual that one would take more time in choosing this point than others. The other two points, however, are prolongation points which require one to visualize the projection of a point or line to another anatomical structure. This type of visualization is more complex than just identifying a structure and thus it is apparent from here that this task thus requires more time in the analytical decision making process. This type of visualization seems to vary amenably to computer simulation learning.
CONCLUSIONS Computer-based surgical simulation holds the promise for broader training, repetitive skills acquisition,
Anastakis DJ, Regehr G, Reznick RK, et al.: Assessment of technical skills transfer from the bench training model to the human model. Amer J Surg 177: 167, 1999 Bholat O, Haluck R, Kutz R, Gorman P, Krummel T: Defining the role of haptic feedback in minimally invasive surgery. In: Medicine Meets Virtual Reality, 1999; pp. 62–66 Brown J, Montgomery K, Latombe J-C, Stephanides M: A microsurgery simulation system. In: Fourth International Conference on Medical Image Computing and ComputerAssisted Intervention (MICCAI 2001), October 2001 Cotin S, Delingette H, Ayache N: A hybrid elastic model allowing real-time cutting, deformations and forcefeedback for surgery training and simulation. Visual Comput 16: 437–452, 2000 Cutting C: A deformer-based surgical simulation program for cleft lip and palate surgery. In: Medicine Meets Virtual Reality (MMVR05), Long Beach, CA, January 2005 Delingette H: Towards realistic soft tissue modeling in medical simulation. In: Proceedings of the IEEE: Special Issue on Surgery Simulation, April 1998, pp. 512–523 Garland M, Heckbert P: Surface simplification using quadratic error metrics. In: ACM SIGGRAPH 97 Conference Proceedings, 1997, pp. 43–52 Gorman PJ, Lieser BS, Murray WB, Haluck RS, Krummel TM: Evaluation of skill acquisition using a force-feedback virtual reality-based surgical trainer. In: Medicine Meets Virtual Reality (MMVR). Newport Beach, CA: IOC Press, 1999 Hassfeld S, Zo¨ller J, Albert FK, Wirtz CR, Knauth M, Mu¨hling J: Preoperative planning and intraoperative navigation in skull base surgery. J Cranio-Maxillofac Surg 26: 220–225, 1998 Lamoral Y, Quirynen M, Peene P, Vanneste F, Lemahieu SF, Baert AL, van Steenberghe D: Computer tomography in the preoperative planning of oral endo-osseous implant surgery. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 153: 505–509, 1990 Montgomery K, Stephanides M, Schendel S: Development and application of a virtual environment for reconstructive surgery. J Computer-Aided Surg 5: 90–97, 2000 ISSN:1092-9088 Montgomery K, Heinrichs L, Bruyns C, Wildermuth S, Hasser C, Ozenne S, Bailey D: Surgical simulator for operative hysteroscopy and endometrial ablation. Computer-Aided Radiology and Surgery (CARS 2001), Berlin, Germany, June 27, 2001 O’Toole R, Playter R, Krummel T, Blank WC, Cornelius NH, Roberts WR, Bell WJ, Raibert M: Measuring and developing suturing technique with a virtual reality surgical simulator. J Amer Coll Surg 189: 114–127, 1999 Pieper S, Laub Jr D, Rosen J: A finite element facial model for simulating plastic surgery. Plastic Reconstr Surg 96: 1100–1105, 1995 Pokrandt P, Bohner P, Knolle JP, Hassfeld S, Raczkowsky J: A communication system supporting simultaneous planning and
ARTICLE IN PRESS 228 Journal of Cranio-Maxillofacial Surgery execution in cranio-maxillo-facial surgery. J Telemed Telecare 2: 3–7, 1996 Quinlan S: Efficient distance computation between nonconvex objects. In: Proceedings of the IEEE International Conference on Robotics and Automation, 1994, pp. 3324–3329 Sorkin S: Distance computation between deformable objects. Honors Thesis, Computer Science Department, Stanford University, June 2000 Zonneveld FW, Lobregt S, van der Meulen JCH, Vaandrager JM: Three-dimensional imaging in craniofacial surgery. World J Surg 13: 328–342, 1989
Dr. S. SCHENDEL National Biocomputation Center Stanford University Stanford, CA 94305 USA E-mail: [email protected]
Paper received 18 November 2003 Accepted 26 April 2005