- Email: [email protected]
Accuracy of anterior cruciate ligament tunnel placement with an active robotic system: A cadaveric study
Accuracy of Anterior Cruciate Ligament Tunnel Placement With an Active Robotic System: A Cadaveric Study Volker Musahl, M.D., Andreas Burkart, M.D., Richard E. Debski, Ph.D., Andrew Van Scyoc, B.S., Freddie H. Fu, M.D., and Savio L-Y. Woo, Ph.D.
Purpose: The objective of this study was to evaluate the accuracy of tunnel placement for ACL reconstruction performed with an active robotic system. Type of Study: Cadaveric analysis. Methods: A reference screw containing 4 fiducials was placed in the femur and tibia of 13 fresh-frozen cadaveric knees. A preoperative plan was developed using images from 3-dimensional computed tomography reconstructions of the knee. The active robotic system then drilled the tunnels. The location and direction of each planned tunnel in the femur and tibia were determined from the preoperative plan. To compare these parameters postoperatively, a mechanical digitizer and a tunnel plug were used. The deviation in location and direction between the planned and drilled tunnel was determined. Results: In preliminary trials, the tibial tunnel was located inaccurately because slippage of the drill bit occurred on the bone at the start of tunnel drilling. This was minimized by decreasing the feed rate of the robot by 75%. For the remaining 10 knees, deviations with respect to the preoperative plan were found of 2.0 ⫾ 1.2 mm and 1.1° ⫾ 0.7° for the intra-articular tibial tunnel location and direction, respectively. For the femur, the deviations were 1.3 ⫾ 0.9 mm for the tunnel location (intra-articular) and 1.0° ⫾ 0.6° for the tunnel direction. Conclusions: The active robotic system is highly accurate for tunnel placement during ACL reconstruction, meaning that the robot drills the tunnels very close to the surgeon’s plan. Comparison to a control group of surgeons could not be made because no preoperative plan is usually created in traditional surgery. However, accuracy values in this study were found to be below the values for precision of repeated tunnel placements reported in the literature. Key Words: Computer-assisted orthopedic surgery—ACL—Accuracy— Tunnel placement—CASPAR.
O
ne of the proposed advantages of computerassisted surgery in anterior cruciate ligament (ACL) reconstruction is improvement in the accuracy of the surgical procedure compared with traditional methods. Outcomes for ACL reconstruction still do not exceed 90% good or excellent results.1 It is also well known that placement of the tunnels in the femur and tibia influences the outcome of ACL reconstruc-
From the Musculoskeletal Research Center, Department of Orthopaedic Surgery, Pittsburgh, Pennsylvania, U.S.A. Address correspondence and reprint requests to Freddie H. Fu, M.D., Musculoskeletal Research Center, Department of Orthopaedic Surgery, E1641 Bio Medical Tower, 210 Lothrop St, PO Box 71199, Pittsburgh, PA 15213, U.S.A. E-mail: [email protected] © 2002 by the Arthroscopy Association of North America 0749-8063/02/1809-3107$35.00/0 doi:10.1053/jars.2002.36110
968
tion procedures.2,3 Specifically, many studies have extensively examined the tunnel positions and found that inappropriate graft placement interfered with graft incorporation and knee function.4-6 Inappropriate positioning of the tunnels can also result in excessive changes in graft length as the knee moves through the range of motion.7,8 The femoral tunnel is the most commonly misplaced. Because the femoral attachment of the ACL is closer to the rotational center of the knee, small errors can be detrimental to knee function.9 Complications such as graft failure, diminished range of motion, and graft impingement with the femoral notch can result.4,10,11 Despite the great variability in tunnel placement between surgeons, each surgeon attempts to consistently place tunnels in the same location from knee to knee.12,13 Initial reports on computer-assisted orthopedic sur-
Arthroscopy: The Journal of Arthroscopic and Related Surgery, Vol 18, No 9 (November-December), 2002: pp 968 –973
TUNNEL PLACEMENT WITH ACTIVE ROBOTIC SYSTEM gery (CAOS) revealed significant reductions of variability in graft tunnel placement with these systems.14-16 Active robotic systems can autonomously drill tunnels for ACL reconstruction procedures.17 A surgical planning station receives computed tomography (CT) data of a patient’s knee, and the surgeon plans the femoral and tibial tunnels.18,19 The tibial tunnel is drilled from outside-to-inside, and a miniarthrotomy allows the femoral tunnel to be drilled separately. Because the goal of CAOS is to increase precision and accuracy, the difference between these terms needs to be understood. Precision is defined as the degree of agreement between repeated results.20 A prior study compared the precision of tunnel placement during ACL reconstruction found with an active robot system with the precision of a traditional technique using an arthroscope and drill guides.21 A significant difference in tibial graft tunnel placement was shown for only 1 of the 4 participating surgeons compared with the active robot system. For the femoral tunnel, the active robotic technique was more precise than 2 of the 4 surgeons. Accuracy is defined as the maximum amount by which a result differs from a true value.20 The goal of this study was to evaluate the accuracy of tunnel placement and drilling using an active robot system. METHODS For this study, 13 fresh-frozen cadaveric knees were used; the mean donor age was 65 years (range, 50 to 75 years). Knees with prior osteotomies, total knee replacements, or posterior cruciate ligament (PCL) deficiency were excluded from the study. Reference screws (U.R.S. Ortho, Rastatt, Germany) with a diameter of 4.5 mm and a target that contained 4 fiducials (Fig 1) were placed in the lateral distal femur and the proximal medial tibia as suggested by the manufacturer. A CT scan (General Electric, 9800, General Electric Medical Systems, Waukesha, WI) was then performed with the knee positioned in hyperextension. These data were transferred to the planning station to develop a preoperative plan. CT Procedure and Surgical Planning An overview scout was initially obtained from approximately 2 cm proximal to the femoral reference screw to 2 cm distal of the tibial reference screw. The first CT spiral (parameters: 140 kV, 130 mA, 3-mm slice collimation, 1° pitch, 1-mm intervals, 0.8 second
969
FIGURE 1. (A) The 4.5-mm cortical screw, (B) CT fiducial target, (C) robot fiducial target, and (D) specially altered target, with (E) additional points for registering and generating a coordinate system, used for digitizing measurements.
rotation time, helical ⫹ Z interpolation) included the entire reference screw and 4 fiducials. The second spiral (140 kV, 120 mA, 3-mm slice collimation, 1.5° pitch, 2-mm intervals, 0.8 second rotation time, helical ⫹ Z interpolation) was obtained between the 2 reference screws. The last spiral (140 kV, 130 mA, 3-mm slice collimation, 1° pitch, 1-mm intervals, 0.8 second rotation time, helical ⫹ Z interpolation) began at the end position of the second spiral and included the entire tibial reference screw. The field-of-view of each image was 250 mm, with a pixel resolution of 512 ⫻ 512. The tunnels for the robotic procedure were planned with the customized software on the planning station (CASPAR, U.R.S. Ortho). This is similar to the procedure recommended when using the quadrant method for the femoral tunnel.22 This method is based on anatomic studies by Bernard et al.22 The location of the center point of the femoral insertion site of the ACL was evaluated with respect to Blumensaat’s line and the width of the femoral condyles. Two parameters are given, expressed in percent values: (1) distance of the center of the insertion to Blumensaat’s line and (2) distance to the posterior margin of the femoral condyle. They were determined to be 25% and 25% for the geometrical center of the femoral insertion site of the ACL.22 The software therefore creates the planned femoral tunnel at (1) 25% and (2) 25% by default. The tibial tunnel is then created on the planning station by creating a line beneath the cortex of the medial tibial plateau that mirrors the anteroposterior (AP) diameter of the tibia.23 Along this line, an indi-
970
V. MUSAHL ET AL. Measurement of Preoperative and Postoperative Tunnels
FIGURE 2. Experimental set-up with (A) the specimen fixed in the (B) custom-made holder and (C) the robotic reamer as used in the operating room.
vidually planned tunnel oriented to the patient’s anatomy can be created; in this study protocol, 43% of the AP diameter was chosen. Tunnel Drilling Procedure The surgical plan was then transferred to the active robot system (CASPAR) using a memory card. The specimens were mounted on a customized jig as prescribed by the manufacturer, and 2 Steinmann pins were inserted into the medial and lateral regions of the femoral epicondyle to obtain rigid fixation of the specimen. The tibia was held with 2 additional Steinmann pins (Fig 2). To orient the robot in space, registration of the femur and tibia was subsequently performed. A special pointer tool was mounted on the robot end-effector to register the reference screws via the 4 fiducials. The error in this methodology has been determined by the manufacturer to be less than 0.5 mm for localizing points within the workspace of the robot. During the drilling procedure, motion of the femur and tibia with respect to the end-effector of the robot was monitored with a video analysis system (Polaris; Northern Digital, Waterloo, ON). The system immediately stops the drilling procedure if excessive or accidental bone motion occurs. Using a 10-mm reamer attached to the end-effector of the robot, the tibial tunnel was drilled first, followed by the femoral tunnel through an extra mini-arthrotomy. Each procedure was recorded using a video camera, and the total time was recorded as well. After the entire procedure was completed, a gross inspection of the specimen was performed and any injuries to intra-articular or extra-articular structures were noted.
To quantitatively evaluate the accuracy of the procedure, the coordinates of the tunnel locations were obtained, creating a coordinate system from the 4 fiducials on each reference screw. Custom software on the planning station provided by the manufacturer determined the coordinates of the fiducial points and tunnel locations with respect to a second coordinate system (in CT). An independently developed methodology was then used to evaluate the locations of the drilled tunnels using a mechanical digitizing device (Microscribe-3DX; Immersion, San Jose, CA; accuracy, 0.23 mm) postoperatively. The locations of each tunnel were determined with respect to the digitizing device by inserting a 10-mm diameter brass plug and digitizing the center point on the surface of each end of the plug (Fig 3). This enabled a repeatable definition of these points.21 A common coordinate system could then be established between the CT (reference screw) and the digitizing device, and accuracy measurements could be taken. Final Evaluation of Drilled Tunnels The preoperative plan could be compared with the drilled tunnel using the coordinates of 2 points on the centerline of each tunnel representing the tunnel locations. These points were used to define a vector representing each tunnel defined in the previously established common coordinate system. The distance (D-deviation) and the angle () between the two tunnels (vectors) were calculated (Fig 4). Measurements were made at the intra-articular tunnel location. Due to the complicated bony geometry and uncertainty in
FIGURE 3. (A) Tibia and (B) femur with (C) reference screw. (D) Tunnel plug in the tibial bone tunnel.
TUNNEL PLACEMENT WITH ACTIVE ROBOTIC SYSTEM
971
the feed rate by 75% to prevent sliding of the drill bit on the cortical bone. RESULTS
FIGURE 4. Methodology to determine the deviation between the axis of the (A) planned tunnel and (B) drilled tunnel in a common coordinate system. D is defined as the deviation in translation and as the deviation in direction.
the definition of tunnel entry points in CT-data and the postoperative measurements, the deviation was calculated as shown in Fig 4. A vector was calculated perpendicular to the planned tunnel, reaching to the vector representing the drilled tunnel. The magnitude of this vector was defined as the deviation, or accuracy. Preliminary Trials Initially, the entire procedure was performed on 3 cadaver knees to examine the protocol and determine the reliability of the measurement method. In these preliminary trials, the tibial tunnel was inaccurately drilled due to a displacement of the drilling device (diamond drill bit) at the tunnel entry point on the cortical bone. These errors were clearly visible on the video recording and reached almost 2 mm in length without being detected by the robotic system. These experiments also revealed that the knee must be flexed more than 120° during drilling of the femoral tunnel. The system had trouble accessing the femoral insertion site and, on occasion, came into contact with the tibial spine when angles less than 120° of flexion were used. For these 3 specimens, the intra-articular tunnel location deviated by 8.8 ⫾ 5.7 mm (range, 4.9 to 15.4 mm) from the preoperative plan, and the direction of the drilled tunnel deviated by 6.3° ⫾ 7.3° (range, 0.8° to 14.7°). On the femoral side, the intra-articular tunnel location deviated by 5.8 ⫾ 4.6 mm (range, 2.0 to 11.1 mm) and the deviation in direction of the drilled tunnel was found to be 5.1° ⫾ 4.8° (range, 0.5° to 10.2°) (Table 1). Therefore, the manufacturer reduced
In 10 cadaveric knees using the reduced feed rate, deviation from the planned tunnel to the drilled tunnel of 2.0 ⫾ 1.2 mm for the intra-articular tunnel location and 1.1° ⫾ 0.7° for the tunnel direction were achieved, respectively, for the tibia. On the femoral side, the intra-articular tunnel location deviated by 1.3 ⫾ 0.9 mm, and the tunnel direction deviated by 1.1° ⫾ 0.6° (Table 2). No sliding of the drill at the tibial entry point was observed. Furthermore, macroscopic inspection showed that the robotic system did not injure other intra-articular or extra-articular structures such as the popliteal vessels, cartilage, PCL, patellar bone, or meniscus during the drilling procedures. The time for tunnel drilling on the tibial side was 4.6 ⫾ 1.4 minutes. On the femoral side, it was 5.5 ⫾ 1.5 minutes (Table 3). Difficulty with the active robot system occurred with 3 knees. Specimen No. 3 had the greatest deviation. For that specimen, deviations of 4.3 mm for the intra-articular tunnel location and 0.5° for tunnel direction were revealed for the tibia. For the femur, the intra-articular tunnel location was determined as 2.8 mm and the tunnel direction as 2.1° from the preoperative plan. Several times throughout the drilling procedure for this specimen, excessive bone motion occurred. Therefore, the active robot system stopped the tibial drilling procedure, and the system was assisted by manually rotating the drill bit. This also explains the increased drilling time for this specimen and indicates that the fixation of the specimen might not have been sufficient. The measurement of the tibial tunnel for specimen No. 8 was not possible for the drilling procedure because the reference screw
TABLE 1. Deviation Between Vectors of Planned and Drilled Tunnels of 3 Preliminary Human Cadaver Knees Tibia
Femur
Specimen
Direction
Location* (mm)
Direction
Location* (mm)
1 2 3 Mean ⫾ SD
14.7° 3.4° 0.8° 6.3° ⫾ 7.4°
15.5 4.9 6.1 8.8 ⫾ 5.8
10.2° 4.4° 0.5° 5.1° ⫾ 4.9°
11.1 4.5 2.0 5.9 ⫾ 4.7
Abbreviation: SD, standard deviation. *Intra-articular location.
972
V. MUSAHL ET AL. TABLE 2. Deviation Between Vectors of Planned and Drilled Tunnels of 10 Human Cadaver Knees Tibia
Femur
Specimen
Direction
Location* (mm)
Direction
Location* (mm)
1 2 3 4 5 6 7 8 9 10 Mean ⫾ SD
1.2° 2.1° 0.5° 0.2° 0.9° 1.3° 1.8° nm 0.4° 1.4° 1.1° ⫾ 0.7°
1.2 1.6 4.3 0.5 0.7 2.3 2.0 nm 3.4 2.4 2.0 ⫾ 1.2
1.5° 1.5° 2.1° 0.7° 0.4° 0.9° 0.9° 1.7° 0.8° 0.3° 1.1° ⫾ 0.6°
0.7 0.2 2.8 0.6 2.4 0.2 1.2 1.5 1.7 1.4 1.3 ⫾ 0.9
Abbreviations: nm, not measured because position of reference screw changed during testing; SD, standard deviation. *Intra-articular location.
turned during the removal of the target with the 4 fiducials. DISCUSSION In this study, accuracy of tunnel placement for ACL reconstruction using an active robot system was evaluated by comparing planned tunnels with drilled tunnels. The deviation at the intra-articular location of the drilled tunnel on the tibia was found to be less than 2.0 mm, and it was less than 1.3 mm for the femur. Similar magnitudes of error have been reported using an augmented reality system for arthroscopic surgery in the knee. The overall system error was 2.3 mm.24 Accuracy for the active robot system was previously reported to be less than 1 mm, when assessing accuTABLE 3. Times for Drilling the Tibial and Femoral Tunnels Drilling Time (min) Specimen
Tibia
Femur
1 2 3 4 5 6 7 8 9 10 Mean ⫾ SD
3.3 4.5 7.0 5.0 2.4 4.2 5.0 6.2 5.0 3.1 4.6 ⫾ 1.4
4.0 4.0 5.0 6.0 8.3 6.2 5.2 4.5 7.3 4.1 5.5 ⫾ 1.5
Abbreviation: SD, standard deviation.
racy with either the active robot25 or postoperative radiographs.19 During preliminary trials, inaccurate placement of the tunnels occurred because the diamond-tipped drill bit slid on the cortical bone of the tibia. The drill bit contacted the tibia at an angle of 60°. Based on our experiments, this angle and soft tissue (periostium) on the surface of the tibia contributed to this sliding. To minimize sliding, a lower feed rate was used in our study, and the system was changed accordingly. An alternative approach may be to increase the angle between the drill bit and the cortical bone. Placement of the tunnels using an active robot system depends not only on the mechanical accuracy of the system but also on the ability of the surgeon to properly use the custom software on the planning station. Surgeons commonly position their tunnels according to anatomic landmarks observed intra-articularly using an arthroscope. The relationship of the intra-articular landmarks and CT-identified landmarks is unknown. Users of this system must learn to properly interpret the CT data to achieve the desired tunnel placement.26 Measurements of accuracy were limited by the precision of the mechanical digitizer (0.23 mm) as well as the relationship between the holes drilled in the reference screws for matching the reference screw and CT coordinate system. The use of a mini-arthrotomy as well as the necessary insertion of 2 reference screws into the patient’s femur and tibia could add morbidity to ACL reconstruction with the active robot. Although the specimen was rigidly fixed to the robot with the addition of pins in the femur and tibia, bone motion occurred during the drilling procedure
TUNNEL PLACEMENT WITH ACTIVE ROBOTIC SYSTEM and was unacceptable in 1 case. In the operating room, such bone motion may be reduced because the leg can be held at the thigh and foot while the knee is fixed to the robot.17,19 In this study, a control group could not be used to compare accuracy because the true value for the planned tunnel could not be assessed using the traditional technique. In a previous study by Burkart et al.,21 however, precision with an active robot system was assessed and could be compared with the traditional technique, because differences in repeated planned tunnels could be easily assessed. The distribution of intra-articular locations on the tibia and femur were reported to be inside a sphere with a radius of 2.0 mm for the active robot system. The values for accuracy reported in this study are therefore within reported ranges. With an active robot system, accurate tunnel placement, especially with regard to tunnel location and direction, can be achieved.21 However, the optimal tunnel location for best outcome, whether in the geometric center of the ACL or elsewhere, remains unknown. Therefore, a statement that active robotic surgery can be highly accurate is valid, but a statement that active robotic surgery is beneficial for orthopedic patients might be premature. After all, robotic surgery is only as accurate as the surgeon who plans the procedure. Furthermore, a question remains regarding how much accuracy is necessary for tunnel placement during ACL reconstruction. Future studies should examine specific factors that can affect the accuracy of this system, such as the registration of bony geometry within the robotic system, motion of the femur and tibia during the drilling procedure, and the stiffness of various drilling tools. The addition of 3-dimensional images to the planning station as well as virtual kinematics will further improve these systems. REFERENCES 1. Eriksson E. How good are the results of ACL reconstruction? Knee Surg Sports Traumatol Arthrosc 1997;5:137 (editorial). 2. Musahl V, Cierpinski T, Hornung H, Hertel P. Sekundaere vordere Kreuzbandplastik nach primaerer Naht und Kreuzbandplastik. Jahrestagung der DGU, Berlin, November 1999. 3. Sati M, Sta¨ ubli HU, Bourquin Y, Kunz M, Ka¨ sermann S, Nolte LP. Clinical integration of new computer assisted technology for arthroscopic ACL replacement. Oper Tech Orthop 2000; 10:40-49. 4. Yaru NC, Daniel DM, Penner D. The effect of tibial attachment site on graft impingement in an anterior cruciate ligament reconstruction. Am J Sports Med 1992;20:217-220. 5. Friedman RL, Feagin JA Jr. Topographical anatomy of the intercondylar roof: A pilot study. Clin Orthop Rel Res 1994; 306:163-170. 6. Morgan CD, Kalman VR, Grawl DM. Definitive landmarks for reproducible tibial tunnel placement in anterior cruciate ligament reconstruction. Arthroscopy 1995;11:275-288.
973
7. Howell SM, Taylor MA. Failure of reconstruction of the anterior cruciate ligament due to impingement by the intercondylar roof. J Bone Joint Surg Am 1993;75:1044-1055. 8. Vergis A, Gillquist J. Graft failure in intra-articular anterior cruciate ligament reconstructions: A review of the literature. Arthroscopy 1995;11:312-321. 9. Harner CD, Giffin JR, Dunteman RC, Annunziata CC, Friedman MJ. Evaluation and treatment of recurrent instability after anterior cruciate ligament reconstruction. J Bone Joint Surg Am 2000;82:1652-1664. 10. Buzzi R, Zaccherotti G, Giron F, Aglietti P. The relationship between the intercondylar roof and the tibial plateau with the knee in extension: Relevance for tibial tunnel placement in anterior cruciate ligament reconstruction. Arthroscopy 1999; 15:625-631. 11. Jalliard R, Lavallee S, Dessenne V. Computer assisted reconstruction of the anterior cruciate ligament. Clin Orthop Rel Res 1998;354:57-64. 12. Sommer C, Friederich NF, Muller W. Improperly placed anterior cruciate ligament grafts: correlation between radiological parameters and clinical results. Knee Surg Sports Traumatol Arthrosc 2000;8:207-213. 13. Kohn D, Busche T, Carls T. Drill hole position in endoscopic anterior cruciate ligament reconstruction: Results of an advanced arthroscopy course. Knee Surg Sports Traumatol Arthrosc 1998;6:13-15 (suppl 1). 14. Desenne V, Lavallee S, Julliard R, Orti R, Martelli S, Cinquin P. Computer-assisted knee anterior cruciate ligament reconstruction: First clinical tests. J Image Guided Surg 1995;1:59-64. 15. Klos TV, Habets RJE, Banks AZ, Banks SA, DeVilee RJJ, Cook FF. Computer assistance in arthroscopic anterior cruciate ligament reconstruction. Clin Orthop Rel Res 1998;354:65-69. 16. Sati M, de Guise JA, Drouin G. Computer assisted knee surgery: Diagnostics and planning of knee surgery. Comput Aided Surg 1997;2:108-123. 17. Gotzen L, Schierl M, Pashminch Azar A, Kober R, Heeckt PF, Petermann J. The CASPAR-system (computer assisted surgical planning and robotics) in the reconstruction of the ACL. Presented at the 9th Annual ESSKA Meeting, London, September 2000. 18. Petermann J, Kober R, Heinze P. Computer-assisted planning and robot-assisted surgery in anterior cruciate ligament reconstruction. Oper Tech Orthop September 2000;10. 19. Bernard M. robot-assisted tunnel placement in ACL surgery. 9th Annual ESSKA Meeting, London, 2000. 20. Beckwith T, Buck N, Marangoni R. Treatment of uncertainties. In: A WP, ed. Mechanical measurements. 1982. 21. Burkart A, Debski RE, McMahon PJ, et al. A comparison of precision for ACL tunnel placement using traditional and robotic technique. Comput Aided Surg 2001;6:270-278. 22. Bernard M, Hertal P, Hornung H, Cierpinski T. Femoral insertion of the ACL: Radiographic quadrant method. Am J Knee Surg 1997;10:14-21 (discussion 21-22). 23. Amis AA, Jakob RP. Anterior cruciate ligament graft positioning, tensioning and twisting. Knee Surg Sports Traumatol Arthrosc 1998;6:2-12 (suppl 1). 24. Tonet O, Megali G, D’Attanasio S, et al. An augmented realitiy navigation system for comptuer assisted arthroscopic surgery of the knee. Computer Assisted Orthopaedic Surgery USA, Pittsburgh, PA, 2000. 25. Petermann J, Schierl M, Niess C, Kober R, Heinze P, Gotzen L. Computerassistierte Planung und roboterassistierte Ersatzplastik des vorderen Kreuzbandes mit dem CASPAR-System. Arthroskopie 2000;13:270-279. 26. Musahl V, Burkart A, Fu F. Tunnelplatzierung: Definition der Anatomischen Topographie. In: Deutsche Gesellschaft fuer Orthopaedie und Traumatologie. Berlin, Germany, October 2001.
Purpose: The objective of this study was to evaluate the accuracy of tunnel placement for ACL reconstruction performed with an active robotic system. Type of Study: Cadaveric analysis. Methods: A reference screw containing 4 fiducials was placed in the femur and tibia of 13 fresh-frozen cadaveric knees. A preoperative plan was developed using images from 3-dimensional computed tomography reconstructions of the knee. The active robotic system then drilled the tunnels. The location and direction of each planned tunnel in the femur and tibia were determined from the preoperative plan. To compare these parameters postoperatively, a mechanical digitizer and a tunnel plug were used. The deviation in location and direction between the planned and drilled tunnel was determined. Results: In preliminary trials, the tibial tunnel was located inaccurately because slippage of the drill bit occurred on the bone at the start of tunnel drilling. This was minimized by decreasing the feed rate of the robot by 75%. For the remaining 10 knees, deviations with respect to the preoperative plan were found of 2.0 ⫾ 1.2 mm and 1.1° ⫾ 0.7° for the intra-articular tibial tunnel location and direction, respectively. For the femur, the deviations were 1.3 ⫾ 0.9 mm for the tunnel location (intra-articular) and 1.0° ⫾ 0.6° for the tunnel direction. Conclusions: The active robotic system is highly accurate for tunnel placement during ACL reconstruction, meaning that the robot drills the tunnels very close to the surgeon’s plan. Comparison to a control group of surgeons could not be made because no preoperative plan is usually created in traditional surgery. However, accuracy values in this study were found to be below the values for precision of repeated tunnel placements reported in the literature. Key Words: Computer-assisted orthopedic surgery—ACL—Accuracy— Tunnel placement—CASPAR.
O
ne of the proposed advantages of computerassisted surgery in anterior cruciate ligament (ACL) reconstruction is improvement in the accuracy of the surgical procedure compared with traditional methods. Outcomes for ACL reconstruction still do not exceed 90% good or excellent results.1 It is also well known that placement of the tunnels in the femur and tibia influences the outcome of ACL reconstruc-
From the Musculoskeletal Research Center, Department of Orthopaedic Surgery, Pittsburgh, Pennsylvania, U.S.A. Address correspondence and reprint requests to Freddie H. Fu, M.D., Musculoskeletal Research Center, Department of Orthopaedic Surgery, E1641 Bio Medical Tower, 210 Lothrop St, PO Box 71199, Pittsburgh, PA 15213, U.S.A. E-mail: [email protected] © 2002 by the Arthroscopy Association of North America 0749-8063/02/1809-3107$35.00/0 doi:10.1053/jars.2002.36110
968
tion procedures.2,3 Specifically, many studies have extensively examined the tunnel positions and found that inappropriate graft placement interfered with graft incorporation and knee function.4-6 Inappropriate positioning of the tunnels can also result in excessive changes in graft length as the knee moves through the range of motion.7,8 The femoral tunnel is the most commonly misplaced. Because the femoral attachment of the ACL is closer to the rotational center of the knee, small errors can be detrimental to knee function.9 Complications such as graft failure, diminished range of motion, and graft impingement with the femoral notch can result.4,10,11 Despite the great variability in tunnel placement between surgeons, each surgeon attempts to consistently place tunnels in the same location from knee to knee.12,13 Initial reports on computer-assisted orthopedic sur-
Arthroscopy: The Journal of Arthroscopic and Related Surgery, Vol 18, No 9 (November-December), 2002: pp 968 –973
TUNNEL PLACEMENT WITH ACTIVE ROBOTIC SYSTEM gery (CAOS) revealed significant reductions of variability in graft tunnel placement with these systems.14-16 Active robotic systems can autonomously drill tunnels for ACL reconstruction procedures.17 A surgical planning station receives computed tomography (CT) data of a patient’s knee, and the surgeon plans the femoral and tibial tunnels.18,19 The tibial tunnel is drilled from outside-to-inside, and a miniarthrotomy allows the femoral tunnel to be drilled separately. Because the goal of CAOS is to increase precision and accuracy, the difference between these terms needs to be understood. Precision is defined as the degree of agreement between repeated results.20 A prior study compared the precision of tunnel placement during ACL reconstruction found with an active robot system with the precision of a traditional technique using an arthroscope and drill guides.21 A significant difference in tibial graft tunnel placement was shown for only 1 of the 4 participating surgeons compared with the active robot system. For the femoral tunnel, the active robotic technique was more precise than 2 of the 4 surgeons. Accuracy is defined as the maximum amount by which a result differs from a true value.20 The goal of this study was to evaluate the accuracy of tunnel placement and drilling using an active robot system. METHODS For this study, 13 fresh-frozen cadaveric knees were used; the mean donor age was 65 years (range, 50 to 75 years). Knees with prior osteotomies, total knee replacements, or posterior cruciate ligament (PCL) deficiency were excluded from the study. Reference screws (U.R.S. Ortho, Rastatt, Germany) with a diameter of 4.5 mm and a target that contained 4 fiducials (Fig 1) were placed in the lateral distal femur and the proximal medial tibia as suggested by the manufacturer. A CT scan (General Electric, 9800, General Electric Medical Systems, Waukesha, WI) was then performed with the knee positioned in hyperextension. These data were transferred to the planning station to develop a preoperative plan. CT Procedure and Surgical Planning An overview scout was initially obtained from approximately 2 cm proximal to the femoral reference screw to 2 cm distal of the tibial reference screw. The first CT spiral (parameters: 140 kV, 130 mA, 3-mm slice collimation, 1° pitch, 1-mm intervals, 0.8 second
969
FIGURE 1. (A) The 4.5-mm cortical screw, (B) CT fiducial target, (C) robot fiducial target, and (D) specially altered target, with (E) additional points for registering and generating a coordinate system, used for digitizing measurements.
rotation time, helical ⫹ Z interpolation) included the entire reference screw and 4 fiducials. The second spiral (140 kV, 120 mA, 3-mm slice collimation, 1.5° pitch, 2-mm intervals, 0.8 second rotation time, helical ⫹ Z interpolation) was obtained between the 2 reference screws. The last spiral (140 kV, 130 mA, 3-mm slice collimation, 1° pitch, 1-mm intervals, 0.8 second rotation time, helical ⫹ Z interpolation) began at the end position of the second spiral and included the entire tibial reference screw. The field-of-view of each image was 250 mm, with a pixel resolution of 512 ⫻ 512. The tunnels for the robotic procedure were planned with the customized software on the planning station (CASPAR, U.R.S. Ortho). This is similar to the procedure recommended when using the quadrant method for the femoral tunnel.22 This method is based on anatomic studies by Bernard et al.22 The location of the center point of the femoral insertion site of the ACL was evaluated with respect to Blumensaat’s line and the width of the femoral condyles. Two parameters are given, expressed in percent values: (1) distance of the center of the insertion to Blumensaat’s line and (2) distance to the posterior margin of the femoral condyle. They were determined to be 25% and 25% for the geometrical center of the femoral insertion site of the ACL.22 The software therefore creates the planned femoral tunnel at (1) 25% and (2) 25% by default. The tibial tunnel is then created on the planning station by creating a line beneath the cortex of the medial tibial plateau that mirrors the anteroposterior (AP) diameter of the tibia.23 Along this line, an indi-
970
V. MUSAHL ET AL. Measurement of Preoperative and Postoperative Tunnels
FIGURE 2. Experimental set-up with (A) the specimen fixed in the (B) custom-made holder and (C) the robotic reamer as used in the operating room.
vidually planned tunnel oriented to the patient’s anatomy can be created; in this study protocol, 43% of the AP diameter was chosen. Tunnel Drilling Procedure The surgical plan was then transferred to the active robot system (CASPAR) using a memory card. The specimens were mounted on a customized jig as prescribed by the manufacturer, and 2 Steinmann pins were inserted into the medial and lateral regions of the femoral epicondyle to obtain rigid fixation of the specimen. The tibia was held with 2 additional Steinmann pins (Fig 2). To orient the robot in space, registration of the femur and tibia was subsequently performed. A special pointer tool was mounted on the robot end-effector to register the reference screws via the 4 fiducials. The error in this methodology has been determined by the manufacturer to be less than 0.5 mm for localizing points within the workspace of the robot. During the drilling procedure, motion of the femur and tibia with respect to the end-effector of the robot was monitored with a video analysis system (Polaris; Northern Digital, Waterloo, ON). The system immediately stops the drilling procedure if excessive or accidental bone motion occurs. Using a 10-mm reamer attached to the end-effector of the robot, the tibial tunnel was drilled first, followed by the femoral tunnel through an extra mini-arthrotomy. Each procedure was recorded using a video camera, and the total time was recorded as well. After the entire procedure was completed, a gross inspection of the specimen was performed and any injuries to intra-articular or extra-articular structures were noted.
To quantitatively evaluate the accuracy of the procedure, the coordinates of the tunnel locations were obtained, creating a coordinate system from the 4 fiducials on each reference screw. Custom software on the planning station provided by the manufacturer determined the coordinates of the fiducial points and tunnel locations with respect to a second coordinate system (in CT). An independently developed methodology was then used to evaluate the locations of the drilled tunnels using a mechanical digitizing device (Microscribe-3DX; Immersion, San Jose, CA; accuracy, 0.23 mm) postoperatively. The locations of each tunnel were determined with respect to the digitizing device by inserting a 10-mm diameter brass plug and digitizing the center point on the surface of each end of the plug (Fig 3). This enabled a repeatable definition of these points.21 A common coordinate system could then be established between the CT (reference screw) and the digitizing device, and accuracy measurements could be taken. Final Evaluation of Drilled Tunnels The preoperative plan could be compared with the drilled tunnel using the coordinates of 2 points on the centerline of each tunnel representing the tunnel locations. These points were used to define a vector representing each tunnel defined in the previously established common coordinate system. The distance (D-deviation) and the angle () between the two tunnels (vectors) were calculated (Fig 4). Measurements were made at the intra-articular tunnel location. Due to the complicated bony geometry and uncertainty in
FIGURE 3. (A) Tibia and (B) femur with (C) reference screw. (D) Tunnel plug in the tibial bone tunnel.
TUNNEL PLACEMENT WITH ACTIVE ROBOTIC SYSTEM
971
the feed rate by 75% to prevent sliding of the drill bit on the cortical bone. RESULTS
FIGURE 4. Methodology to determine the deviation between the axis of the (A) planned tunnel and (B) drilled tunnel in a common coordinate system. D is defined as the deviation in translation and as the deviation in direction.
the definition of tunnel entry points in CT-data and the postoperative measurements, the deviation was calculated as shown in Fig 4. A vector was calculated perpendicular to the planned tunnel, reaching to the vector representing the drilled tunnel. The magnitude of this vector was defined as the deviation, or accuracy. Preliminary Trials Initially, the entire procedure was performed on 3 cadaver knees to examine the protocol and determine the reliability of the measurement method. In these preliminary trials, the tibial tunnel was inaccurately drilled due to a displacement of the drilling device (diamond drill bit) at the tunnel entry point on the cortical bone. These errors were clearly visible on the video recording and reached almost 2 mm in length without being detected by the robotic system. These experiments also revealed that the knee must be flexed more than 120° during drilling of the femoral tunnel. The system had trouble accessing the femoral insertion site and, on occasion, came into contact with the tibial spine when angles less than 120° of flexion were used. For these 3 specimens, the intra-articular tunnel location deviated by 8.8 ⫾ 5.7 mm (range, 4.9 to 15.4 mm) from the preoperative plan, and the direction of the drilled tunnel deviated by 6.3° ⫾ 7.3° (range, 0.8° to 14.7°). On the femoral side, the intra-articular tunnel location deviated by 5.8 ⫾ 4.6 mm (range, 2.0 to 11.1 mm) and the deviation in direction of the drilled tunnel was found to be 5.1° ⫾ 4.8° (range, 0.5° to 10.2°) (Table 1). Therefore, the manufacturer reduced
In 10 cadaveric knees using the reduced feed rate, deviation from the planned tunnel to the drilled tunnel of 2.0 ⫾ 1.2 mm for the intra-articular tunnel location and 1.1° ⫾ 0.7° for the tunnel direction were achieved, respectively, for the tibia. On the femoral side, the intra-articular tunnel location deviated by 1.3 ⫾ 0.9 mm, and the tunnel direction deviated by 1.1° ⫾ 0.6° (Table 2). No sliding of the drill at the tibial entry point was observed. Furthermore, macroscopic inspection showed that the robotic system did not injure other intra-articular or extra-articular structures such as the popliteal vessels, cartilage, PCL, patellar bone, or meniscus during the drilling procedures. The time for tunnel drilling on the tibial side was 4.6 ⫾ 1.4 minutes. On the femoral side, it was 5.5 ⫾ 1.5 minutes (Table 3). Difficulty with the active robot system occurred with 3 knees. Specimen No. 3 had the greatest deviation. For that specimen, deviations of 4.3 mm for the intra-articular tunnel location and 0.5° for tunnel direction were revealed for the tibia. For the femur, the intra-articular tunnel location was determined as 2.8 mm and the tunnel direction as 2.1° from the preoperative plan. Several times throughout the drilling procedure for this specimen, excessive bone motion occurred. Therefore, the active robot system stopped the tibial drilling procedure, and the system was assisted by manually rotating the drill bit. This also explains the increased drilling time for this specimen and indicates that the fixation of the specimen might not have been sufficient. The measurement of the tibial tunnel for specimen No. 8 was not possible for the drilling procedure because the reference screw
TABLE 1. Deviation Between Vectors of Planned and Drilled Tunnels of 3 Preliminary Human Cadaver Knees Tibia
Femur
Specimen
Direction
Location* (mm)
Direction
Location* (mm)
1 2 3 Mean ⫾ SD
14.7° 3.4° 0.8° 6.3° ⫾ 7.4°
15.5 4.9 6.1 8.8 ⫾ 5.8
10.2° 4.4° 0.5° 5.1° ⫾ 4.9°
11.1 4.5 2.0 5.9 ⫾ 4.7
Abbreviation: SD, standard deviation. *Intra-articular location.
972
V. MUSAHL ET AL. TABLE 2. Deviation Between Vectors of Planned and Drilled Tunnels of 10 Human Cadaver Knees Tibia
Femur
Specimen
Direction
Location* (mm)
Direction
Location* (mm)
1 2 3 4 5 6 7 8 9 10 Mean ⫾ SD
1.2° 2.1° 0.5° 0.2° 0.9° 1.3° 1.8° nm 0.4° 1.4° 1.1° ⫾ 0.7°
1.2 1.6 4.3 0.5 0.7 2.3 2.0 nm 3.4 2.4 2.0 ⫾ 1.2
1.5° 1.5° 2.1° 0.7° 0.4° 0.9° 0.9° 1.7° 0.8° 0.3° 1.1° ⫾ 0.6°
0.7 0.2 2.8 0.6 2.4 0.2 1.2 1.5 1.7 1.4 1.3 ⫾ 0.9
Abbreviations: nm, not measured because position of reference screw changed during testing; SD, standard deviation. *Intra-articular location.
turned during the removal of the target with the 4 fiducials. DISCUSSION In this study, accuracy of tunnel placement for ACL reconstruction using an active robot system was evaluated by comparing planned tunnels with drilled tunnels. The deviation at the intra-articular location of the drilled tunnel on the tibia was found to be less than 2.0 mm, and it was less than 1.3 mm for the femur. Similar magnitudes of error have been reported using an augmented reality system for arthroscopic surgery in the knee. The overall system error was 2.3 mm.24 Accuracy for the active robot system was previously reported to be less than 1 mm, when assessing accuTABLE 3. Times for Drilling the Tibial and Femoral Tunnels Drilling Time (min) Specimen
Tibia
Femur
1 2 3 4 5 6 7 8 9 10 Mean ⫾ SD
3.3 4.5 7.0 5.0 2.4 4.2 5.0 6.2 5.0 3.1 4.6 ⫾ 1.4
4.0 4.0 5.0 6.0 8.3 6.2 5.2 4.5 7.3 4.1 5.5 ⫾ 1.5
Abbreviation: SD, standard deviation.
racy with either the active robot25 or postoperative radiographs.19 During preliminary trials, inaccurate placement of the tunnels occurred because the diamond-tipped drill bit slid on the cortical bone of the tibia. The drill bit contacted the tibia at an angle of 60°. Based on our experiments, this angle and soft tissue (periostium) on the surface of the tibia contributed to this sliding. To minimize sliding, a lower feed rate was used in our study, and the system was changed accordingly. An alternative approach may be to increase the angle between the drill bit and the cortical bone. Placement of the tunnels using an active robot system depends not only on the mechanical accuracy of the system but also on the ability of the surgeon to properly use the custom software on the planning station. Surgeons commonly position their tunnels according to anatomic landmarks observed intra-articularly using an arthroscope. The relationship of the intra-articular landmarks and CT-identified landmarks is unknown. Users of this system must learn to properly interpret the CT data to achieve the desired tunnel placement.26 Measurements of accuracy were limited by the precision of the mechanical digitizer (0.23 mm) as well as the relationship between the holes drilled in the reference screws for matching the reference screw and CT coordinate system. The use of a mini-arthrotomy as well as the necessary insertion of 2 reference screws into the patient’s femur and tibia could add morbidity to ACL reconstruction with the active robot. Although the specimen was rigidly fixed to the robot with the addition of pins in the femur and tibia, bone motion occurred during the drilling procedure
TUNNEL PLACEMENT WITH ACTIVE ROBOTIC SYSTEM and was unacceptable in 1 case. In the operating room, such bone motion may be reduced because the leg can be held at the thigh and foot while the knee is fixed to the robot.17,19 In this study, a control group could not be used to compare accuracy because the true value for the planned tunnel could not be assessed using the traditional technique. In a previous study by Burkart et al.,21 however, precision with an active robot system was assessed and could be compared with the traditional technique, because differences in repeated planned tunnels could be easily assessed. The distribution of intra-articular locations on the tibia and femur were reported to be inside a sphere with a radius of 2.0 mm for the active robot system. The values for accuracy reported in this study are therefore within reported ranges. With an active robot system, accurate tunnel placement, especially with regard to tunnel location and direction, can be achieved.21 However, the optimal tunnel location for best outcome, whether in the geometric center of the ACL or elsewhere, remains unknown. Therefore, a statement that active robotic surgery can be highly accurate is valid, but a statement that active robotic surgery is beneficial for orthopedic patients might be premature. After all, robotic surgery is only as accurate as the surgeon who plans the procedure. Furthermore, a question remains regarding how much accuracy is necessary for tunnel placement during ACL reconstruction. Future studies should examine specific factors that can affect the accuracy of this system, such as the registration of bony geometry within the robotic system, motion of the femur and tibia during the drilling procedure, and the stiffness of various drilling tools. The addition of 3-dimensional images to the planning station as well as virtual kinematics will further improve these systems. REFERENCES 1. Eriksson E. How good are the results of ACL reconstruction? Knee Surg Sports Traumatol Arthrosc 1997;5:137 (editorial). 2. Musahl V, Cierpinski T, Hornung H, Hertel P. Sekundaere vordere Kreuzbandplastik nach primaerer Naht und Kreuzbandplastik. Jahrestagung der DGU, Berlin, November 1999. 3. Sati M, Sta¨ ubli HU, Bourquin Y, Kunz M, Ka¨ sermann S, Nolte LP. Clinical integration of new computer assisted technology for arthroscopic ACL replacement. Oper Tech Orthop 2000; 10:40-49. 4. Yaru NC, Daniel DM, Penner D. The effect of tibial attachment site on graft impingement in an anterior cruciate ligament reconstruction. Am J Sports Med 1992;20:217-220. 5. Friedman RL, Feagin JA Jr. Topographical anatomy of the intercondylar roof: A pilot study. Clin Orthop Rel Res 1994; 306:163-170. 6. Morgan CD, Kalman VR, Grawl DM. Definitive landmarks for reproducible tibial tunnel placement in anterior cruciate ligament reconstruction. Arthroscopy 1995;11:275-288.
973
7. Howell SM, Taylor MA. Failure of reconstruction of the anterior cruciate ligament due to impingement by the intercondylar roof. J Bone Joint Surg Am 1993;75:1044-1055. 8. Vergis A, Gillquist J. Graft failure in intra-articular anterior cruciate ligament reconstructions: A review of the literature. Arthroscopy 1995;11:312-321. 9. Harner CD, Giffin JR, Dunteman RC, Annunziata CC, Friedman MJ. Evaluation and treatment of recurrent instability after anterior cruciate ligament reconstruction. J Bone Joint Surg Am 2000;82:1652-1664. 10. Buzzi R, Zaccherotti G, Giron F, Aglietti P. The relationship between the intercondylar roof and the tibial plateau with the knee in extension: Relevance for tibial tunnel placement in anterior cruciate ligament reconstruction. Arthroscopy 1999; 15:625-631. 11. Jalliard R, Lavallee S, Dessenne V. Computer assisted reconstruction of the anterior cruciate ligament. Clin Orthop Rel Res 1998;354:57-64. 12. Sommer C, Friederich NF, Muller W. Improperly placed anterior cruciate ligament grafts: correlation between radiological parameters and clinical results. Knee Surg Sports Traumatol Arthrosc 2000;8:207-213. 13. Kohn D, Busche T, Carls T. Drill hole position in endoscopic anterior cruciate ligament reconstruction: Results of an advanced arthroscopy course. Knee Surg Sports Traumatol Arthrosc 1998;6:13-15 (suppl 1). 14. Desenne V, Lavallee S, Julliard R, Orti R, Martelli S, Cinquin P. Computer-assisted knee anterior cruciate ligament reconstruction: First clinical tests. J Image Guided Surg 1995;1:59-64. 15. Klos TV, Habets RJE, Banks AZ, Banks SA, DeVilee RJJ, Cook FF. Computer assistance in arthroscopic anterior cruciate ligament reconstruction. Clin Orthop Rel Res 1998;354:65-69. 16. Sati M, de Guise JA, Drouin G. Computer assisted knee surgery: Diagnostics and planning of knee surgery. Comput Aided Surg 1997;2:108-123. 17. Gotzen L, Schierl M, Pashminch Azar A, Kober R, Heeckt PF, Petermann J. The CASPAR-system (computer assisted surgical planning and robotics) in the reconstruction of the ACL. Presented at the 9th Annual ESSKA Meeting, London, September 2000. 18. Petermann J, Kober R, Heinze P. Computer-assisted planning and robot-assisted surgery in anterior cruciate ligament reconstruction. Oper Tech Orthop September 2000;10. 19. Bernard M. robot-assisted tunnel placement in ACL surgery. 9th Annual ESSKA Meeting, London, 2000. 20. Beckwith T, Buck N, Marangoni R. Treatment of uncertainties. In: A WP, ed. Mechanical measurements. 1982. 21. Burkart A, Debski RE, McMahon PJ, et al. A comparison of precision for ACL tunnel placement using traditional and robotic technique. Comput Aided Surg 2001;6:270-278. 22. Bernard M, Hertal P, Hornung H, Cierpinski T. Femoral insertion of the ACL: Radiographic quadrant method. Am J Knee Surg 1997;10:14-21 (discussion 21-22). 23. Amis AA, Jakob RP. Anterior cruciate ligament graft positioning, tensioning and twisting. Knee Surg Sports Traumatol Arthrosc 1998;6:2-12 (suppl 1). 24. Tonet O, Megali G, D’Attanasio S, et al. An augmented realitiy navigation system for comptuer assisted arthroscopic surgery of the knee. Computer Assisted Orthopaedic Surgery USA, Pittsburgh, PA, 2000. 25. Petermann J, Schierl M, Niess C, Kober R, Heinze P, Gotzen L. Computerassistierte Planung und roboterassistierte Ersatzplastik des vorderen Kreuzbandes mit dem CASPAR-System. Arthroskopie 2000;13:270-279. 26. Musahl V, Burkart A, Fu F. Tunnelplatzierung: Definition der Anatomischen Topographie. In: Deutsche Gesellschaft fuer Orthopaedie und Traumatologie. Berlin, Germany, October 2001.