- Email: [email protected]
Development of a motion-controlled in vitro elbow testing system
ELSEVIER
Journal of Orthopaedic Research 21 (2003) 4 0 5 4 1 1
Journal of Orthopaedic Research www.elsevier.comllocate1orthres
Development of a motion-controlled in vitro elbow testing system Cynthia E. Dunning
Karen D. Gordon Graham J.W. King James A. Johnson a,b3c,d,e,*
aib,d,
a,d,e,
Bioengineering Researcli Luhoratorj3 Lawson Health Restlurch Institute, Hond & Upper Limb Centre, St. Josc~ph'sHeulth Cure London, London, Ontario, Cunadu Department of Mc4iaiiical and Matc,riuls Engineering, The Uniuersit.v oj Western Ontario, Lon(lon, Ontario, Cunudo Department uf' Biomedical Engineering, The tiniaersiry .f Western Ontario. London. Onturio, Cumdo Dtyurirrient of' Surgery, The Uniuwsiiy of' Uresisl(wiOniurio, London. Ofiturio, Cunudu Departnient of Mrdicul Biophysics, The tiiiiwrsiiy of Western Ontario, London, Onturio, Cmcrdu
Abstract
Joint simulators can be used to study motion pathways of a human joint, to investigate changes in joint stability following injury, and to formulate improved reconstructive and rehabilitative procedures. Our objectives were: to develop a laboratory-based, motion-controlled elbow testing apparatus capable of simulating tendon (muscle) loading and displacement in a cadaveric specimen; to describe its performance while testing stable and unstable elbows; and to compare its operation to that of a previously designed load-controlled device. Velocity control of a pneumatic actuator was achieved using a custom-written, closed-loop feedback controller. This actuator was incorporated into an elbow testing system that used additional pneumatic actuators and a combination of motion- and load-control to achieve desired motions. Simulations achieved with this apparatus demonstrated small magnitudes of error in actuator position and highly repeatable flexion pathways with the specimens positioned in vertical, varus, and valgus orientations. The repeatability i n motion pathways generated in both a stable and unstable elbow model was equivalent to or better than for similar tests performed using the load-controlled system, and the velocity of the resulting elbow motion was more reproducible. 0 2003 Orthopaedic Research Society. Published by Elsevier Science Ltd. All rights reserved.
Introduction In vitro testing in the laboratory plays a n important role in advancing patient care by helping t o develop new treatment techniques a n d improved rehabilitation protocols. Due to a lack of adequate theoretical models, these investigations are important in elucidating injury mechanisms. Moreover, these studies should be useful in validating new surgical and rehabilitative treatments prior to their clinical implementation. A number of laboratory-based systems capable of simulating active elbow motion in vitro have been designed. Some of these devices are load-controlled, where specific force magnitudes are applied directly to bones o r tendons to produce joint loading a n d motion. Johnson
* Corresponding author. Address: Bioengineering Research Laboratory, Lawson Health Research Institute. Hand & Upper Limb Centre, St. Joseph's Health Care London, London. Ontario. Canada. Tel.: +1-519-646-601 I : fax: +I-519-646-6049. E-rnriil citflrrrss: [email protected] (J.A. Johnson).
et al. [lo] described a loading device incorporating pneumatic actuators to apply sufficient loads t o relevant tendons to produce unassisted elbow flexion-extension and forearm pronation-supination with the arm in the vertical orientation. Other elbow simulators are motioncontrolled devices, since they achieve simulated joint movement by controlling bone o r tendon displacement using a motor o r actuator. Bottlang et al. [3,4] and Madey et al. [12] described devices that used DC motors connected directly to the bone(s) of the forearm to move the elbow through the flexion-extension cycle. No motion-controlled testing device currently exists to produce elbow motion through tendon loading. Past experience in o u r laboratory has shown that a loadcontrolled elbow testing system is a practical approach to simulate joint flexion [lo]. In some joint positions, however, inertial effects can make it very difficult to simulate smooth, gentle joint motions using only loadcontrol. F o r example, in the varus orientation (i.e. the humerus is fixed in a horizontal position with the medial aspect of the elbow inferior), the carrying angle formed
0736-0266/03/$ - see front matter 0 2003 Orthopaedic Research Society. Published by Elsevier Science Ltd. All rights reserved doi:l0.10161S0736-0266(02)00233-4
406
c‘ E Dunning et ul I Journal of Orthopuedc RcJseuvch21 (2003) 405-411
between the ulna and humerus tends to produce spontaneous, rapid elbow flexion even with very small magnitudes of load applied to the elbow flexor muscles. In addition, the lack of adequate and predictable resistance during loading challenges load-controlled motion. The use of motion-controlled testing systems can help minimize these effects by maintaining tendon excursion (and thus joint rotation) at a constant velocity as opposed to applying a prescribed tendon force. Therefore, the purpose of this work was threefold: to design and develop a motion-controlled testing system to produce elbow motion; to evaluate the performance of this system in both the stable and unstable elbows; and to compare the elbow motion pathways it generates to those achieved using a load-controlled testing system.
PNEUMATIC ACTUATORS IN-LINE LOAD CELL
BASE PLATE
SECONDARY ALIGNMENT SYSTEM
CABLES TRANSMITTER ALIGNMENT RING
Methods The motion-controlled elbow simulator (Fig. 1) is machined from polyethylene, Delrin“, and 3 16L stainless steel to avoid interference with the electromagnetic tracking device [14], whose receivers are secured to the specimen (see below) and whose transmitter is secured to the base plate of the simulator. A cadaveric specimen is attached to the device via the humeral mounting clamp. Cables that have been sutured to five tendons of interest (biceps, brachialis, brachioradialis, triceps, and pronator teres) are each connected to a dedicated actuator. This simulator incorporates a two-part alignment system to control the lines-of-action of the cables. The primary alignment system consists of two separate units: an alignment guide (located within the humeral clamp), and an alignment ring (located near the transmitter). Both units contain pulley mechanisms that incorporate a two degree-offreedom (2 DOF) design, which enables them to accommodate changes in the lines-of-action that occur as the elbow moves through a range of motion. That is, these mechanisms are free to rotate about the axis of the approaching cable. As the angle of approach changes, they rotate to ensure their pulleys remain in contact with the cable. Cables attached to the brachioradialis and pronator teres pass through the alignment ring on route to their respective actuators. Moment arms for the cables attached to the biceps (3.3 cm), brachialis (2.0 cm), and triceps (1.6 cm) are maintained by the alignment guide, positioned approximately 4 cm proximal to the elbow joint [2]. A secondary alignment system, which uses simple pulleys, ensures that each Cable is in line with the piston of its dedicated actuator. The height of each pulley in the secondary alignment system can be adjusted to match the height of its actuator. The entire base plate is mounted o n a polyethylene stand via a 2 D O F hinge. This allows the simulator to be rotated into varus and valgns orientations (i.e. the humerus is fixed in a horizontal position with the medial aspect of the elbow inferior or superior, respectively), as well as into the vertical position (i.e. the humerus is vertical and the forearm is free to move through an arc of flexion) [8]. To initiate motion-control, the tendon of the muscle that supplies the largest percentage of the total applied load responsible for producing the motion is designated as the “prime mover”. The percentage depends upon the muscle-loading ratio, which is developed from EMG and pCSA data as has been previously reported [8]; two of these ratios are listed in Table 1. Using the biceps (BI) loading ratio, the biceps is the prime mover for both pronated and supinated flexion. However, the brachialis muscle is the prime mover according to the brachialis (BR) loading ratio. Therefore, both the brachiah and biceps are connected to pneumatic actuators (model PFC-096-XLP, BIMBA Ltd., Cambridgeshire, UK) that contain an internal linear resistive transducer (LRT) to provide position feedback as required for motion control. These position feedback actuators can also be operated under load-control. The remaining 3 tendons are connected to actuators (model 096-DX, BIMBA L.td., Cambridgeshire, UK) that operate under load-control only (Le. do not contain an LRT).
PRIMARY
- ALIGNMENT
HUMERAL MOUNTING CLAMP
SYSTEM
ALIGNMENT GUIDE RECEIVER ---#m~
i-.*‘“
Fig. 1. The motion-controlled elbow simulator is shown with the base plate in the vertical position. A universal hinge attached to the base plate also allows the simulator to be placed in varus and valgus orientations, where the weight of the forearm applies a provocative moment about the elbow joint. A cadaveric specimen is attached to the device via a humeral mounting clamp. Cables, sutured to five tendons of interest, are connected to independent pneumatic actuators. Primary and secondary alignment systems that contain adjustable pulleys are used to maintain proper (physiologic) lines-of-action for each cables. An electromagnetic tracking device, consisting of a receiver fixed to the ulna and a transmitter secured to the base plate, is used to measure joint motion.
Control of each actuator is achieved using a custom-written LabVIEW software program (Fig. 2). A key feature of this code is a digital proportional-integral-derivative (PID) controller used within a realtime feedback control loop to move the piston of the prime mover actuator at a constant velocity (i.e. controls the rate of tendon excursion during the flexion arc). The controller acts to reduce or eliminate error between the actuator’s current position (i.e. the process variable) to its desired position (i.e. the set point). Controller output is a voltage signal sent to a dedicated proportional-pressure-controller (PPC) (model #PPC5C-AAA-AGCB-BBB-JD, MAC Valves, Wixon, MI), that adjusts the pressure of the air supplied to the cylinder as required to achieve the desired actuator position. The controlling software also receives input from a custom-built, strain gauge based, 1 D O F load cell used to monitor the force applied by the prime mover at 200 Hz. This signal is used to apportion load to the load-controlled actuators in accordance with the loading ratio selected; i.e., the controlling software sends appropriate voltage signals to independent PPCs connected to the each of the simulator’s load-controlled actuators to simulate muscle tension. Testing can be performed with the forearm maintained in either supination or pronation. This is accomplished by initially activating the appropriate tendons using a load-controlled protocol, as previously described [8,10]. For example, to position the arm in supination, load is applied to the biceps tendon using a ramp function over a 5 s period. Simultaneously, a ramp load is applied to the triceps to keep
C.E. Dunning et al. I Journal of Orthopaedic Research 21 (2003 j 405 4 1 I
407
Table 1 Loading ratios for simulated active elbow flexion Motion
Loading ratio
Biceps (%)
Brachialis (YO)
Brachioradalis (%I)
Triceps ((XI)
Supinated flexion
BI BR
26 31
21 42
9 I1
44 16
Pronated flexion
BI BR
25 22
21 41
8 16
46 21
Note: BI = biceps loading ratio; BR = brachialis loading ratio. B1 loading ratio: E M G data from Caldwell and Van Leemputte [6] and pCSA data from Amis et al. [I]. BR loading ratio: EMG data from Funk et al. [9] and pCSA data from Amis et al. [l].
ELECTROMAGNETIC TRACKING SYSTEM DATA INITIALIZATION USERINPUTS
MOTION DATA
---,, - + OUT P UT
(based on load cell feedback and loadino ratio selected)
DATA FILES
VOLTAGE SIGNAL
VOLTAGE SIGNAL CONTROLLING COMPUTER POSITION
(generated by PID controller based on LRT feedback)
PPC
LOAD CONTROL FEEDBACK REGULATOR
Fig. 2. The main controlling computer operating with cnstom-designed LabVIEW software is a central feature of the motion-controlled simulator. It receives inputs from the user that include PID controller gains, desired actuator velocity, forearm position (supination, pronation), simulator position (vertical, varus, valgus), loading ratio, threshold load values, and load cell calibration data. It also receives feedback information from the load cell and the LRT on the prime mover actuator and uses it to apportion loads to the load-controlled actuators and provide an error signal to the PID controller, respectively. The computer regulates voltage to the PPCs attached to both the prime mover and the load-controlled actuators. In addition, it initializes the electromagnetic tracking device to collect data once the prime mover is activated, and stores the kinematic data, and the position and load values from the prime mover actuator.
the elbow in a fully extended position. The magnitudes of these loads are determined using an iterative procedure to determine the minimum magnitudes of load necessary to achieve both full supination and full extension. After the 5 s, the motion-controlled algorithm is activated to initiate controlled elbow flexion, and the data collection begins. This algorithm ensures that the biceps load will not fall below the iteratively determined threshold value needed to maintain the supinated position. Triceps activation (indicated by the loading ratios in Table 1) is not maintained throughout the flexion arc when simulating motion in the vertical orientation. Rather, once the motion-controlled movement begins, the triceps load is ramped down to 0 N over a 5 s interval. In the varus and valgus orientations, however, the triceps load is maintained at a constant value throughout the flexion arc. A similar procedure was followed for testing in the pronated forearm position, where the pronator teres tendon was initially activated (as opposed to the biceps) and a threshold value of load for this tendon was determined. Experimental testing was conducted to evaluate the performance of the elbow testing system. Eight fresh-frozen cadaveric upper limbs were tested (mean age: 67 19 years; range: 29-85 years) in accor-
*
dance with the rules and regulations of our institution and government regulatory agency. Specimens were thawed to room temperature (23 f 2 “C) prior to testing. Initial specimen preparation involved exposing and dissecting the tendons of the biceps, brachialis, brachioradialis, triceps, and pronator teres. Ticron #2 suture (Sherwood Medical, St. Louis, MO) was secured to each distal tendon using a Bunnell technique. The suture was then tied to a length of stainless steel cable via a nylon washer. Cables from the brachioradialis and pronator teres tendons were routed beneath the skin to the lateral and medial epicondyles, respectively. Holes were drilled through each epicondyle into the humcral canal. Specially designed Delrin@ sleeves were inserted into these holes (to minimize friction), and the cables passed through the sleeves and out the proximal end of the humeral canal. Lightweight Delrin@’pedestals, designed to accept receivers from the ‘Flock of Birds’ (Ascension Technologies, Burlington, VT) electromagnetic tracking device, were secured to the distal radius and ulna using bone screws. Each specimen was secured in the humeral clamp, the tendon cables were fed through the appropriate alignment pulleys and attached to dedicated actuators, and the tracking device receivers were attached to their pedestals. With the simulator placed in the vertical orientation, load-controlled testing was performed to determine the threshold values as described above. Motion-control testing began by performing an iterative procedure to select appropriate PID controller gains with actuator backpressure set to 276 kPa, and the desired piston velocity set to 5 mmls. These magnitudes of pressure and velocity were based on pilot tests conducted to determine the desired rate of controlled elbow motion. The gain values were adjusted to achieve a gentle arc of smooth elbow flexion. Once determined, the gain values remained constant for the duration of testing. The testing protocol was designed to investigate effects of loading ratio, reduced loads to the non-prime mover tendons, simulator orientation, and ligament sectioning on the performance of the motioncontrolled simulator. Each test was repeated five times consecutively to quantify repeatability. With the forearm maintained in supination, simulated active flexion was first performed by designating the brachialis as the prime mover and apportioning load ( i s . load control) to the biceps and brachioradialis according to the BR loading ratio (Table 1). The loads to the load-controlled tendons were then reduced to 1/3 and then 2/3 of the values prescribed by the BR loading ratio. Motions employing the BI loading ratio, which designates the biceps as the prime mover, were also performed. Similar motions were performed with the forearm maintained in pronation. The simulator was then rotated into the valgus orientation (i.e. using the aforementioned 2 D O F hinge), where the BR loading ratio was used to achieve five simulated active flexion motions with the forearm in supination. Elbow flexion with the forearm in pronation was attempted with the simulator in this orientation, but was unsuccessful. Testing in the varus orientation was also performed using the BR loading ratio, but with the forearm maintained in pronation. In this position, elbow flexion with the forearm in supination could not be performed. The threshold values used as limits for the biceps and pronator teres loads in the valgus and varns orientations, respectively, were roughly 50% of those used in the vertical position. An unstable elbow was then created by complete transection of the anterior and posterior joint capsules and the radial and lateral ulnar collateral ligaments (collectively referred to as the lateral collateral ligament or LCL), and the entire testing protocol was repeated. At the completion of testing, each specimen
408
C. E. Dunning et ul. I Journal of Orthopuedic Reseavch 21 (2003) 405-411
was disarticulated at the wrist and elbow to digitize the necessary bony landmarks used to create anatomic bone coordinate systems [ I I]. This allowed motion data collected by the tracking device to be expressed in clinically relevant terms of flexion-extension, varus-valgus angulation, and internal-external rotation according to an Euler Z Y-X analysis. The performance of the motion-controlled simulator was evaluated using three separate parameters. Root-mean-square (RMS) error was used to quantify the deviation between the desired and actual positions attained by the motion-controlled actuator throughout the flexion arc. To quantify the ability of the simulator to repeatedly produce the same flexion position as a function of time, the standard deviation of flexion angle from the five successive trials was calculated at each second throughout the duration of the simulated movement. The maximum standard deviation, regardless of the time at which it occurred, was used as a measure of repeatability. Repeatability of the elbow motion pathways generated from the Euler analysis (varus- valgus angulation and internal-external rotation as a function of flexion angle) was also evaluated using the maximum standard deviation in five trials and compared to the results obtained with the load-controlled simulator [8]. All three sets of data (RMS error, maximum standard deviation in flexion angle versus time, and maximum standard deviation in elbow motion pathways) were analyzed statistically using two-way repeated measure analyses of variance and post-hoc Student-Newman-Keuls tests with 2 = 0.05.
a w 1 v)
a . a
n --
I
BR
113BR
213BR
BI
I
VERTICAL
LEI VALGUS
LOADING MODE
U
a w 1 v)
I . Results
a
0 -
R M S error in actuator position
I
BR
113BR
213BR
BI
1
VERTICAL
Excellent agreement between the desired and actual positions of the motion control actuator was obtained as indicated by the small magnitude of (RMS) errors (Fig. 3). For the 8 specimens tested, the average RMS values calculated in the vertical position ranged from 0.7 mm (intact elbow, supinated forearm, BI loading ratio) to 1.6 mm (LCL cut, pronated forearm, BI loading ratio). Comparing the 1/3, 2/3 and BR ratios, the RMS error with the forearm maintained in supination was significantly greater for the 1/3 ratio than either the 2/3 (p =0.003) or full (p = 0.002) ratios (Fig. 3A). In the pronated forearm position, RMS error was significantly greater for both the 1/3 (p < 0.001) and 2/3 0, = 0.003) compared to the BR ratio (Fig. 3B). No difference in the magnitude of the RMS errors was found when comparing the B1 and BR loading ratios in either supination (p = 0.05) or pronation (p = 0.33). For all of the tests performed in the vertical position, cutting the LCL had no significant effect on the error measurements (p > 0.28), although the magnitudes of the error and their standard deviations generally tended to increase with ligament sectioning. In the varus and valgus orientations, the range in the average RMS error measured was from 0.9 mm (LCL cut, pronated forearm, varus orientation) to 1.5 mm (LCL cut, supinated forearm, valgus orientation). When the BR loading ratio was used in supination, the error was significantly greater in the valgus orientation than the vertical position (p = 0.005). For this subset of data, cutting the LCL also significantly increased the error (p = 0.004). RMS error was significantly larger in the vertical position than the varus orientation when the BR loading ratio was used with the
rn VARUS
LOADING MODE Fig. 3 . The means and standard deviations ( n = 8) of RMS error between the desired and actual positions attained by the motion-controlled actuator are shown for the various loading modes tested in both the intact and LCL deficient elbows with the forearm maintained in (A) supination and (B) pronation.
forearm maintained in pronation (p = 0.006), but there was no effect of ligament sectioning (p = 0.60).
Repeatability of flexion angle versus time In the vertical position, the results were highly repeatable (Fig. 4) with the average maximum standard deviation values as small as 0.9" (intact elbow, supinated forearm, 2/3 BR loading ratio) and never greater than 2.8" (LCL cut, pronated forearm, 1/3 BR loading ratio). Significant differences in repeatability were found when comparing the 1/3, 2/3, and BR loading ratios with the forearm maintained in supination (p = 0.001) (Fig. 4A). Tests performed with both the 113 (p = 0.004) and 213 (p = 0.001) ratios were more repeatable than those using the BR loading ratio. The loading ratio used did not significantly affect repeatability when the forearm was maintained in pronation (p = 0.50) (Fig. 4B). No differences in repeatability were found between the BI and BR loading ratios in either supination (p = 0.94) or pronation (p = 0.81). With the arm in the vertical orientation, the effect of cutting the LCL was significant in supination 0,= 0.01; LCL cut was less repeatable), but not in pronation (p = 0.83). In the varus and valgus orientations, the average maximum standard deviation
C.E. Dunning et ul. I Journal of Orthopuedic Reseurch 21 (2003j 405411
n
u)
3
I
S
BR
'
113BR
2ABR
81 I I
VERTICAL LOADING MODE
I
0
5 -
n
* 2
n -
-
I
BR
113BR 2ABR
I
81
I
VERTICAL
values were larger than those found in the vertical position, ranging from 5.5" (intact, supinated forearm, valgus orientation) to 8.4" (LCL cut, pronated forearm, varus orientation). When the BR loading ratio was employed, placing the elbow in a valgus or varus orientation produced a significant decrease in repeatability when compared to the vertical position data (p = 0.023 and p = 0.008, respectively).
BR I
VALGUS
Ln
LL
409
rn
VARUS
Assessing motion pathivuys Repeatability of varus-valgus angulation and internal-external rotation as a function of flexion angle was assessed using the BR loading ratio and compared with data [8] generated using the same loading ratio on a load-controlled simulator (Fig. 5). Repeatability of the motion pathway using motion-control was similar to or better than that obtained using load-control (p < 0.13). The motion pathways obtained with the current motioncontrolled simulator are not shown, but are similar to those previously published [8].
LOADING MODE
Fig. 4. The repeatability of flexion angle versus time for both the intact and LCL deficient elbow is shown using the means and standard deviations (n = 8) of the maximum standard deviation in five successive trials of simulated active flexion using the various loading modes with the I'orearm maintained in (A) supination and (B) pronation.
-2-
(A) INTACT: V-V ANGULATION 1.0
0.0
-
PRONATED
0
v
0.8
-
-I
0.6 -
v)
1.0
(B) LCL CUT: V-V ANGULATION I
LL 0.4
PRONATED
SUPINATED
e2 5
0.8
-
0.6
-
0.4
-
SUPINATED
-
I-
I-
0
-2
0'4k 0.2
x 5
a
This is the first joint simulator, to our knowledge, to employ motion-controlled pneumatic actuators and to
T
sf 2
Discussion
-
v)
0
Fig. 5. These graphs show the mean (and standard deviation) of the maximum standard deviations found in five successive trials of elbow flexion using the BR loading ratio when measuring varus-valgus (V-V) angulation (A & B), and internakxternal (ILE) rotation (C & D) ( n = 8). In all = 0.005; P[H)< 0.001; cases, the repeatability achieved using motion control was equal to or better than that attained wlth load-control p") = 0.13; p,") = 0.02). The load-control data is based on experiments reported in Dunning et al. [8].
410
C.E. Dunning el ul. I Journal of Orthopurdic Reseurclz 21 (2003) 405-411
introduce a novel method of designating one tendon as the prime mover to achieve velocity control, apportioning smaller magnitudes of load to other muscles in response to the prime mover load on the basis of a loading ratio. Unlike simulators designed to simulate knee motion [7,13] that were able to replicate targeted flexion time-histories, a similar approach cannot yet be applied to the upper limb, since the time-histories of joint motion and force are unknown. In the shoulder simulators of Cain et al. [5] and Sharkey et al. [15], joint motion was achieved using a motion-controlled device attached directly to bone (while constant loads were applied to muscles surrounding the shoulder). A similar approach was used in the elbow simulators described by Bottlang et al. [3,4] and Madey et al. [I21 The wrist simulator developed by Werner et al. [I61 used both position and force feedback hydraulic actuators to achieve targeted wrist motions, but did not apportion muscle loads in accordance with a loading ratio. Our data show the ability of such motion-controlled devices to produce repeatable motion pathways, suggesting advantages to their use for laboratory investigations of joint kinematics. A key component of this new elbow testing system is the custom-written software developed to control motion. This software controls and/or incorporates each component of the testing system in a synchronized fashion: the motion control of the prime mover actuator, the load control of the remaining actuators, the input/output of the prime mover load cell, and the input/ output of the electromagnetic tracking device. In this study, reproducibility of the velocity of motion achieved with the arm placed in the vertical orientation was better than that achieved in a previous study employing a similar approach but using a load-controlled device [lo]. Currently, with the desired velocity of the prime mover actuator set to 5 m d s , the average angular velocity of the elbow (calculated by assuming a linear flexion-time relationship between 30" and 120" of flexion) was 35.8 f 5.8"/s. That is, the velocity coefficient of variation was approximately 16%. In the previous study [lo], similar velocities were achieved but the coefficient of variation was larger due to the inability of loadcontrol to precisely describe joint motion. The scatter observed in the velocity data in the current investigation can be attributed to a number of factors, including differences in arm geometries since, strictly speaking, our model controlled the rate of tendon displacement as opposed to flexion angle versus time. The magnitude of the loads applied to the non-prime mover tendons significantly affected the quality of the resulting motion-control (as quantified by RMS error), as shown by comparing the BR loading ratio to the 1/3 and 2/3 BR ratios. Although the differences were small, they demonstrated that increasing the magnitudes of these other tendon loads decreases the RMS error be-
tween the actual and desired actuator positions. This may be because, as these loads increase, less force from the prime mover is required to achieve the overall resultant load necessary to produce the desired motion. Not only is the payload on the prime mover decreased, but also the load variations experienced by the prime mover through the flexion arc are less dramatic (since they are shared among the various tendons). The repeatability of the flexion angle versus time was not affected by increasing the magnitudes of the non-prime mover tendon loads with the forearm maintained in pronation, but decreased slightly when these assistive loads were increased in the supinated forearm position. This difference is difficult to explain and, based on general trends observed in Fig. 4, is most likely inconsequential. Motion control is achieved by designating one of the tendons as a prime mover. Thc selection of a prime mover is based on the loading ratio employed to apportion load among the relevant tendons. The use of loading ratios based on EMG and pCSA data is somewhat controversial, since it is unclear whether they represent true physiologic loading [8]. However, Dunning et al. [8] showed that, provided the motion occurs unassisted, the selection of a loading ratio may not be important. In this study, the same motions in the vertical orientation were repeated using two different prime movers (BI prime mover is biceps; BR prime mover is brachialis). Neither the RMS crror in the actuator's position nor the repeatability of flexion angle as a function of time, was different between these two loading ratios. The ability to achieve motion-control is apparently not dependent upon thc selection of the prime mover tendon, supporting the previous findings regarding the importance of active loading ratios determined using the load-controlled simulator [8]. Unlike the previous load-controlled design, this system is able to produce simulated active motion in the varus and valgus orientations with the forearm in pronation and supination, respectively. Achieving optimum control in these positions is more difficult than in the vertical position. With the arm in the varus orientation, the RMS error was significantly less than a corresponding motion in the vertical position. In the valgus position, however, RMS error increased when the elbow was moved from the vertical position. These results may be explained by the effect of the elbow's carrying angle on elbow flexion. When the arm is in the varus orientation, the elbow has a tendency to flex spontaneously. Therefore, in this orientation, a constant load must be applied to the triceps tendon, and very little load on the brachialis tendon is required throughout the entire flexion arc. Since the brachialis load does not vary considerably, it may be easier for the motion-controlled actuator to reach its desired target positions. In the valgus orientation, a constant load was also applied to
C.E. Dunning et al. I Journal of Orthupuedic Research 21 (2003) 405-411
the triceps tendon throughout the flexion arc, but this may not have been necessary since spontaneous flexion is not a problem in this position. Perhaps the presence of this triceps load resulted in the increased RMS error when compared to data from the vertical position. The effect of this antagonist load requires further study. It should be noted that although the RMS error values were statistically different when the varus and valgus orientations were compared to the vertical position, the actual magnitudes of these errors were similar and, in practical terms, may be considered quite comparable. The differences seen in the repeatability of flexion angle versus time, however, were comparatively larger in magnitude. In the case of the intact forearm, these differences may also be explained by the effect of the constant triceps load. The further decrease in repeatability measured following LCL sectioning may be attributed to the resulting decrease in elbow stability, especially in the varus orientation. Although the application of tendon loads helps to increase joint stability [8], the varus orientation in particular is a very provocative position for the LCL deficient elbow (i.e. LCL is an important varus stabilizer). In 2 of the 8 arms tested, the application of muscle loads could not keep the elbow from visibly dislocating when simulated active motion was attempted in this orientation. This helps to explain not only the decrease in the repeatability of the flexion data, but also the relatively large standard deviation associated with the measurement (Fig. 4). Controlling forearm position in the varus and valgus orientations is more difficult than in the vertical position and may be considered a limitation of this simulator. In the varus orientation, flexion could only be performed with the forearm maintained in pronation; in the valgus orientation, the forearm was maintained in supination. This is due to the influence of gravity on the radius and hand. For example, to overcome the influence of gravity in the valgus orientation, a very large load has to be applied to the pronator teres to position the forearm in pronation. This has two effects: it influences the speed of the subsequent flexion motion since the pronator teres also acts as a weak elbow flexor, and it applies a large stress on the suture-tendon interface (where the actuator cable is attached) that often resulted in suture failure. These issues will have to be addressed in future studies. In summary, a motion-controlled elbow testing system employing pneumatic actuators has been developed to simulate active elbow flexion in the vertical, varus, or valgus orientations. It produces flexion versus time curves that are highly repeatable, regardless of the loading ratio employed or the prime mover selected. The elbow motion pathways produced are at least as repeatable as those obtained using a load-controlled sim-
41 1
ulator, and the flexion pathway occurs at a slower, more controlled rate. This device will be an important tool in laboratory-based studies aimed at increasing knowledge related to kinematics of the stable and unstable elbow. Acknowledgements The authors wish to acknowledge the funding support of the Natural Science and Engineering Research Council of Canada and the Canadian Institute for Health Research, and the technical assistance of Mr. Louis Ferreira. References [I] Amis AA, Dowson D, Wright V. Muscle strengths and musculoskeletal geometry of the upper limb. Eng Med 1979;8:41-8. [2] An K N , Hui FC, Morrey BF, Linscheid RL, Chao EY. Muscles across the elbow joint: a biomechanical analysis. J Biomech 198 1: 14:659-~69. [3] Bottlang M, Madey SM, Steyers CM, Marsh JL, Brown TD. Assessment of elbow joint kinematics in passive motion by electromagnetic motion tracking. J Orthop Res 2000:18: I95 202. [4] Bottlang M, O’Rourke M R , Steyers CM, Marsh JL, Brown T D . Radiographic landmarks of the rotation axis of the humero-ulnar articulation. Trans ORS 1999;24( l):367. [5] Cain PR, Mutschler TA, Fu FH, Lee SK. Anterior stability of the glenohumeral joint. A dynamic model. Am J Sports Med 1987: 1 5: 144-8. [6] Caldwell GE, Van Leemputte M. Elbow torques and EMG patterns of flexor muscles during different isometric tasks. Electromyogr Clin Neurophysiol 1991:31:43345. [7] DiAngelo DJ, Harrington IA. Design of a dynamic multi-purpose joint simulator. Adv Bioeng, ASME 1992;22:107-10. [8] Dunning CE, Duck TR, King GJ, Johnson JA. Simulated active control produces repeatable motion pathways of the elbow in an in vitro testing system. J Biomech 2001;34:103948. [9] Funk DA, An KN, Morrey BF, Daube JR. Electromyographic analysis of muscles across the elbow joint. J Orthop Res 1987; 5:529-38. [lo] Johnson JA, Rath DA, Dunning CE, Roth SE. King GJ. Simulation of elbow and forearm motion in vitro using a load controlled testing apparatus. J Biomech 2000;33:635-9. [ I l l King GJ, Zarzour ZD, Rath DA, Dunning CE, Patterson SD, Johnson JA. Metallic radial head arthroplasty improves valgus stability of the elbow. Clin Orthop 1999:114 25. [I21 Madey SM, Bottlang M , Steyers CM, Marsh JL. Brown T D . Hinged external fixation of the elbow: optimal axis alignment to minimize motion resistance. J Orthop Trauma 2000; 14:41-7. [I31 McLean CA, Ahmed AM. Design and development of an unconstrained dynamic knee simulator. J Biomech Eng 1993; 115:144--8. [14] Milne AD, Chess DG, Johnson JA, King GJ. Accuracy of an electromagnetic tracking device: a study of the optimal range and metal interference. J Biomech 1996;29:791-3. [I51 Sharkey NA, Marder RA. Hanson PB. The entire rotator cuff contributes to elevation of the arm. J Orthop Res 1994:12:699 708. [I61 Werner FW, Palmer AK, Somerset JH, Tong JJ, Gillison DB, Fortino MD, et al. Wrist joint motion simulator. J Orthop Res 1996;14:639 46.
Journal of Orthopaedic Research 21 (2003) 4 0 5 4 1 1
Journal of Orthopaedic Research www.elsevier.comllocate1orthres
Development of a motion-controlled in vitro elbow testing system Cynthia E. Dunning
Karen D. Gordon Graham J.W. King James A. Johnson a,b3c,d,e,*
aib,d,
a,d,e,
Bioengineering Researcli Luhoratorj3 Lawson Health Restlurch Institute, Hond & Upper Limb Centre, St. Josc~ph'sHeulth Cure London, London, Ontario, Cunadu Department of Mc4iaiiical and Matc,riuls Engineering, The Uniuersit.v oj Western Ontario, Lon(lon, Ontario, Cunudo Department uf' Biomedical Engineering, The tiniaersiry .f Western Ontario. London. Onturio, Cumdo Dtyurirrient of' Surgery, The Uniuwsiiy of' Uresisl(wiOniurio, London. Ofiturio, Cunudu Departnient of Mrdicul Biophysics, The tiiiiwrsiiy of Western Ontario, London, Onturio, Cmcrdu
Abstract
Joint simulators can be used to study motion pathways of a human joint, to investigate changes in joint stability following injury, and to formulate improved reconstructive and rehabilitative procedures. Our objectives were: to develop a laboratory-based, motion-controlled elbow testing apparatus capable of simulating tendon (muscle) loading and displacement in a cadaveric specimen; to describe its performance while testing stable and unstable elbows; and to compare its operation to that of a previously designed load-controlled device. Velocity control of a pneumatic actuator was achieved using a custom-written, closed-loop feedback controller. This actuator was incorporated into an elbow testing system that used additional pneumatic actuators and a combination of motion- and load-control to achieve desired motions. Simulations achieved with this apparatus demonstrated small magnitudes of error in actuator position and highly repeatable flexion pathways with the specimens positioned in vertical, varus, and valgus orientations. The repeatability i n motion pathways generated in both a stable and unstable elbow model was equivalent to or better than for similar tests performed using the load-controlled system, and the velocity of the resulting elbow motion was more reproducible. 0 2003 Orthopaedic Research Society. Published by Elsevier Science Ltd. All rights reserved.
Introduction In vitro testing in the laboratory plays a n important role in advancing patient care by helping t o develop new treatment techniques a n d improved rehabilitation protocols. Due to a lack of adequate theoretical models, these investigations are important in elucidating injury mechanisms. Moreover, these studies should be useful in validating new surgical and rehabilitative treatments prior to their clinical implementation. A number of laboratory-based systems capable of simulating active elbow motion in vitro have been designed. Some of these devices are load-controlled, where specific force magnitudes are applied directly to bones o r tendons to produce joint loading a n d motion. Johnson
* Corresponding author. Address: Bioengineering Research Laboratory, Lawson Health Research Institute. Hand & Upper Limb Centre, St. Joseph's Health Care London, London. Ontario. Canada. Tel.: +1-519-646-601 I : fax: +I-519-646-6049. E-rnriil citflrrrss: [email protected] (J.A. Johnson).
et al. [lo] described a loading device incorporating pneumatic actuators to apply sufficient loads t o relevant tendons to produce unassisted elbow flexion-extension and forearm pronation-supination with the arm in the vertical orientation. Other elbow simulators are motioncontrolled devices, since they achieve simulated joint movement by controlling bone o r tendon displacement using a motor o r actuator. Bottlang et al. [3,4] and Madey et al. [12] described devices that used DC motors connected directly to the bone(s) of the forearm to move the elbow through the flexion-extension cycle. No motion-controlled testing device currently exists to produce elbow motion through tendon loading. Past experience in o u r laboratory has shown that a loadcontrolled elbow testing system is a practical approach to simulate joint flexion [lo]. In some joint positions, however, inertial effects can make it very difficult to simulate smooth, gentle joint motions using only loadcontrol. F o r example, in the varus orientation (i.e. the humerus is fixed in a horizontal position with the medial aspect of the elbow inferior), the carrying angle formed
0736-0266/03/$ - see front matter 0 2003 Orthopaedic Research Society. Published by Elsevier Science Ltd. All rights reserved doi:l0.10161S0736-0266(02)00233-4
406
c‘ E Dunning et ul I Journal of Orthopuedc RcJseuvch21 (2003) 405-411
between the ulna and humerus tends to produce spontaneous, rapid elbow flexion even with very small magnitudes of load applied to the elbow flexor muscles. In addition, the lack of adequate and predictable resistance during loading challenges load-controlled motion. The use of motion-controlled testing systems can help minimize these effects by maintaining tendon excursion (and thus joint rotation) at a constant velocity as opposed to applying a prescribed tendon force. Therefore, the purpose of this work was threefold: to design and develop a motion-controlled testing system to produce elbow motion; to evaluate the performance of this system in both the stable and unstable elbows; and to compare the elbow motion pathways it generates to those achieved using a load-controlled testing system.
PNEUMATIC ACTUATORS IN-LINE LOAD CELL
BASE PLATE
SECONDARY ALIGNMENT SYSTEM
CABLES TRANSMITTER ALIGNMENT RING
Methods The motion-controlled elbow simulator (Fig. 1) is machined from polyethylene, Delrin“, and 3 16L stainless steel to avoid interference with the electromagnetic tracking device [14], whose receivers are secured to the specimen (see below) and whose transmitter is secured to the base plate of the simulator. A cadaveric specimen is attached to the device via the humeral mounting clamp. Cables that have been sutured to five tendons of interest (biceps, brachialis, brachioradialis, triceps, and pronator teres) are each connected to a dedicated actuator. This simulator incorporates a two-part alignment system to control the lines-of-action of the cables. The primary alignment system consists of two separate units: an alignment guide (located within the humeral clamp), and an alignment ring (located near the transmitter). Both units contain pulley mechanisms that incorporate a two degree-offreedom (2 DOF) design, which enables them to accommodate changes in the lines-of-action that occur as the elbow moves through a range of motion. That is, these mechanisms are free to rotate about the axis of the approaching cable. As the angle of approach changes, they rotate to ensure their pulleys remain in contact with the cable. Cables attached to the brachioradialis and pronator teres pass through the alignment ring on route to their respective actuators. Moment arms for the cables attached to the biceps (3.3 cm), brachialis (2.0 cm), and triceps (1.6 cm) are maintained by the alignment guide, positioned approximately 4 cm proximal to the elbow joint [2]. A secondary alignment system, which uses simple pulleys, ensures that each Cable is in line with the piston of its dedicated actuator. The height of each pulley in the secondary alignment system can be adjusted to match the height of its actuator. The entire base plate is mounted o n a polyethylene stand via a 2 D O F hinge. This allows the simulator to be rotated into varus and valgns orientations (i.e. the humerus is fixed in a horizontal position with the medial aspect of the elbow inferior or superior, respectively), as well as into the vertical position (i.e. the humerus is vertical and the forearm is free to move through an arc of flexion) [8]. To initiate motion-control, the tendon of the muscle that supplies the largest percentage of the total applied load responsible for producing the motion is designated as the “prime mover”. The percentage depends upon the muscle-loading ratio, which is developed from EMG and pCSA data as has been previously reported [8]; two of these ratios are listed in Table 1. Using the biceps (BI) loading ratio, the biceps is the prime mover for both pronated and supinated flexion. However, the brachialis muscle is the prime mover according to the brachialis (BR) loading ratio. Therefore, both the brachiah and biceps are connected to pneumatic actuators (model PFC-096-XLP, BIMBA Ltd., Cambridgeshire, UK) that contain an internal linear resistive transducer (LRT) to provide position feedback as required for motion control. These position feedback actuators can also be operated under load-control. The remaining 3 tendons are connected to actuators (model 096-DX, BIMBA L.td., Cambridgeshire, UK) that operate under load-control only (Le. do not contain an LRT).
PRIMARY
- ALIGNMENT
HUMERAL MOUNTING CLAMP
SYSTEM
ALIGNMENT GUIDE RECEIVER ---#m~
i-.*‘“
Fig. 1. The motion-controlled elbow simulator is shown with the base plate in the vertical position. A universal hinge attached to the base plate also allows the simulator to be placed in varus and valgus orientations, where the weight of the forearm applies a provocative moment about the elbow joint. A cadaveric specimen is attached to the device via a humeral mounting clamp. Cables, sutured to five tendons of interest, are connected to independent pneumatic actuators. Primary and secondary alignment systems that contain adjustable pulleys are used to maintain proper (physiologic) lines-of-action for each cables. An electromagnetic tracking device, consisting of a receiver fixed to the ulna and a transmitter secured to the base plate, is used to measure joint motion.
Control of each actuator is achieved using a custom-written LabVIEW software program (Fig. 2). A key feature of this code is a digital proportional-integral-derivative (PID) controller used within a realtime feedback control loop to move the piston of the prime mover actuator at a constant velocity (i.e. controls the rate of tendon excursion during the flexion arc). The controller acts to reduce or eliminate error between the actuator’s current position (i.e. the process variable) to its desired position (i.e. the set point). Controller output is a voltage signal sent to a dedicated proportional-pressure-controller (PPC) (model #PPC5C-AAA-AGCB-BBB-JD, MAC Valves, Wixon, MI), that adjusts the pressure of the air supplied to the cylinder as required to achieve the desired actuator position. The controlling software also receives input from a custom-built, strain gauge based, 1 D O F load cell used to monitor the force applied by the prime mover at 200 Hz. This signal is used to apportion load to the load-controlled actuators in accordance with the loading ratio selected; i.e., the controlling software sends appropriate voltage signals to independent PPCs connected to the each of the simulator’s load-controlled actuators to simulate muscle tension. Testing can be performed with the forearm maintained in either supination or pronation. This is accomplished by initially activating the appropriate tendons using a load-controlled protocol, as previously described [8,10]. For example, to position the arm in supination, load is applied to the biceps tendon using a ramp function over a 5 s period. Simultaneously, a ramp load is applied to the triceps to keep
C.E. Dunning et al. I Journal of Orthopaedic Research 21 (2003 j 405 4 1 I
407
Table 1 Loading ratios for simulated active elbow flexion Motion
Loading ratio
Biceps (%)
Brachialis (YO)
Brachioradalis (%I)
Triceps ((XI)
Supinated flexion
BI BR
26 31
21 42
9 I1
44 16
Pronated flexion
BI BR
25 22
21 41
8 16
46 21
Note: BI = biceps loading ratio; BR = brachialis loading ratio. B1 loading ratio: E M G data from Caldwell and Van Leemputte [6] and pCSA data from Amis et al. [I]. BR loading ratio: EMG data from Funk et al. [9] and pCSA data from Amis et al. [l].
ELECTROMAGNETIC TRACKING SYSTEM DATA INITIALIZATION USERINPUTS
MOTION DATA
---,, - + OUT P UT
(based on load cell feedback and loadino ratio selected)
DATA FILES
VOLTAGE SIGNAL
VOLTAGE SIGNAL CONTROLLING COMPUTER POSITION
(generated by PID controller based on LRT feedback)
PPC
LOAD CONTROL FEEDBACK REGULATOR
Fig. 2. The main controlling computer operating with cnstom-designed LabVIEW software is a central feature of the motion-controlled simulator. It receives inputs from the user that include PID controller gains, desired actuator velocity, forearm position (supination, pronation), simulator position (vertical, varus, valgus), loading ratio, threshold load values, and load cell calibration data. It also receives feedback information from the load cell and the LRT on the prime mover actuator and uses it to apportion loads to the load-controlled actuators and provide an error signal to the PID controller, respectively. The computer regulates voltage to the PPCs attached to both the prime mover and the load-controlled actuators. In addition, it initializes the electromagnetic tracking device to collect data once the prime mover is activated, and stores the kinematic data, and the position and load values from the prime mover actuator.
the elbow in a fully extended position. The magnitudes of these loads are determined using an iterative procedure to determine the minimum magnitudes of load necessary to achieve both full supination and full extension. After the 5 s, the motion-controlled algorithm is activated to initiate controlled elbow flexion, and the data collection begins. This algorithm ensures that the biceps load will not fall below the iteratively determined threshold value needed to maintain the supinated position. Triceps activation (indicated by the loading ratios in Table 1) is not maintained throughout the flexion arc when simulating motion in the vertical orientation. Rather, once the motion-controlled movement begins, the triceps load is ramped down to 0 N over a 5 s interval. In the varus and valgus orientations, however, the triceps load is maintained at a constant value throughout the flexion arc. A similar procedure was followed for testing in the pronated forearm position, where the pronator teres tendon was initially activated (as opposed to the biceps) and a threshold value of load for this tendon was determined. Experimental testing was conducted to evaluate the performance of the elbow testing system. Eight fresh-frozen cadaveric upper limbs were tested (mean age: 67 19 years; range: 29-85 years) in accor-
*
dance with the rules and regulations of our institution and government regulatory agency. Specimens were thawed to room temperature (23 f 2 “C) prior to testing. Initial specimen preparation involved exposing and dissecting the tendons of the biceps, brachialis, brachioradialis, triceps, and pronator teres. Ticron #2 suture (Sherwood Medical, St. Louis, MO) was secured to each distal tendon using a Bunnell technique. The suture was then tied to a length of stainless steel cable via a nylon washer. Cables from the brachioradialis and pronator teres tendons were routed beneath the skin to the lateral and medial epicondyles, respectively. Holes were drilled through each epicondyle into the humcral canal. Specially designed Delrin@ sleeves were inserted into these holes (to minimize friction), and the cables passed through the sleeves and out the proximal end of the humeral canal. Lightweight Delrin@’pedestals, designed to accept receivers from the ‘Flock of Birds’ (Ascension Technologies, Burlington, VT) electromagnetic tracking device, were secured to the distal radius and ulna using bone screws. Each specimen was secured in the humeral clamp, the tendon cables were fed through the appropriate alignment pulleys and attached to dedicated actuators, and the tracking device receivers were attached to their pedestals. With the simulator placed in the vertical orientation, load-controlled testing was performed to determine the threshold values as described above. Motion-control testing began by performing an iterative procedure to select appropriate PID controller gains with actuator backpressure set to 276 kPa, and the desired piston velocity set to 5 mmls. These magnitudes of pressure and velocity were based on pilot tests conducted to determine the desired rate of controlled elbow motion. The gain values were adjusted to achieve a gentle arc of smooth elbow flexion. Once determined, the gain values remained constant for the duration of testing. The testing protocol was designed to investigate effects of loading ratio, reduced loads to the non-prime mover tendons, simulator orientation, and ligament sectioning on the performance of the motioncontrolled simulator. Each test was repeated five times consecutively to quantify repeatability. With the forearm maintained in supination, simulated active flexion was first performed by designating the brachialis as the prime mover and apportioning load ( i s . load control) to the biceps and brachioradialis according to the BR loading ratio (Table 1). The loads to the load-controlled tendons were then reduced to 1/3 and then 2/3 of the values prescribed by the BR loading ratio. Motions employing the BI loading ratio, which designates the biceps as the prime mover, were also performed. Similar motions were performed with the forearm maintained in pronation. The simulator was then rotated into the valgus orientation (i.e. using the aforementioned 2 D O F hinge), where the BR loading ratio was used to achieve five simulated active flexion motions with the forearm in supination. Elbow flexion with the forearm in pronation was attempted with the simulator in this orientation, but was unsuccessful. Testing in the varus orientation was also performed using the BR loading ratio, but with the forearm maintained in pronation. In this position, elbow flexion with the forearm in supination could not be performed. The threshold values used as limits for the biceps and pronator teres loads in the valgus and varns orientations, respectively, were roughly 50% of those used in the vertical position. An unstable elbow was then created by complete transection of the anterior and posterior joint capsules and the radial and lateral ulnar collateral ligaments (collectively referred to as the lateral collateral ligament or LCL), and the entire testing protocol was repeated. At the completion of testing, each specimen
408
C. E. Dunning et ul. I Journal of Orthopuedic Reseavch 21 (2003) 405-411
was disarticulated at the wrist and elbow to digitize the necessary bony landmarks used to create anatomic bone coordinate systems [ I I]. This allowed motion data collected by the tracking device to be expressed in clinically relevant terms of flexion-extension, varus-valgus angulation, and internal-external rotation according to an Euler Z Y-X analysis. The performance of the motion-controlled simulator was evaluated using three separate parameters. Root-mean-square (RMS) error was used to quantify the deviation between the desired and actual positions attained by the motion-controlled actuator throughout the flexion arc. To quantify the ability of the simulator to repeatedly produce the same flexion position as a function of time, the standard deviation of flexion angle from the five successive trials was calculated at each second throughout the duration of the simulated movement. The maximum standard deviation, regardless of the time at which it occurred, was used as a measure of repeatability. Repeatability of the elbow motion pathways generated from the Euler analysis (varus- valgus angulation and internal-external rotation as a function of flexion angle) was also evaluated using the maximum standard deviation in five trials and compared to the results obtained with the load-controlled simulator [8]. All three sets of data (RMS error, maximum standard deviation in flexion angle versus time, and maximum standard deviation in elbow motion pathways) were analyzed statistically using two-way repeated measure analyses of variance and post-hoc Student-Newman-Keuls tests with 2 = 0.05.
a w 1 v)
a . a
n --
I
BR
113BR
213BR
BI
I
VERTICAL
LEI VALGUS
LOADING MODE
U
a w 1 v)
I . Results
a
0 -
R M S error in actuator position
I
BR
113BR
213BR
BI
1
VERTICAL
Excellent agreement between the desired and actual positions of the motion control actuator was obtained as indicated by the small magnitude of (RMS) errors (Fig. 3). For the 8 specimens tested, the average RMS values calculated in the vertical position ranged from 0.7 mm (intact elbow, supinated forearm, BI loading ratio) to 1.6 mm (LCL cut, pronated forearm, BI loading ratio). Comparing the 1/3, 2/3 and BR ratios, the RMS error with the forearm maintained in supination was significantly greater for the 1/3 ratio than either the 2/3 (p =0.003) or full (p = 0.002) ratios (Fig. 3A). In the pronated forearm position, RMS error was significantly greater for both the 1/3 (p < 0.001) and 2/3 0, = 0.003) compared to the BR ratio (Fig. 3B). No difference in the magnitude of the RMS errors was found when comparing the B1 and BR loading ratios in either supination (p = 0.05) or pronation (p = 0.33). For all of the tests performed in the vertical position, cutting the LCL had no significant effect on the error measurements (p > 0.28), although the magnitudes of the error and their standard deviations generally tended to increase with ligament sectioning. In the varus and valgus orientations, the range in the average RMS error measured was from 0.9 mm (LCL cut, pronated forearm, varus orientation) to 1.5 mm (LCL cut, supinated forearm, valgus orientation). When the BR loading ratio was used in supination, the error was significantly greater in the valgus orientation than the vertical position (p = 0.005). For this subset of data, cutting the LCL also significantly increased the error (p = 0.004). RMS error was significantly larger in the vertical position than the varus orientation when the BR loading ratio was used with the
rn VARUS
LOADING MODE Fig. 3 . The means and standard deviations ( n = 8) of RMS error between the desired and actual positions attained by the motion-controlled actuator are shown for the various loading modes tested in both the intact and LCL deficient elbows with the forearm maintained in (A) supination and (B) pronation.
forearm maintained in pronation (p = 0.006), but there was no effect of ligament sectioning (p = 0.60).
Repeatability of flexion angle versus time In the vertical position, the results were highly repeatable (Fig. 4) with the average maximum standard deviation values as small as 0.9" (intact elbow, supinated forearm, 2/3 BR loading ratio) and never greater than 2.8" (LCL cut, pronated forearm, 1/3 BR loading ratio). Significant differences in repeatability were found when comparing the 1/3, 2/3, and BR loading ratios with the forearm maintained in supination (p = 0.001) (Fig. 4A). Tests performed with both the 113 (p = 0.004) and 213 (p = 0.001) ratios were more repeatable than those using the BR loading ratio. The loading ratio used did not significantly affect repeatability when the forearm was maintained in pronation (p = 0.50) (Fig. 4B). No differences in repeatability were found between the BI and BR loading ratios in either supination (p = 0.94) or pronation (p = 0.81). With the arm in the vertical orientation, the effect of cutting the LCL was significant in supination 0,= 0.01; LCL cut was less repeatable), but not in pronation (p = 0.83). In the varus and valgus orientations, the average maximum standard deviation
C.E. Dunning et ul. I Journal of Orthopuedic Reseurch 21 (2003j 405411
n
u)
3
I
S
BR
'
113BR
2ABR
81 I I
VERTICAL LOADING MODE
I
0
5 -
n
* 2
n -
-
I
BR
113BR 2ABR
I
81
I
VERTICAL
values were larger than those found in the vertical position, ranging from 5.5" (intact, supinated forearm, valgus orientation) to 8.4" (LCL cut, pronated forearm, varus orientation). When the BR loading ratio was employed, placing the elbow in a valgus or varus orientation produced a significant decrease in repeatability when compared to the vertical position data (p = 0.023 and p = 0.008, respectively).
BR I
VALGUS
Ln
LL
409
rn
VARUS
Assessing motion pathivuys Repeatability of varus-valgus angulation and internal-external rotation as a function of flexion angle was assessed using the BR loading ratio and compared with data [8] generated using the same loading ratio on a load-controlled simulator (Fig. 5). Repeatability of the motion pathway using motion-control was similar to or better than that obtained using load-control (p < 0.13). The motion pathways obtained with the current motioncontrolled simulator are not shown, but are similar to those previously published [8].
LOADING MODE
Fig. 4. The repeatability of flexion angle versus time for both the intact and LCL deficient elbow is shown using the means and standard deviations (n = 8) of the maximum standard deviation in five successive trials of simulated active flexion using the various loading modes with the I'orearm maintained in (A) supination and (B) pronation.
-2-
(A) INTACT: V-V ANGULATION 1.0
0.0
-
PRONATED
0
v
0.8
-
-I
0.6 -
v)
1.0
(B) LCL CUT: V-V ANGULATION I
LL 0.4
PRONATED
SUPINATED
e2 5
0.8
-
0.6
-
0.4
-
SUPINATED
-
I-
I-
0
-2
0'4k 0.2
x 5
a
This is the first joint simulator, to our knowledge, to employ motion-controlled pneumatic actuators and to
T
sf 2
Discussion
-
v)
0
Fig. 5. These graphs show the mean (and standard deviation) of the maximum standard deviations found in five successive trials of elbow flexion using the BR loading ratio when measuring varus-valgus (V-V) angulation (A & B), and internakxternal (ILE) rotation (C & D) ( n = 8). In all = 0.005; P[H)< 0.001; cases, the repeatability achieved using motion control was equal to or better than that attained wlth load-control p") = 0.13; p,") = 0.02). The load-control data is based on experiments reported in Dunning et al. [8].
410
C.E. Dunning el ul. I Journal of Orthopurdic Reseurclz 21 (2003) 405-411
introduce a novel method of designating one tendon as the prime mover to achieve velocity control, apportioning smaller magnitudes of load to other muscles in response to the prime mover load on the basis of a loading ratio. Unlike simulators designed to simulate knee motion [7,13] that were able to replicate targeted flexion time-histories, a similar approach cannot yet be applied to the upper limb, since the time-histories of joint motion and force are unknown. In the shoulder simulators of Cain et al. [5] and Sharkey et al. [15], joint motion was achieved using a motion-controlled device attached directly to bone (while constant loads were applied to muscles surrounding the shoulder). A similar approach was used in the elbow simulators described by Bottlang et al. [3,4] and Madey et al. [I21 The wrist simulator developed by Werner et al. [I61 used both position and force feedback hydraulic actuators to achieve targeted wrist motions, but did not apportion muscle loads in accordance with a loading ratio. Our data show the ability of such motion-controlled devices to produce repeatable motion pathways, suggesting advantages to their use for laboratory investigations of joint kinematics. A key component of this new elbow testing system is the custom-written software developed to control motion. This software controls and/or incorporates each component of the testing system in a synchronized fashion: the motion control of the prime mover actuator, the load control of the remaining actuators, the input/output of the prime mover load cell, and the input/ output of the electromagnetic tracking device. In this study, reproducibility of the velocity of motion achieved with the arm placed in the vertical orientation was better than that achieved in a previous study employing a similar approach but using a load-controlled device [lo]. Currently, with the desired velocity of the prime mover actuator set to 5 m d s , the average angular velocity of the elbow (calculated by assuming a linear flexion-time relationship between 30" and 120" of flexion) was 35.8 f 5.8"/s. That is, the velocity coefficient of variation was approximately 16%. In the previous study [lo], similar velocities were achieved but the coefficient of variation was larger due to the inability of loadcontrol to precisely describe joint motion. The scatter observed in the velocity data in the current investigation can be attributed to a number of factors, including differences in arm geometries since, strictly speaking, our model controlled the rate of tendon displacement as opposed to flexion angle versus time. The magnitude of the loads applied to the non-prime mover tendons significantly affected the quality of the resulting motion-control (as quantified by RMS error), as shown by comparing the BR loading ratio to the 1/3 and 2/3 BR ratios. Although the differences were small, they demonstrated that increasing the magnitudes of these other tendon loads decreases the RMS error be-
tween the actual and desired actuator positions. This may be because, as these loads increase, less force from the prime mover is required to achieve the overall resultant load necessary to produce the desired motion. Not only is the payload on the prime mover decreased, but also the load variations experienced by the prime mover through the flexion arc are less dramatic (since they are shared among the various tendons). The repeatability of the flexion angle versus time was not affected by increasing the magnitudes of the non-prime mover tendon loads with the forearm maintained in pronation, but decreased slightly when these assistive loads were increased in the supinated forearm position. This difference is difficult to explain and, based on general trends observed in Fig. 4, is most likely inconsequential. Motion control is achieved by designating one of the tendons as a prime mover. Thc selection of a prime mover is based on the loading ratio employed to apportion load among the relevant tendons. The use of loading ratios based on EMG and pCSA data is somewhat controversial, since it is unclear whether they represent true physiologic loading [8]. However, Dunning et al. [8] showed that, provided the motion occurs unassisted, the selection of a loading ratio may not be important. In this study, the same motions in the vertical orientation were repeated using two different prime movers (BI prime mover is biceps; BR prime mover is brachialis). Neither the RMS crror in the actuator's position nor the repeatability of flexion angle as a function of time, was different between these two loading ratios. The ability to achieve motion-control is apparently not dependent upon thc selection of the prime mover tendon, supporting the previous findings regarding the importance of active loading ratios determined using the load-controlled simulator [8]. Unlike the previous load-controlled design, this system is able to produce simulated active motion in the varus and valgus orientations with the forearm in pronation and supination, respectively. Achieving optimum control in these positions is more difficult than in the vertical position. With the arm in the varus orientation, the RMS error was significantly less than a corresponding motion in the vertical position. In the valgus position, however, RMS error increased when the elbow was moved from the vertical position. These results may be explained by the effect of the elbow's carrying angle on elbow flexion. When the arm is in the varus orientation, the elbow has a tendency to flex spontaneously. Therefore, in this orientation, a constant load must be applied to the triceps tendon, and very little load on the brachialis tendon is required throughout the entire flexion arc. Since the brachialis load does not vary considerably, it may be easier for the motion-controlled actuator to reach its desired target positions. In the valgus orientation, a constant load was also applied to
C.E. Dunning et al. I Journal of Orthupuedic Research 21 (2003) 405-411
the triceps tendon throughout the flexion arc, but this may not have been necessary since spontaneous flexion is not a problem in this position. Perhaps the presence of this triceps load resulted in the increased RMS error when compared to data from the vertical position. The effect of this antagonist load requires further study. It should be noted that although the RMS error values were statistically different when the varus and valgus orientations were compared to the vertical position, the actual magnitudes of these errors were similar and, in practical terms, may be considered quite comparable. The differences seen in the repeatability of flexion angle versus time, however, were comparatively larger in magnitude. In the case of the intact forearm, these differences may also be explained by the effect of the constant triceps load. The further decrease in repeatability measured following LCL sectioning may be attributed to the resulting decrease in elbow stability, especially in the varus orientation. Although the application of tendon loads helps to increase joint stability [8], the varus orientation in particular is a very provocative position for the LCL deficient elbow (i.e. LCL is an important varus stabilizer). In 2 of the 8 arms tested, the application of muscle loads could not keep the elbow from visibly dislocating when simulated active motion was attempted in this orientation. This helps to explain not only the decrease in the repeatability of the flexion data, but also the relatively large standard deviation associated with the measurement (Fig. 4). Controlling forearm position in the varus and valgus orientations is more difficult than in the vertical position and may be considered a limitation of this simulator. In the varus orientation, flexion could only be performed with the forearm maintained in pronation; in the valgus orientation, the forearm was maintained in supination. This is due to the influence of gravity on the radius and hand. For example, to overcome the influence of gravity in the valgus orientation, a very large load has to be applied to the pronator teres to position the forearm in pronation. This has two effects: it influences the speed of the subsequent flexion motion since the pronator teres also acts as a weak elbow flexor, and it applies a large stress on the suture-tendon interface (where the actuator cable is attached) that often resulted in suture failure. These issues will have to be addressed in future studies. In summary, a motion-controlled elbow testing system employing pneumatic actuators has been developed to simulate active elbow flexion in the vertical, varus, or valgus orientations. It produces flexion versus time curves that are highly repeatable, regardless of the loading ratio employed or the prime mover selected. The elbow motion pathways produced are at least as repeatable as those obtained using a load-controlled sim-
41 1
ulator, and the flexion pathway occurs at a slower, more controlled rate. This device will be an important tool in laboratory-based studies aimed at increasing knowledge related to kinematics of the stable and unstable elbow. Acknowledgements The authors wish to acknowledge the funding support of the Natural Science and Engineering Research Council of Canada and the Canadian Institute for Health Research, and the technical assistance of Mr. Louis Ferreira. References [I] Amis AA, Dowson D, Wright V. Muscle strengths and musculoskeletal geometry of the upper limb. Eng Med 1979;8:41-8. [2] An K N , Hui FC, Morrey BF, Linscheid RL, Chao EY. Muscles across the elbow joint: a biomechanical analysis. J Biomech 198 1: 14:659-~69. [3] Bottlang M, Madey SM, Steyers CM, Marsh JL, Brown TD. Assessment of elbow joint kinematics in passive motion by electromagnetic motion tracking. J Orthop Res 2000:18: I95 202. [4] Bottlang M, O’Rourke M R , Steyers CM, Marsh JL, Brown T D . Radiographic landmarks of the rotation axis of the humero-ulnar articulation. Trans ORS 1999;24( l):367. [5] Cain PR, Mutschler TA, Fu FH, Lee SK. Anterior stability of the glenohumeral joint. A dynamic model. Am J Sports Med 1987: 1 5: 144-8. [6] Caldwell GE, Van Leemputte M. Elbow torques and EMG patterns of flexor muscles during different isometric tasks. Electromyogr Clin Neurophysiol 1991:31:43345. [7] DiAngelo DJ, Harrington IA. Design of a dynamic multi-purpose joint simulator. Adv Bioeng, ASME 1992;22:107-10. [8] Dunning CE, Duck TR, King GJ, Johnson JA. Simulated active control produces repeatable motion pathways of the elbow in an in vitro testing system. J Biomech 2001;34:103948. [9] Funk DA, An KN, Morrey BF, Daube JR. Electromyographic analysis of muscles across the elbow joint. J Orthop Res 1987; 5:529-38. [lo] Johnson JA, Rath DA, Dunning CE, Roth SE. King GJ. Simulation of elbow and forearm motion in vitro using a load controlled testing apparatus. J Biomech 2000;33:635-9. [ I l l King GJ, Zarzour ZD, Rath DA, Dunning CE, Patterson SD, Johnson JA. Metallic radial head arthroplasty improves valgus stability of the elbow. Clin Orthop 1999:114 25. [I21 Madey SM, Bottlang M , Steyers CM, Marsh JL. Brown T D . Hinged external fixation of the elbow: optimal axis alignment to minimize motion resistance. J Orthop Trauma 2000; 14:41-7. [I31 McLean CA, Ahmed AM. Design and development of an unconstrained dynamic knee simulator. J Biomech Eng 1993; 115:144--8. [14] Milne AD, Chess DG, Johnson JA, King GJ. Accuracy of an electromagnetic tracking device: a study of the optimal range and metal interference. J Biomech 1996;29:791-3. [I51 Sharkey NA, Marder RA. Hanson PB. The entire rotator cuff contributes to elevation of the arm. J Orthop Res 1994:12:699 708. [I61 Werner FW, Palmer AK, Somerset JH, Tong JJ, Gillison DB, Fortino MD, et al. Wrist joint motion simulator. J Orthop Res 1996;14:639 46.