Dunbar 2012 Accuracy of Dynamic Tactile Guided UKA .pdf
Original filename: Dunbar 2012- Accuracy of Dynamic Tactile Guided UKA.pdf
Title: Accuracy of Dynamic Tactile-Guided Unicompartmental Knee Arthroplasty
Author: Nicholas J. Dunbar BSc
This PDF 1.7 document has been generated by Elsevier / Acrobat Distiller 8.1.0 (Windows), and has been sent on pdf-archive.com on 07/06/2017 at 15:59, from IP address 162.1.x.x.
The current document download page has been viewed 521 times.
File size: 610 KB (7 pages).
Privacy: public file
Download original PDF file
The Journal of Arthroplasty Vol. 27 No. 5 2012
Accuracy of Dynamic Tactile-Guided
Unicompartmental Knee Arthroplasty
Nicholas J. Dunbar, BSc,* Martin W. Roche, MD,y Brian H. Park, BSc,*
Sharon H. Branch, BSc,z Michael A. Conditt, PhD,z and Scott A. Banks, PhD*
Abstract: Unicompartmental knee arthroplasty (UKA) can achieve excellent clinical and
functional results for patients having single-compartment osteoarthritis. However, UKA is
considered to be technically challenging to perform, and malalignment of implant components
significantly contributes to UKA failures. It has been shown that surgical navigation and tactile
robotics could be used to provide very accurate component placement when the bones were
rigidly fixed in a stereotactic frame during preparation. The purpose of this investigation was
to determine the clinically realized accuracy of UKA component placement using surgical
navigation and tactile robotics when the bones are free to move. A group of 20 knees receiving
medial UKA with dynamically referenced tactile-robotic assistance was studied. Implant
placement errors were comparable with those achieved using tactile robotics with rigid stereotactic
fixation. Keywords: unicompartmental knee arthroplasty, surgical navigation, robotics, tactile
© 2012 Elsevier Inc. All rights reserved.
Unicompartmental knee arthroplasty (UKA) is a successful treatment for managing anteromedial osteoarthritis. Early UKA techniques relied heavily on surgical
judgment while manually placing cutting blocks and
extramedullary guides to align the components to the
bone surfaces. Clinical studies pertaining to these
surgical techniques suggest that there have been
significant improvements over the last 30 years ;
early studies reported satisfactory clinical outcomes at
more than 2 to 5 years of 50% to 70% [2-5], whereas
more recent studies reported success rates of 90% or
higher [6-10]. Higher success rates have been attributed
to strict patient selection criteria, the development of
intramedullary guides for condylar component positioning and other advances in surgical technique, improved
implant designs, better polymer material processing, and
other factors. Several authors have suggested that
patient recovery time also has decreased with the
From the *Department of Mechanical & Aerospace Engineering, University
of Florida, Gainesville, Florida; yDepartment of Orthopaedic Surgery, Holy
Cross Hospital, Fort Lauderdale, Florida; and zMako Surgical Corp, Fort
Supplementary material available at www.arthroplastyjournal.org.
Submitted January 25, 2011; accepted September 16, 2011.
The Conflict of Interest statement associated with this article can be
found at doi:10.1016/j.arth.2011.09.021.
Reprint requests: Scott A. Banks, PhD, Department of Mechanical &
Aerospace Engineering, University of Florida, MAE-A Room 318,
Gainesville, FL 32611-6250.
© 2012 Elsevier Inc. All rights reserved.
introduction of minimally invasive and quadricepssparing surgical techniques [11,12].
These advances are not without consequences. Minimally invasive techniques reduce joint visualization
during surgery with the potential to compromise implant
alignment accuracy [1,6,13-15]. Achieving accurate component positioning with minimally invasive technique is
technically demanding, and it has been shown that improper limb or implant alignment is the primary cause of early
UKA failure associated with aseptic implant loosening
[9,16-18], excessive polyethylene wear [16,19], and disease progression in the noninvolved compartment [9,16].
Computer-assisted navigation and tactile-robotic assistance recently have been demonstrated to provide
accurate implant positioning. Tactile robots permit
surgeons freely to remove bone within a preplanned
cutting volume and prohibit the surgeon from removing
bone outside that volume. In essence, tactile-robotic
tools act as virtual cutting guides or templates. Cobb et al
 first reported a prospective comparison of a tactileguided, robot-assisted UKA with conventional UKA
performed with manual instrumentation. Their robotic
system used static referencing that required rigid
intraoperative fixation of the femur and tibia to a
stereotactic frame. Implant position errors relative to
the planned position averaged 1.1 mm and 2.5° with
robotic assistance compared with 2.2 mm and 5.5°
conventionally along any axis. Overall tibiofemoral
coronal plane alignment was within 2° for every case
performed with robotic assistance.
804 The Journal of Arthroplasty Vol. 27 No. 5 May 2012
Fig. 1. The dynamically referenced tactile-guided robotic
system uses reflective marker arrays to monitor the instantaneous position and orientation of the femur and tibia during
surgery (shown during a cadaver surgical procedure).
Dynamic-referencing tactile-guidance robotic systems
recently have been introduced for performing UKA 
(Fig. 1). These systems use optical motion capture
technology to dynamically track marker arrays fixed to
the femur and tibia, allowing the surgeon to freely adjust
limb position and orientation during tactile-guided bone
cutting. This reduces setup time and complexity,
minimizes the required tissue exposure, and allows the
surgeon more normally to perform the procedure .
However, dynamic referencing adds sources of measurement uncertainty, and it has not been proven
clinically that a dynamically referenced procedure can
be performed with the same component positioning
accuracy as a statically referenced procedure. Thus, the
goal of this project was to determine if tactile-guided,
robot-assisted UKA with dynamic referencing achieves
similar in vivo accuracy results to Cobb et al , who
used statically referenced robotic procedures.
Materials and Methods
We retrospectively collected data on the first 50
patients who received medial UKA with robot-assisted
bone preparation with a dynamically referenced system
(TGS; MAKO Surgical Corp, Fort Lauderdale, Fla) at 1
center. This was the first system used clinically, and a
single surgeon (M. R.) performed all cases. Each patient
received a traditional cobalt-chrome femoral implant
with an all-polyethylene tibial inlay component (Unicondylar Knee System; StelKast, Inc, McMurray, Penn).
Computer-assisted preoperative planning of implant
positioning was based on patient-specific 3-dimensional
(3D) computed tomographic (CT) scans. The surgeon
approved the plan for implant alignment before surgery
and maintained the ability to change the plan at any
time during surgery. Tibial and femoral tracking arrays
were placed with intracortical pins, and the articular
surfaces were probed to register the preoperative plan
with the surgical anatomy. After osteophyte removal,
dynamic measurement of tibiofemoral motion permitted
quantitative fine-tuning of implant component placement to achieve appropriate and uniform joint laxity
over the flexion arc. Bone removal was performed using
graphical and tactile feedback with a virtually constrained robotic arm equipped with a high-speed burr.
An optical probe was used to confirm that cut bone
surfaces conformed to the surgical plan and that trial
implants were placed in the desired 3D locations. After
these checks, the definitive implants were placed with
cement. The operative time (skin-to-skin) averaged 93
minutes (SD, 15 minutes).
Institutional review board approval was obtained
prospectively to collect deidentified information for all
patients. From the 50 knee group:
• preoperative and postoperative clinical records
were available for all 50 knees,
• preoperative CT scans were available for all 50 knees,
• postoperative CT scans were available for 24 knees
because many patients declined to have postoperative scans and some patients mistakenly were not
prescribed postoperative scans, and
• surgical plans were available for 35 cases.
Of the 24 cases with postoperative CT scans, 4 cases
had surgical plans that were used for teaching or
demonstrations postoperatively, erasing the actual
intraoperative plan. Thus, complete data were available
for only 20 knees (19 patients, 1 bilateral). Of these
patients, 9 were male and 10 were female with an
average age of 71 years (range, 49-92 years), an average
height of 168 cm (range, 155-185 cm), an average
weight of 80 kg (range, 44-111 kg), and an average body
mass index of 28 ± 5 kg/m 2. Preoperative range of
motion and Knee Society Scores (pain/function) averaged 119° and 39/60 points, respectively (Table 1).
Cobb et al  used 2-dimensional projections and
registration of the preoperative and postoperative 3D
CT scans to quantify the 3D implant placement errors.
The details of our measurement technique differ but
produce the same measured parameters. We compared
Table 1. Clinical Range of Motion (ROM) and Knee Society
Scores for Knees with UKA After Dynamically Referenced
119 ± 10
97 ± 14
126 ± 8
130 ± 6
127 ± 4
127 ± 4
125 ± 8
39 ± 10
74 ± 13
82 ± 9
83 ± 9
85 ± 10
87 ± 4
88 ± 10
60 ± 11
63 ± 5
71 ± 17
71 ± 21
75 ± 16
78 ± 13
75 ± 16
99 ± 17
137 ± 16
153 ± 21
154 ± 26
161 ± 21
165 ± 14
163 ± 17
Accuracy of Robotic UKA Dunbar et al
Fig. 2. Implants were segmented from the postoperative CT scans, and their 3D positions were compared with the preoperatively
planned positions. From left to right, it can be seen how the segmented CT volume is replaced with a model of the implant
component, and then the difference in positions of the implant models is assessed (dark vs light implants in far right image).
3D preoperative plans and 3D postoperative CT scans
directly to quantify the difference between the 3D
planned implant positions and the actual postoperative
positions of the tibial and femoral components (Fig. 2).
The CT scan protocol included 200 slices taken through
the knee (approximately 10 cm above and below the
knee center) with a maximum slice thickness of 1 mm.
Combined bone and implant 3D models were created by
segmenting the CT images using active contour models
with image intensity threshold filtering (ITK-SNAP,
www.itksnap.org). Some manual editing was required
during segmentation to correct contours for streaking
artifacts from the metallic implants.
The placement errors of the implanted components
relative to their planned positions were determined for
each tibial and femoral component using global modelbased 3D registration. Registration was performed
using a 3D-to-3D iterative closest point algorithm
(Geomagic Studio; Geomagic, Inc, Morrisville, NC).
Bone models from the preoperative and postoperative
scans were registered first to establish a common
reference frame. The position and orientation of the
Fig. 3. Implant placement errors can be visualized by superimposing the planned implant position (light) and the actual
implant position (dark). The top row of images of the femoral (left) and tibial (right) components show the cases with the smallest
errors, the second row shows cases with average placement errors, and the bottom row shows the cases with the highest
806 The Journal of Arthroplasty Vol. 27 No. 5 May 2012
Table 2. Root Mean Square Implant Placement Errors Comparing the Preoperative Ideal Placement and the Surgically
Internal/external rotation (deg)
Cobb et al—Manual Technique
Cobb et al—Robotic Technique
surgically placed implants were then measured relative
to the planned implant positions and orientations (Fig.
2). All displacements were quantified using 4 × 4
homogeneous transformation matrices, from which
translations and rotations were expressed in standard
anatomical parameters. Root mean square (RMS)
errors were used to quantify average alignment
accuracy and dispersion.
The CT-based implant position measurements were
validated by comparison with identical measures from
3D laser scans of 5 plastic knee specimens implanted
with femoral and tibial unicondylar components (Fig. 3).
The same surgeon performed dynamic tactile-guided
UKA on each set of plastic bones, and postoperative CT
scans and laser scans were obtained. Implant position
measurements relative to the bone were compared
using global model-based 3D registration.
The validation study showed that RMS differences
between CT-based and laser scan implant position
measurements were within 0.8 mm and 0.9° and 0.9
mm and 1.7° in all directions for the femoral and tibial
Postoperative range of motion and Knee Society
Scores at 3-year follow-up averaged 125° and 88/75,
respectively (Table 1).
Surgical RMS errors for femoral component placement were within 1.6 mm and 3.0° in all directions of
the planned implant position, respectively (Table 2,
Figs. 3 and 4; Fig. 4 is available online at www.
arthroplastyjournal.org). Average RMS errors for tibial
component placement were within 1.6 mm and 3.0° in
all directions. Tibial and femoral component RMS
varus/valgus errors were 1.5° and 2.6°, respectively.
Average tibial posterior slope was within 1.9° relative to
the planned position.
Poorly aligned UKA components can lead to aseptic
loosening, excessive polyethylene wear, and disease
progression to the noninvolved compartment
[9,16,18,19]. Robot-assisted surgical techniques have
been shown to improve implant placement after UKA
compared with conventional manual techniques by
improving bone preparation accuracy [20,23]. Dynamic-referencing techniques, using motion capture
and reflective markers fixed to the bones, allow for
free bone movement during surgery. The purpose of
this study was to evaluate the clinically achieved
accuracy of robot-assisted positioning of UKA components using dynamic bone tracking and tactile
guidance of the cutting burr. We found that implant
positioning errors using dynamic bone tracking and
tactile robotics were very similar to placement errors
previously reported for statically referenced robotassisted UKA techniques.
The study sample of 20 knees was drawn from a
single-surgeon's initial clinical series of 50 medial UKA
with robotic-assisted bone preparation. This was the first
clinical series of patients using the dynamically referenced tactile-guided robotic system, and our 40%
inclusion rate reflects several unforeseen difficulties.
Our analysis required preoperative and postoperative
CT scans and the computer surgical plan, and these
complete data sets were available only for 20 of the first
50 patients. Many patients declined to have postoperative CT scans or, mistakenly, were not prescribed
postoperative scans. The robotic software was revised
and updated several times during the series, and the
surgical system frequently was used for teaching and
demonstrations. The result was that valid surgical
plans could be retrieved for only 20 of the 24 subjects
who had postoperative CT scans. Our 20 knee study
cohort includes all patients with complete data, but this
may represent a biased sample compared with the 50
The reported implant positioning errors represent
the sum of all errors associated with the surgical and
data measurement processes, including bone preparation errors, implant placement and cementing errors,
and errors associated with the CT-based 3D measurements. Thus, the reported implant position errors are
conservative upper bounds of the actual errors due
only to the robot-assisted preparation of bone. The
surgeon's learning curve was not investigated within
this sample but may have influenced the results. Future
studies will have to address the possibility that tactileguided robot technology might reduce or eliminate
Accuracy of Robotic UKA Dunbar et al
the effect of surgeon experience on the accuracy of
The implant accuracy results generally were within
3 mm or 3° of the preoperative planned positions (Fig. 4;
available online at www.arthroplastyjournal.org). Five
knees had individual femoral component placement
errors greater than 4°, and no knees had translation
errors greater than 4 mm. There were also 5 knees with
tibial components having single-degree-of-freedom errors greater than 4°, with all 5 knees showing greater
than 4° of axial rotation error in the placement of the
tibial component. Importantly, none of the tibial implants
showed varus/valgus placement errors greater than 3°.
Our results with dynamically referenced robotics can
be compared directly with the data from Cobb et al 
for traditional manual and statically referenced robotic
surgery. The robot-assisted implants from our study
were more accurately placed compared with the
traditional manual implants of Cobb et al for each
translation and rotation (Table 2). On average, the
accuracy of dynamically referenced and statically referenced implant placement differed by 0.4 mm/0.4° or less
(Table 1). Qualitatively, it appears that the femoral
component proximal/distal placement errors were
slightly greater for the dynamically referenced implants
(1.2 mm) than statically referenced implants (0.6 mm).
Proximal-distal placement is strongly influenced by
the cement layer thickness, a variable unaccounted for
by either our analyses or that of Cobb et al and perhaps an issue deserving further study. These findings
suggest that the average difference in implant placement
accuracy between dynamically referenced and statically
referenced tactile-guided UKA surgery is less than
0.5 mm or 0.5°.
Several groups have reported their results for computer-navigated UKA using manual bone cutting instruments. It is relevant to review these data to
understand if the robotic approach provides any
accuracy benefit vs manual instruments with navigation. Jenny et al  reported navigated UKA with both
traditional and minimally invasive approaches. They
defined 5 measures for nominal placement of the components as 0° to 5° coronal angle between the femoral
and tibial components and within 3° of specific sagittal
and coronal plane targets for the femoral and tibial
components. Jenny et al  reported that 60% of
knees with a traditional navigated approach and 62%
of knees with a minimally invasive navigated approach
satisfied all 5 criteria. Seventy-five percent of our
femoral components and 90% of our tibial components
would have satisfied the alignment success criteria of
Jenny et al, which did not include targets for axial
rotation. Seon et al  reported a similar analysis
for navigated minimally invasive UKA with similar
success criteria. They reported 87% and 81% successful femoral coronal and sagittal alignment and 94% and
87% successful tibial coronal and sagittal alignment,
respectively. Our results, for comparison, are 85%
and 80% successful femoral coronal and sagittal alignment and 100% and 90% successful tibial component
coronal and sagittal alignment, respectively. It would
seem fair to conclude that UKA with dynamically
referenced tactile robotics is at least as accurate as
computer-navigated minimally invasive UKA with
Robotic preparation of bone is only 1 factor influencing placement accuracy for cemented arthroplasty
components. After the cavity is robotically prepared,
cement is introduced and the implant is seated, usually
with poorly controlled loads or alignment applied during
cement curing. Bone pockets for inset tibial components
are necessarily larger than the implant to provide a
roughly 2-mm circumferential cement mantle. These
oversized pockets and uncontrolled placement of implant alignment and load can easily result in several
degrees and several millimeters of implant placement
variation. Ongoing efforts to include implant alignment
features within the bone pockets, to control the
placement and loading of implants during cement
curing, and to directly monitor and control component
placement have the possibility to further improve
implant placement accuracy in conjunction with robotic
Dynamically referenced tactile robotics provides a new
tool to accurately prepare bone with minimally invasive
approaches. Our results suggest that excellent UKA
implant placement accuracy can be achieved, comparable with that demonstrated for statically referenced
tactile robotics. The study cohort was drawn from the
first group of patients operated on by a single surgeon
using this technique, suggesting that good implant
alignment is achieved in what normally would be
considered a learning phase. Finally, these patients
were treated with the first approved version of this
new tool. One can reasonably expect that further
refinement of this robotic technology will enhance the
accuracy and usability of this tool.
1. Bert JM. Unicompartmental knee replacement. Orthop
Clin N Am 2005;36:513.
2. Cameron HU, Hunter GA, Welsh RP, et al. Uncompartmental knee replacement. Clin Orthop Relat Res 1981;
3. Insall J, Aglietti P. A five to seven-year follow-up of
unicondylar arthroplasty. J Bone Joint Surg Am 1980;62:
4. Laskin RS. Unicompartmental tibiofemoral resurfacing
arthroplasty. J Bone Joint Surg Am 1978;60:182.
5. Mallory TH, Danyi J. Unicompartmental total knee
arthroplasty: a five to nine year follow-up study of 42
procedures. Clin Orthop 1983;175:135.
808 The Journal of Arthroplasty Vol. 27 No. 5 May 2012
6. Hamilton WG, Ammeen D, Engh CA, et al. Learning curve
with minimally invasive unicompartmental knee arthroplasty. J Arthroplasty 2009.
7. Keene G, Forster MC. Modern unicompartmental knee
replacement. Curr Orthop 2005;19:428.
8. Lustig S, Paillot JL, Servien E, et al. Cemented all
polyethylene tibial insert unicompartimental knee arthroplasty: a long term follow-up study. Orthop Trauma: Surg
9. Saenz CL, McGrath MS, Marker DR, et al. Early failure of
a unicompartmental knee arthroplasty design with an
all-polyethylene tibial component. Knee 2010;17:53.
10. Voss F, Sheinkop MB, Galante JO, et al. Miller-Galante
unicompartmental knee arthroplasty at 2-to 5-year follow-up evaluations. J Arthroplasty 1995;10:6.
11. Argenson JA, Flecher X. Minimally invasive unicompartmental knee arthroplasty. Knee 2004;11:341.
12. Hamilton WG, Collier MB, Tarabee E, et al. Incidence and
reasons for reoperation after minimally invasive unicompartmental knee arthroplasty. J Arthroplasty 2006;
21(6 Suppl 2).
13. Cossey AJ, Spriggins AJ. The use of computer-assisted
surgical navigation to prevent malalignment in unicompartmental knee arthroplasty. J Arthroplasty 2005;
14. Fisher DA, Watts M, Davis KE. Implant position in knee
surgery: a comparison of minimally invasive, open
unicompartmental, and total knee arthroplasty. J Arthroplasty 2003;18(7 Suppl 1).
15. Muller PE, Pellengahr C, Witt M, et al. Inﬂuence of
minimally invasive surgery on implant positioning and the
functional outcome for medial unicompartmental knee
arthroplasty. J Arthroplasty 2004;19:3.
16. Aleto TJ, Berend ME, Ritter MA, et al. Early failure of
unicompartmental knee arthroplasty leading to revision.
J Arthroplasty 2008;23:2.
17. Riebel GD, Werner FW, Ayers DC, et al. Early failure of the
femoral component in unicompartmental knee arthroplasty. J Arthroplasty 1995;10:5.
18. Schai PA, Suh JT, Thornhill TS, et al. Unicompartmental
knee arthroplasty in middle-aged patients: a 2- to 6-year
follow-up evaluation. J Arthroplasty 1998;13:4.
19. Collier WB, Engh CA, McAuley JP, et al. Factors
associated with the loss of thickness of polyethylene
tibial bearings after knee arthroplasty. J Bone Joint Surg
20. Cobb J, Henckel J, Gomes P, et al. Hands-on robotic
unicompartmental knee replacement: a prospective, randomised controlled study of the acrobot system. J Bone
Joint Surg Br 2006;882.
21. Conditt MA, Roche MW. Minimally invasive robotic-armguided unicompartmental knee arthroplasty. J Bone Joint
Surg Am 2009;91:63.
22. Pearle AD, O'Loughlin PF, Kendoff DO. Robot-assisted
unicompartmental knee arthroplasty. J Arthroplasty 2010;
23. Lonner JH, John TK, Conditt TA. Robotic arm–assisted
UKA improves tibial component alignment: a pilot study.
Clin Orthop Relat Res 2010;468:141.
24. Jenny JY, Ciobanu E, Boeri C. The rationale for navigated
minimally invasive unicompartmental knee replacement.
Clin Orthop Relat Res 2007;463:58.
25. Seon JK, Song EK, Park SJ, et al. Comparison of
minimally invasive unicompartmental knee arthroplasty
with or without a navigation system. J Arthroplasty
Accuracy of Robotic UKA Dunbar et al
Fig. 4. Distribution of implant placement errors for the femur (left column) and tibia (right column) for varus/valgus,
flexion/extension, and internal/external rotations (A) and for medial/lateral, anterior/posterior, and proximal/distal positions (B).
The dynamically referenced, robot-assisted data (gray, from present study) are compared with the reported accuracy figures from
Cobb et al  for fixed reference, robot-assisted (black) and traditionally instrumented manual (white) procedures.