An Approach to Robotic Testing of the Wrist Using Three-Dimensional Imaging and a Hybrid Testing Methodology

Abstract

Robotic technology is increasingly used for sophisticated in vitro testing designed to understand the subtleties of joint biomechanics. Typically, the joint coordinate systems in these studies are established via palpation and digitization of anatomic landmarks. We are interested in wrist mechanics in which overlying soft tissues and indistinct bony features can introduce considerable variation in landmark localization, leading to descriptions of kinematics and kinetics that may not appropriately align with the bony anatomy. In the wrist, testing is often performed using either load or displacement control with standard material testers. However, these control modes either do not consider all six degrees-of-freedom (DOF) or reflect the nonlinear mechanical properties of the wrist joint. The development of an appropriate protocol to investigate complexities of wrist mechanics would potentially advance our understanding of normal, pathological, and artificial wrist function. In this study, we report a novel methodology for using CT imaging to generate anatomically aligned coordinate systems and a new methodology for robotic testing of wrist. The methodology is demonstrated with the testing of 9 intact cadaver specimens in 24 unique directions of wrist motion to a resultant torque of 2.0 N·m. The mean orientation of the major principal axis of range of motion (ROM) envelope was oriented 12.1 ± 2.7 deg toward ulnar flexion, which was significantly different (p < 0.001) from the anatomical flexion/extension axis. The largest wrist ROM was 98 ± 9.3 deg in the direction of ulnar flexion, 15 deg ulnar from pure flexion, consistent with previous studies [1,2]. Interestingly, the radial and ulnar components of the resultant torque were the most dominant across all directions of wrist motion. The results of this study showed that we can efficiently register anatomical coordinate systems from CT imaging space to robotic test space adaptable to any cadaveric joint experiments and demonstrated a combined load-position strategy for robotic testing of wrist.

Introduction

Robotic manipulators have been used in orthopedic and biomechanical research to determine joint kinetics, joint kinematics, and ligament and tendon function [3–22]. Typically, the robot is programmed to move adjacent bones based on the location and orientation of a joint coordinate system transformed into the robot's built-in coordinate space. Defining an anatomically appropriate joint coordinate system and placing that joint coordinate system within the reference space of the robot enables the robot to move in a manner that simulates joint motion.

A robust joint coordinate system is required to define the location and orientation of the specimen with respect to the robotic test device to ensure repeatable and standardized testing. Frey et al. estimated knee joint pose from reconstructed CT scans and located them in robot space using aluminum tubes with screws [6]. More recent studies have utilized portable coordinate measuring systems [7,9,11,13–15,17,18] and three-dimensional (3D) spatial digitizers [19,20] to digitize bony landmarks in the construction of joint coordinate systems. These digitizers are easy to use, fast, and accurate at locating bony landmarks that can be precisely and repeatedly palpated and digitized. However, in some joints, such as the wrist, palpating the necessary bony landmarks can be challenging due to the complex bony anatomy and overlying soft tissues, which make the landmarks inaccessible. Such challenges can lead to inaccurate digitization, introducing error into the generated coordinate systems, confounding robot control, and the interpretation of the data.

To address these challenges, imaging and motion capture methods have been used in conjunction with robotic testing for kinematic data acquisition. Fraysse et al. measured individual carpal bone and wrist kinematics on a Stewart platform (hexapod robot) using bone-pins with reflective marker-sets, recording kinematic data using an optical motion capture system [12]. Limitations of this approach include measurement errors due to marker dropout and bone-pin movement due to surrounding tissues. Chokhandre et al. used water-based registration marker sets as fiducial markers in magnetic resonance imaging images to align robot testing space to imaging space [14]. The specimen coordinate system was established in robot space by digitizing palpable knee landmarks, and the test data were transformed to imaging space after mechanical testing. While a robust approach in specific applications, the use of water-filled markers still requires direct access to the bony landmarks. A method that does not rely on manual landmark digitization would be useful.

In addition to the importance of coordinate system definition, selection of an appropriate control protocol is an essential requirement of the experimental study design. The typically used methods of simple load or displacement control are not well suited to the nonlinear behavior of cadaver joints. For example, small displacements in high stiffness regions can produce large changes in loads, whereas small changes in applied loads in low stiffness regions can produce large displacements. Previously, hybrid load–displacement protocols have been successfully used to address these limitations in knee [21] and spine [10,22] testing. Test protocols developed to reproduce in vivo motion of the ovine stifle using parallel robot system addressed some of the issues associated with using displacement control [23]. In vitro testing of wrist has commonly involved manipulating wrist movements by applying forces through tendons [24–30]. Other studies have used custom specimen fixturing and standard materials testing systems to apply pure moments to rotate the wrist without artificially constraining its response [1,31,32]. Carpal kinematics has been studied using a hexapod, but limitations of the displacement control strategy used were not addressed [12]. In that study, the wrist was rotated about a fixed center of rotation by restricting joint coordinate translations, which likely altered joint loads as the wrist was rotated around nonphysiological axes. The passive stiffness of the wrist joint is known to be nonlinear and to depend on the direction of motion, being extremely compliant about the neutral position but much stiffer at rotations approaching the limits of the ranges of motion [1,33]. A more appropriate control strategy to address these challenges would be a hybrid load–displacement method that prescribes wrist motions while simultaneously minimizing joint loads. Hybrid control would enable the robot to measure and respond to joint loads while mimicking natural wrist movements.

To address these issues, in this paper, we describe an imaging-based methodology for coordinate system generation and a hybrid test methodology for implementing robotic testing in the upper extremity. The methodology is demonstrated with robotic testing of intact human cadaveric wrist specimens.

Methods

Specimen Preparation.

Nine upper extremity specimens (right handed, male, average age: 58 yr old) were acquired from a licensed nontransplant tissue provider. The specimens were transected distal to the elbow and at the metacarpophalangeal joints. In each specimen, two Kirschner wires were used to fix the forearm in neutral pronosupination under fluoroscopic guidance. The flexor carpi radialis, flexor carpi ulnaris, extensor carpi ulnaris, and extensor carpi radialis longus and brevis were isolated for later attachment to constant force springs. The radius, ulna, and the metacarpals were potted in fast setting urethane resin (SmoothOn, Macungie, PA), such that the dorsal surface of the hand was aligned with the dorsal surface of the forearm at the time of potting. After the resin set, six blind holes were drilled into each of the resin blocks (3.5 mm dia., x ∼ 10 mm deep) to be used as fiducial markers to facilitate registration of anatomical points from computed tomography (CT) global coordinate space (CT_GCS) to the robot's global coordinate system (ROB_GCS). The fiducial marker holes were drilled to align with the locations of a series of noncoplanar holes in the robot mounting fixtures, providing access for the system's digitizing probe.

Image Acquisition and Processing.

Each potted specimen was imaged using a clinical CT scanner (HighSpeed Advantage 16, GE^®, Milwaukee, WI). Image resolution was 0.4 mm × 0.4 mm in-plane and 0.625 mm along the Z-axis. The radius (RAD), second metacarpal (MC2), third metacarpal (MC3), fourth metacarpal (MC4), and the fiducial marker holes were segmented semi-automatically using MIMICS v19 (Materialise^®, Leuven, Belgium), and 3D-meshed surface models were exported. Coordinates (x, y, z) of the anatomical landmarks for RAD, MC2, MC3, MC4, and their respective fiducial markers (RF1-RF6 and MF1-MF6, respectively), were digitized from the exported models in the CT_GCS (Fig. 1) using Geomagic Studio v12 (3D Systems^®, Rock Hill, SC).

Fig. 1.

Three-dimensional wrist bone models in CT global coordinate space (CT_GCS: X_CT, Y_CT and Z_CT), with landmarks for anatomical coordinate system definition (red) and fiducial markers (blue) for the radius and metacarpals. Anatomical landmarks on the radius included: SN—base of concavity of sigmoid notch; RS—radial styloid; RC—most distal point on line fit through centroids of radial diaphysis (shaft); and RI—radius intersection, the projection of RC on distal articular surface of radius. For the metacarpals, the anatomical landmarks included: M1—the distal-most point on the line fit through centroids of MC₃; M2—the proximal-most point on the line fit through centroids of MC₃; M3—the centroid at the base of MC₂; and, M4—the centroid at the base of MC₄. The fiducial points are labeled RF1 to RF6 for the radial potting and MF1 to MF6 for the metacarpal potting.

Robotic Test Setup.

A serial linkage, six-axis industrial robot (KUKA KR 6 R700, Augsburg, DE) was used for testing. The robot was integrated with a multi-axis load cell at its end effector (ATI, Apex, NC) and was controlled by simVITRO musculoskeletal simulator software (Cleveland Clinic, Cleveland, OH, see Tables 1 and 2 for robot and load cell specs, respectively) [7]. Stiffness of the KUKA KR6 series models (6 kg payload) was determined by the manufacturer via the application of 100 N loads to the end effector of an R900 model robot, which has longer second and fourth links than the R700 model (by 85 mm and 55 mm, respectively). The stiffnesses of the R900 are: K_X ≈ 180 N/mm, K_Y ≈ 116 N/mm, and K_Z ≈ 131 N/mm in each of the three orthogonal directions. The R700 model, which was used in this study, is stiffer than the R900 model given its shorter second and fourth links. The stiffness of the custom mounting fixtures, including robot base and end-effector clamps was approximately ∼230 N/mm. The overall stiffness of the test system is two orders of magnitude stiffer than the wrists tested in this study.

Table 1.

KUKA Robot KR 6 R700 Technical Specifications

Features	Data
Max. reach	706.7 mm
Max. payload	6 kg
Pose repeatability (ISO 9283)	±0.03 mm
Number of axes	6
Weight	50 kg
Volume of working envelope	1.36 m3
Axis 1	ROM: ±170 deg, Speed: 360 deg/s
Axis 2	ROM: +45 deg to –190 deg, Speed: 300 deg/s
Axis 3	ROM: +156 deg to –120 deg, Speed: 360 deg/s
Axis 4	ROM: ± 185 deg, Speed: 381 deg/s
Axis 5	ROM: ± 120 deg, Speed: 388 deg/s
Axis 6	ROM: ± 350 deg, Speed: 615 deg/s
Cycle time (1 kg Payload)	138 cycles/min

Table 2.

 Six-DOF force-torque sensor (ATI load cell SI-580-20) technical specifications

Sensing ranges	FX, FY	FZ	TX, TY	TZ
	580 N	1160 N	20 N·m	20 N·m
Resolution	FX, FY	FZ	TX, TY	TZ
	¼ N	¼ N	1/188 N·m	1/376 N·m
Stiffness	KX, KY	KZ	KTX, KTY	KTZ
	7.4 × 107 N/m	9.8 × 107 N/m	1.7 × 104 N·m/rad	3.5 × 104 N·m/rad

Each specimen was thawed for 10–12 h prior to testing. The proximally potted forearm was mounted to the pedestal, rigidly affixed to the robot base, and the distally potted metacarpals were mounted to the robot end effector (Fig. 2). The four isolated tendons were attached to constant force springs (4.45 N each, McMaster-Carr, Robbinsville, NJ) whose magnitudes were consistent with the loads reported in tendon excursion and gliding studies [34]. After mounting, a 6DOF digitizing probe (Optotrak, NDI, Waterloo, ON, Canada; 3D accuracy: 0.1 mm and resolution: 0.01 mm) was used to establish the position of load cell and the specimen in the robot's global coordinate system (ROB_GCS). The fiducial drill holes were digitized in the ROB_GCS and a transformation matrix from CT_GCS to ROB_GCS was computed within simVITRO to register the anatomical landmarks via iterative closest point algorithm [35].

Fig. 2.

Overview of the KUKA robot wrist test configuration with a mounted specimen and its coordinate systems. The wrist specimen is mounted to the robot in a horizontal position to enable testing about the complete ROM envelope. The radius is mounted robot's base plate via a raised pedestal and the metacarpals are mounted to robot end-effector. The metacarpal coordinate system (X_M, Y_M, and Z_M) and the radial coordinate system (X_R, Y_R, and Z_R) are defined within the robot test space (ROB_GCS: X_ROB, Y_ROB, and Z_ROB) through registration of the fiducial marker points.

Radial and metacarpal coordinate systems (MCS, Fig. 2) were defined by simVITRO in the ROB_GCS using the registered anatomical landmarks (Fig. 1). The radius-based coordinate system was constructed as: Z_R—positive radially, defined by SN and RS; Y_R—positive proximally, along the long axis of the radial shaft, defined by RI and RC; X_R—positive radially, generated by crossing Y_R and Z_R; and O_R—the point on Z_R closest to Y_R The MCS was constructed as: Y_M—positive proximally, defined by M₁ and M₂; Z_M—positive radially, defined by M₃ and M₄; X_M—Positive volarly generated by crossing Y_M and Z_M, and the origin, O_M defined as the midpoint of M₃ and M₄.

Hybrid Testing Methodology.

Our test strategy is based on Raibert and Craig [36], who combine force and torque information with positional data to simultaneously comply with the specified position and force trajectory constraints for a particular task. We apply this method to the wrist joint, where torque constraints are overlaid on the prescribed rotational position of the wrist.

After mounting and digitizing, a joint coordinate system (JCS) was constructed within simVITRO, describing the position of MCS with respect to the radius-based coordinate system, consistent with ISB recommendations [37]. The JCS was constructed with the flexion–extension axis (Z) fixed in the radius, the pronation–supination axis (Y) fixed in metacarpals, and the radio ulnar deviation axis (X) allowed to float. The calculated rotations and translations of the JCS define the offset of the metacarpals relative to the radius. The offset was then refined to bring the wrists to the neutral position, as follows: (1) flexion–extension and radial-ulnar deviation were set to 0 deg in position control; (2) pronation/supination was set to 0 N·m in load control; and (3) all three joint forces (volar/dorsal, radial/ulnar and compression/distraction) were set to 0 N in load control. This ensured the wrist was in an anatomical neutral position, as well as in a near-zero force/torque state.

The wrists were rotated from the neutral position to the extremes of 24 distinct directions of wrist motion: the classic orthogonal directions of flexion, extension, radial deviation, ulnar deviation, plus 20 coupled directions at increments of 15 deg from the orthogonal directions (e.g., ulnar flexion, ulnar extension, radial flexion, and radial extension). For example, a direction of 45 deg in ulnar flexion would be equally composed of flexion and ulnar deviation. Pronation/supination was maintained at 0 deg for during motion in all 24 wrist directions. The wrists were rotated in each direction at a rate of 1 deg/s to the point where the resultant torque was 2 N·m. The resultant torque was calculated as the vector sum of the three torque components reported in the MCS and was defined to be the terminating criteria, representing the total effective joint torque generated from all three rotational degrees-of-freedom (DOF). Joint loading was controlled to 0 N to avoid constraining the location of the rotation axes, thus permitting physiological motions. All specimens were preconditioned for two cycles from 50 deg flexion to 50 deg extension. Data were acquired digitally at rate of 50 Hz. Finally, 6 trajectories from total of 216 trajectories (9 specimens' × 24 directions) were terminated at torque values between 1.5 N·m and 2 N·m due to impingement of the robotic segments in those directions.

Data Analysis.

Wrist kinematics was reported with the Z–X–Y Cardan-angle sequence. The torque–rotation curves were analyzed to determine range of motion (ROM), neutral zone (NZ), and stiffness (KROM) in each direction. To do so, the loading portion of torque–rotation data was fit with a fifth-order polynomial (MATLAB, Mathworks^®, Natick, MA). Wrist rotations were calculated using Pythagorean theorem $\sqrt{(\mathrm{F}{\mathrm{E})}^2+(\mathrm{RU}{\mathrm{D})}^2}$ as pronation–supination was set to 0 deg. The ROMs and NZs were then defined as the wrist rotations at torque values of 2 N·m and 0.5 N·m, respectively. A few tests were terminated between 1.5 N·m and 2 N·m due to limitations in the robot's motion envelope. In these cases, the final rotation values were extrapolated from the fitted polynomial curve. For KROM calculations, the final tangential stiffness was calculated as the slope between points of the fitted polynomial curve at 1.5 N·m and 2 N·m.

The ROM, NZ, and KROM (mean and ±Std Dev) are presented as polar plots of the 24 directions of wrist motions. The sequential sets of these values defined the ROM, NZ, and KROM “envelopes” of the wrist. The major principal axes of the sets described the primary orientation of the envelopes [1]. Unpaired t-tests were used to determine if the orientation of the major principal axis was different from the anatomic flexion/extension axis (defined as 0 deg). A one-way analysis of variance with Holm-Sidak posthoc analysis (Sigmaplot v14, Systat, San Jose, CA) was used to determine if the maximum ROM, NZ, and KROM values differed for the four principal directions of flexion–extension and radial-ulnar deviation. Statistical significance was set at p = 0.05.

Results

The ROM envelopes for the wrists were generally elliptical in shape with major principal axis oblique to the anatomical axes of flexion–extension (Fig. 3). The mean orientation of the major principal axis of the ROM envelope was angled 12.1 ± 2.7 deg toward ulnar flexion, which was significantly different (p < 0.001) from pure flexion/extension. The largest wrist ROM was 98 ± 9.3 deg in the direction of ulnar flexion (15 deg from pure flexion), which was significantly greater than the ROM in pure flexion (75.3 ± 17.5 deg, p < 0.001) or pure extension (82.7 ± 9.6 deg, p = 0.017). Mean maximum ulnar and radial deviation ROMs were 41.9 ± 5.9 deg and 25.3 ± 5.9 deg, respectively. The neutral zone envelope was oriented by 11.5 ± 3.6 deg toward radial extension-ulnar flexion and significantly different (p < 0.001) from pure flexion/extension (Fig. 4). The largest neutral zone was ROM 74.4 ± 12.1 deg in the direction of ulnar flexion (15 deg from flexion). This was significantly larger than the neutral zone values for pure flexion (45 ± 13.5 deg, p < 0.001), radial deviation (15.5 ± 5.1 deg, p < 0.001), and ulnar rotation (33.4 ± 6.1 deg, p < 0.001), though it was not significantly different than extension (65.2 ± 7.5 deg, p = 0.105).

Fig. 3.

Mean (±1 SD) passive ROM of the wrist at 2 N·m applied torque. The ROM envelope described by its major principal axis was oriented oblique to anatomical flexion–extension axes from radial extension to ulnar flexion.

Fig. 4.

Mean NZ envelope of the wrist at 0.5 N·m, which was similar to the mean range of motion envelope (Fig. 3)

The stiffness envelope was oriented orthogonal to the ROM envelope, with the major principal axis oriented from ulnar extension to radial flexion, –80.6 ± 5 deg from flexion (Fig. 5). The wrists were stiffest in the direction of radial flexion, with a mean of 0.32 N·m/deg, while they were least stiff in the direction of ulnar flexion with a mean of 0.11 N·m/deg.

Fig. 5.

Mean (±1 SD) envelope of wrist stiffness (KROM) was oriented approximately orthogonal to the range of motion and NZ envelopes, with greatest stiffness in the direction of radial flexion

The components of the resultant torque recorded to move the wrists along each of the 24 motion paths were not proportional with the direction of motion but were instead dominated by radial and ulnar torque components (Fig. 6).

Fig. 6.

Components of the mean torque envelopes required to manipulate the wrist specimens in each of the 24 directions of wrist motion: (a) flexion and extension torque, (b) radial and ulnar torque, and (c) pronation and supination torque

Discussion

The objectives of this study were to describe an image-based methodology for coordinate system generation and a hybrid control strategy for robotic testing of the wrist. The image-based coordinate system methodology was developed to accurately generate coordinate systems in the wrist because its multiple bones and complex morphology make direct digitization of bony landmarks extremely challenging. The hybrid control strategy for determining wrist mechanical properties was developed because the wrist is dominated by two rotational DOF with large neutral zones that are directionally dependent.

Previous robotic studies [6,12,14] have utilized CT scans and motion capture technology to estimate anatomic pose before testing, or to register testing coordinate systems to anatomy after data acquisition. In our approach, anatomical coordinate systems are registered into the robot space and anatomic manipulations are based directly on CT-generated anatomic coordinate systems. Joint kinematics is calculated directly in the robot software, obviating the need for motion capture technology for data acquisition. Three-dimensional digitizers are only required for fiducial marker digitization to register the anatomical points from imaging space. This approach is not restricted to the wrist and could easily be adapted for use in other joints (e.g., knee, spine) and with other imaging modalities (e.g., magnetic resonance imaging). We used drill holes geometry as our fiducial markers, however, any imageable fiducial markers that do not produce artifacts (e.g., tantalum beads with CT images) could be used to register anatomical points from the imaging modality coordinate system to the robotic coordinate system. The location of the fiducial markers can be chosen randomly, provided they can be easily accessible and digitized. Accessibility of the fiducials is more important than their location, as long as they are rigidly fixed to the corresponding bone.

Previous in vitro studies of the wrist adopted either load or position control strategies, as the tests were based on achieving wrist motion by applying forces on muscles and tendons [24–28,30] or used specially designed test setups with limited degrees-of-freedom [1,31,32]. A recent robotic wrist study employed position control with restricted range of motion (<20 deg) in both flexion–extension and radio ulnar directions [12]. Wrist translations were not permitted, constraining normal physiological kinematics, which may have resulted in increased joint loads. Constraining wrist rotation axes can result in induced loading. In an Instron-based wrist ROM study, Crisco et al. [1] reported axial loads of ±20 N as the actuator was held constant in a fixed position at wrist neutral position. In contrast, our approach was designed to minimize forces by allowing translations. Wrist joint translations are reported to be minimal during normal motion [38–40]; however, the axis of rotation changes with direction of wrist rotations [41–43]. To characterize the wrist joint throughout its complete range of motion, we controlled all three rotational DOFs in position control for 24 wrist directions of wrist ROM testing. We chose to keep wrist pronation/supination fixed at 0 deg as studies have shown varying amount of pronation/supination based on the activity [44–47]. In previous wrist studies, the specimens were unmounted, remounted, and manually realigned before testing in each direction [1,31]. Our CT-based method allowed us to define an anatomical neutral position for all specimens before they were moved sequentially in the 24 directions without repositioning. This improves the repeatability of neutral position definition in repeated-measures study designs (e.g., intact, after ligament transection, device implantation, etc.), reducing the potential of variation in the data. We confirmed neutral pose repeatability by unmounting, remounting, and digitizing a single specimen (male, right wrist, age <60) ten times. The standard deviation in the neutral position was 0.01 deg in flexion–extension, 0.01 deg in radio ulnar deviation and 0.02 deg in pronation–supination; 0.77 mm in volar-dorsal translation, 0.71 mm in radio ulnar translation, and 0.33 mm in proximal–distal translation.

The ability to record individual torques contributing to wrist motion is another advantage over previous studies [1]. This helps in evaluating the role of each rotational DOF during specific wrist motions. We did observe an average resultant torque of 0.1 N·m at neutral position across all directions and specimens, which was ∼5% of our termination criteria of 2 N·m and within our tolerance limit. We monitored wrist torques in all three rotational DOF and terminated our motion trajectories at resultant torques of 2 N·m. Interestingly, the components of the applied torques were not proportional to the direction of wrist motion (Fig. 6). For example, if the wrist was a true universal joint (i.e., two pure rotational DOFs), motions in the direction of wrist flexion and radial deviation would generate pure flexion and radial deviation torques, respectively. Further studies will be required to understand the complexities of the torque envelopes recorded in Fig. 6.

There are few studies of the complete ROM of the wrist that can be compared directly to our results. Our average wrist ROM data of 75.3 deg in flexion, 82.3 deg in extension, 41.9 deg in ulnar rotation and 25 deg radial deviation is similar to intact wrist data from the literature [1,2]. Our finding of the largest mean ROM of 174 deg from radial extension to ulnar flexion is close to the 178 deg previously reported by Crisco et al., who performed wrist ROM testing using a standard biaxial materials testing system [1]. Our wrist motion envelope depicting an oblique orientation from radial extension to ulnar flexion is in general agreement with previous studies [1,48], and is consistent with the direction of wrist motion referred to as the dart thrower's motion [49]. Our stiffness findings at extremes of motion is similar to values reported in a previous study [1], and the overall shape and envelope oriented from ulnar extension to radial flexion is also consistent with previous findings [1,33]. We attribute the small differences in data between studies to discrepancies in study protocol and measurements taken at different wrist rotations. The overall consistency with data from the literature [1,2,33,48] provides confidence in our hybrid control strategy and its use for future robotic wrist studies. The robot used in the study is a 6DOF serial linkage manipulator. Serial robots although less stiff than parallel robots offer larger working volume, and this is especially relevant to the work in the wrist. Wrist joints have nonlinear stiffness with large neutral zone and ROM of ∼130 deg in flexion–extension and ∼40 deg in radio ulnar deviation [50].

Limitations of our study include not being able to distinguish midcarpal and radiocarpal motions from overall wrist motion, as we measured the position of the metacarpals relative to radius. Midcarpal and radiocarpal motions can be measured using additional marker sets or using motion capture systems in conjunction with the robotic testing. Our measurements were also limited to neutral forearm rotation, and therefore, it remains unclear if other forearm positions of supination and pronation would alter our results. In a recent study, wrist range of motion in neutral, full pronation, and full supination was not found to be significantly different [31]. Finally, we chose all right-sided male specimens for the testing. Due to the differences in bone size between men and women, it is possible that the magnitude of the values we measured may differ with sex. However, the methodology described here would not be affected.

The imaging-based methodology in this study provides an efficient approach to generate specimen-specific anatomical coordinate systems for in vitro robotic testing and is easily adaptable to any other anatomical joint and imaging modality for future experiments. The described hybrid control strategy for wrist testing presents a multidimensional control for the study of wrist biomechanics—a highly lax joint—in all six degrees-of-freedom, allowing for measurements of all three rotations and translations. Future studies can incorporate this methodology and strategy to advance our understanding of the complexities of healthy wrist mechanics and the effects of trauma, pathologies, and wrist implants.

Funding Data

• National Institute of Arthritis and Musculoskeletal and Skin Diseases of the National Institutes of Health (Award No. P30-GM122732 (Bio-engineering Core); Funder ID: 10.13039/100000069).