In Vivo Quantification of Hip Arthrokinematics during Dynamic Weight-bearing Activities using Dual Fluoroscopy

Summary

Dual fluoroscopy accurately captures in vivo dynamic motion of human joints, which can be visualized relative to reconstructed anatomy (e.g., arthrokinematics). Herein, a detailed protocol to quantify hip arthrokinematics during weight-bearing activities of daily living is presented, including the integration of dual fluoroscopy with traditional skin marker motion capture.

Abstract

Several hip pathologies have been attributed to abnormal morphology with an underlying assumption of aberrant biomechanics. However, structure-function relationships at the joint level remain challenging to quantify due to difficulties in accurately measuring dynamic joint motion. The soft tissue artifact errors inherent in optical skin marker motion capture are exacerbated by the depth of the hip joint within the body and the large mass of soft tissue surrounding the joint. Thus, the complex relationship between bone shape and hip joint kinematics is more difficult to study accurately than in other joints. Herein, a protocol incorporating computed tomography (CT) arthrography, three-dimensional (3D) reconstruction of volumetric images, dual fluoroscopy, and optical motion capture to accurately measure the dynamic motion of the hip joint is presented. The technical and clinical studies that have applied dual fluoroscopy to study form-function relationships of the hip using this protocol are summarized, and the specific steps and future considerations for data acquisition, processing, and analysis are described.

Introduction

The number of total hip arthroplasty (THA) procedures performed on adults aged 45-64 years suffering from hip osteoarthritis (OA) more than doubled between 2000 and 2010¹. Based on the increases in THA procedures from 2000 to 2014, a recent study predicted that the overall number of yearly procedures may triple over the next twenty years². These large increases in THA procedures are alarming considering that current treatment costs exceed $18 billion annually in the United States alone³.

Developmental dysplasia of the hip (DDH) and femoroacetabular impingement syndrome (FAIS), which describe an under- or over-constrained hip, respectively, are believed to be the primary etiology of hip OA⁴. The high prevalence of these structural hip deformities in individuals undergoing THA was initially described more than three decades ago⁵. Still, the relationship between abnormal hip anatomy and osteoarthritis is not well understood. One challenge to improving the working understanding of the role of deformities in the development of hip OA is that abnormal hip morphology is very common amongst asymptomatic adults. Notably, studies have observed morphology associated with cam-type FAIS in approximately 35% of the general population⁶, 83% of senior athletes⁷, and more than 95% of collegiate male athletes⁸. In another study of female collegiate athletes, 60% of participants had radiographic evidence of cam FAIS, and 30% had evidence of DDH⁹.

Studies demonstrating a high prevalence of deformities amongst individuals without hip pain point to the possibility that morphology commonly associated with FAIS and DDH may be a natural variant that only becomes symptomatic under certain conditions. However, the interaction between hip anatomy and hip biomechanics is not well understood. Notably, there are known difficulties with measuring hip joint motion using traditional optical motion capture technology. First, the joint is relatively deep within the body, such that the location of the hip joint center is difficult to both identify and track dynamically using optical skin marker motion capture, with errors on the same order of magnitude as the radius of the femoral head^10,11. Second, the hip joint is surrounded by large soft tissue bulk, including subcutaneous fat and muscle, that moves relative to the underlying bone, resulting in soft tissue artifact^12,13,14. Finally, using optical tracking of skin markers, kinematics are evaluated relative to generalized anatomy and thus do not provide insight into how subtle morphological differences might affect the biomechanics of the joint.

To address the lack of accurate kinematics in combination with subject-specific bone morphology, both single and dual fluoroscopy systems have been developed for analyzing other natural joint systems^15,16,17. However, this technology has only recently been applied to the native hip joint, likely due to the difficulty in acquiring high-quality images through the soft tissue surrounding the hip. The methodology to accurately measure in vivo hip joint motion and display this motion relative to subject-specific bone anatomy is described herein. The resulting arthrokinematics provide an unparalleled ability to investigate the subtle interplay between bone morphology and biomechanics.

Herein, the procedures for acquiring and processing dual fluoroscopy images of the hip during activities of daily living have been described. Owing to the desire to capture whole-body kinematics with optical marker tracking simultaneously with dual fluoroscopy images, the data collection protocol requires coordination between several sources of data. Calibration of the dual fluoroscopy system utilizes plexiglass structures implanted with metallic beads that can be directly identified and tracked as markers. In contrast, dynamic bone motion is tracked using markerless tracking, which utilizes only the CT-based radiographic density of the bones to define orientation. Dynamic motion is then tracked simultaneously using dual fluoroscopy and motion capture data that are spatially and temporally synced.

The systems are synced spatially during calibration through concurrent imaging of a cube with both reflective markers and implanted metal beads and the generation of a common coordinate system. The systems are synced temporally for each activity or capture through the use of a split electronic trigger, which sends a signal to end the recording of the dual fluoroscopy cameras and interrupts a constant 5 V input to the motion capture system. This coordinated protocol enables the quantification of the position of body segments that fall outside the combined field of view of the dual fluoroscopy system, expression of kinematic results relative to gait-normalized events, and characterization of the soft tissue deformation around the femur and pelvis.

Protocol

Procedures outlined in this protocol were approved by the University of Utah Institutional Review Board.

1. CT arthrogram imaging

1. Arthrogram¹⁸
   1. Schedule a trained musculoskeletal radiologist to perform the arthrogram directly prior to the scheduled CT imaging.
   2. Position the participant on the table with the hip of interest in the field of view of a clinical fluoroscope. Place sandbags on either side of the ankle to prevent rotation of the leg and hip.
   3. Prepare the skin to create a sterile environment. Mark the location where the needle will be inserted (femoral head-neck junction) and anesthetize the soft tissue at the injection site with 2-5 mL of 1% lidocaine.
   4. Prepare a solution of 20 mL of 1% lidocaine, 10 mL of iohexol injection, and 0.1 mL of 1 mg/mL (1:1000) epinephrine in a 30 mL luer lock syringe.
   5. Two to five minutes after the lidocaine injection, insert a spinal needle just until it contacts the femoral neck; verify the location of the needle by fluoroscopy. Inject a small amount of the prepared solution (<5 mL) and ensure that the injected fluid is contained within the joint capsule with an image from fluoroscopy.
   6. Inject 20-30 mL of the contrast mixture. When additional resistance to the injection is observed, have a study team member manually apply traction to the hip by pulling on the participant's ankle while the participant grasps the headboard of the table to resist upper body movement. Inject the remaining contrast mixture, as appropriate.
   7. Verify by fluoroscopy that the contrast agent fills the joint space and covers the femoral head when traction is applied.
   8. Transfer the patient to the CT scanner in a wheelchair or bed to minimize the loss of contrast within the joint capsule.
2. Traction and CT imaging
   1. Help the participant into a supine position on the CT gantry.
   2. Place the hare traction splint device under the leg of interest, ensuring that the proximal padded bar rests just distal to the ischium. Attach the hook and loop straps around the thigh and ankle of the participant and apply light traction.
   3. Acquire a scout image and set the field of view to include the entire pelvis and proximal femurs to just below the lesser trochanter for the hips. Set a separate field of view to include the distal femurs and proximal tibias for the knees.
   4. Apply additional traction (have one member of the research team pull on the ankle while another tightens the strap of the hare traction splint) to ensure separation of the joint space. Acquire images at 120 kVp, 1.0 mm slice thickness, 200 - 400 mAs for the hip and 120 kVp, 3.0 mm slice thickness, and 150 mAs for the knees. Use CARE Dose, an automated exposure control that modulates tube current according to image quality, to minimize the radiation burden to the participant.
   5. Release and remove the hare traction splint device. Assist the participant to a standing position and ensure they feel comfortable putting weight and being mobile on the limb before allowing them to leave.

2. Dual fluoroscopy imaging

1. System setup
   1. Apply anthropometrics¹⁹ to estimate the height of the hip joint based on the participant's reported height and use this measurement to estimate the desired height of the center of the field of view of the system.
   2. Position the image intensifiers approximately 50° from one another on the side of the instrumented treadmill corresponding to the hip of interest (Figure 1).
   3. Position the X-ray emitters to be pointed towards the image intensifiers. Ensure that the distance between the emitter source and the face of the image intensifiers is approximately 100-110 cm.
      NOTE: The recommended distance between the emitter source and the face of the image intensifiers will vary based on system specification and the collimator in the X-ray emitter.
   4. Connect the center of the face of the image intensifier and the corresponding X-ray emitter of each fluoroscope pair using strings or measuring tapes. Verify that the strings (or tapes) cross at the desired location (i.e., in the expected location of the hip joint).
   5. Affix the plate with three lasers to the emitter and the mirror to the image intensifier. Turn on the lasers and refine the alignment of each emitter and image intensifier based on the reflection of the lasers back to the laser source.
2. Calibration images
   1. Prepare for the use of radiation by donning lead and placing signage on the entrances to the room. Minimize exposure by having staff wear protection that includes a leaded vest, skirt, gloves, and glasses. Turn on the fluoroscopes and allow the systems to warm up, as necessary.
   2. For all calibration images, set the fluoroscopes to 64 kVp and 1.4-1.6 mA, or as otherwise desired.
   3. Open the camera control software on the computer and select the appropriate cameras as slave and master. Use external syncing to the master camera from the slave camera to sync the two cameras.
      NOTE: For all recorded activities, save the same frames from both dual fluoroscopy cameras; frames are identified with a number representing the number of frames prior to the electronic trigger signal.
   4. Verify the alignment of the system by affixing a circular metal washer to the center of the image intensifier and attaching the crosshair fixture to the emitter.
      ​NOTE: Once alignment is verified, it is important to avoid contacting the system.
   5. Attach the plexiglass grid to one of the image intensifiers using screws; minimize the force applied in this process to avoid altering the alignment. Acquire fluoroscopy images and save 100 image frames from each dual fluoroscopy camera of the grid. Remove the grid, and repeat the process for the other image intensifier.
   6. Place the 3D calibration cube within the combined field of view of the two fluoroscopes. To do this, place the cube on a stool or platform that is radio-translucent and visually verify that most or all of the cube is within the field of view. Orient the cube such that the calibration beads do not overlap for either dual fluoroscopy camera view. Acquire images and save 100 image frames of the cube.
   7. Before moving the cube, measure and record the approximate location of the cube's origin from each emitter using the coordinate system of the cube. Remove the cube and any associated platform.
   8. Measure and record the distance between the emitter source and the face of the image intensifier for each fluoroscope.
   9. Attach the beaded plexiglass to a long rod or ruler with a rubber band and move it randomly to provide movements ranging the entire field of view of the system. Ensure that the research staff is mindful of the path of radiation and wear protection to minimize exposure (see step 2.2.1). Save 100 image frames of the motion.
   10. Reset the imaging clock used to track exposure time.
3. Static trial and adjustment of parameters
   1. Measure the height of the greater trochanter to ensure that the system height is appropriate for the participant.
      1. Palpate the thigh to find the bony prominence of the greater trochanter and locate the most superior point, as is possible.
      2. As the superior greater trochanter is approximately at the same height as the hip joint, measure the height from the floor to this point and compare it to the height estimation used to set up the dual fluoroscopy system.
      3. If necessary, adjust the system height and recalibrate while the participant is being prepped for data capture.
   2. Familiarize the participant with the fluoroscopy system and inform them that they must notify the research team if they come into contact with any of the equipment during the imaging session, as contact with the system negatively affects the accuracy of their data.
   3. Have the participant step onto the treadmill and stand within the field of view of the dual fluoroscopy system. Check participant alignment from the perspective of each emitter and take note of this position from the perspective of where each member of the research team will be standing or sitting during imaging.
   4. Estimate the imaging parameters (kVp and mA of each emitter and the exposure of the dual fluoroscopy cameras) based on the body mass index (BMI) of the participant and set each fluoroscope accordingly.
      NOTE: For the referenced cohort, fluoroscopy settings ranged from 78 to 104 kVp and 1.9-3.2 mA with camera exposures of 4.5-7.0 ms.
   5. Acquire images of the participant during standing and assess the images for contrast and field of view.
      ​NOTE: Increased kVp is associated with increased X-ray scatter (increases noise and reduces contrast), lower image resolution, and lower contrast.
   6. Adjust the parameters and/or participant alignment and repeat image acquisition, as necessary.
   7. Save 100 frames of the final images to use as a static trial.
4. Dynamic trials (Figure 2)
   1. Prior to the start of the dual fluoroscopy imaging, have the participant walk a known distance while being timed. Use this to determine the self-selected walking speed for both level and incline walking on the treadmill.
   2. Have the participant don a leaded thyroid collar to protect the thyroid.
   3. During dynamic acquisitions, have the researcher manning the dual fluoroscopy camera control at the dual fluoroscopy workstation step behind the lead shield and watch the participant through the viewing window of the shield (Figure 3).
   4. For the performance of all walking trials:
      1. Inform the participant prior to starting the belt of the treadmill. Ramp the speed of the treadmill up to the appropriate walking speed and let the participant's gait normalize prior to collecting images.
      2. For each walking activity, acquire and save at least two full gait cycles.
      3. For the inclined walking activity, have the participant step off the treadmill. Unlock the treadmill, set the incline to 5°, and relock the treadmill before having the participant step back onto the treadmill to perform the activity.
      4. Repeat the imaging, such that the activity is recorded twice.
      5. Repeat the same process (step 2.4.4.3) to lower the treadmill upon completion of the activity.
   5. For the pivot activities:
      1. Have the participant rotate their body position and feet approximately 45° from the front of the treadmill opposite of the direction of the pivot. If desired, ensure that each foot is placed entirely on a single belt of the dual-belt treadmill to allow straightforward processing of the force plate data.
      2. Have the participant perform several pivots to and from their end range of motion while watching for the alignment of the pelvis at the end range of motion. Ensure that the motion is performed smoothly as the pivot does not require acceleration to achieve the final position.
      3. Based on the position of the pelvis at the end range of motion, have the participant rotate and/or translate their feet such that the pelvis is facing forward on the treadmill and the hip of interest is in the middle of the combined field of view of the fluoroscopes at the end of the pivot.
      4. Once the position is optimized, have the participant perform the pivot during dual fluoroscopy imaging and save all frames where the femur and pelvis are visible in both dual fluoroscopy camera views (approximately 200-400 frames) centered about the end range of motion, capturing as much of the pivot as possible.
      5. Repeat the imaging, such that the activity is recorded twice.
   6. For the abduction-adduction activity:
      1. Have the participant stand in the field of view of the fluoroscopes and raise the leg of interest approximately 45° out to their side. Remind the participant to avoid torso motion and reduce the range of motion, if necessary.
      2. Acquire and save all frames where the femur and pelvis are visible in both dual fluoroscopy camera views (approximately 200-400 frames).
      3. Repeat the imaging, such that the activity is recorded twice.
   7. For the dynamic hip joint center or star-arc activity²⁰
      1. Have the participant stand in the field of view of the dual fluoroscopy system and raise and lower their leg anteriorly and at 45° increments of 180°, ending with a posterior raise and lower of their leg. Prior to placing their leg back down onto the ground, have the participant circumduct their leg and return to a standing position.
   8. Once the participant is comfortable with the motion and can complete it in approximately 6-8 s, acquire and save images of the activity.
      NOTE: Only one activity is captured with dual fluoroscopy due to the length of the trial.
5. Additional calibration images
   1. If at any point during the data collection, the participant believes they may have come into contact with any part of the fluoroscopic equipment, image the grids and cube and save all the files for calibration.
   2. Upon completion of the data collection, image the grids and cube and save all files for calibration to serve as a backup if any issues arise with the initial calibration.

3. Skin marker motion capture and instrumented treadmill

1. System setup
   1. Focus the optical motion capture system on the treadmill (Figure 3). Due to the potential issues with visualizing the participant while in the field of view of the dual fluoroscopy system, be prepared to precisely position the infrared cameras to ensure accurate visualization (Figure 2).
   2. Turn on the system and use a set of markers to ensure that the dual fluoroscopy system does not prevent visualization of the desired field of view.
   3. Check that the markers are clear and circular and adjust the focus of the infrared cameras, as necessary.
   4. Ensure that the fluoroscopes are covered to reduce any reflective surfaces. Review each infrared camera and mask the camera view if the reflective objects cannot be covered.
   5. Set up the motion capture software to read in an external 5 V signal from the electronic trigger used to end camera acquisition of the dual fluoroscopy system. Use this trigger to temporally sync the data from the two systems.
2. Calibration
   1. Once the system is on and ready, use the active calibration wand to simultaneously calibrate the optical and infrared motion capture cameras. Ensure that the entire region within the dual fluoroscopy system is thoroughly captured during the calibration while avoiding contact with any equipment.
      NOTE: Wand motions resembling tossing food in a frying pan have worked well.
   2. Due to the obstructions caused by the dual fluoroscopy system, calibration values may be worse than usually observed for optical motion capture. Perform the calibration so that all the infrared cameras have image errors less than 0.2.
      NOTE: The image error for the video camera will be higher, although still less than 0.5. The video camera is not specifically used for any quantification of motion, only for visual recording of the motion capture.
   3. During the acquisition of the cube trial for dual fluoroscopy, also capture the cube with the motion capture infrared cameras. Ensure that the cube has reflective markers affixed to it for the position to be imaged with cameras from both the motion capture and dual fluoroscopy systems.
3. Marker set and placement
   1. Before the arrival of the participant, cut and apply double-sided tape (toupee tape) to the base of 21 spherical reflective skin markers. To ensure the longevity of the markers, ensure that the tape or any skin does not come in contact with the reflective markers.
   2. For each of the five marker plates (two on the shank, two on the thigh, one on the back; Figure 4), apply spray glue to the skin side of the fabric strap and wrap it tightly around the participant. Check with the participant that the straps feel tight (but are not uncomfortable). Clean hands of any excess spray glue before adhering the rest of the marker set.
   3. Apply five markers, used only for calibration, to the clavicle, medial knees, and medial malleoli, respectively.
   4. Apply the remaining 16 markers to the anterior superior iliac spines (ASIS), posterior superior iliac spines (PSIS), greater trochanter of the femur being imaged, shoulders, sternum, lateral knees, lateral malleoli, and feet (Figure 4).
   5. Ask the participant to inform the study team if any markers or straps become loose during the data capture.
4. Static trial
   1. In conjunction with the static standing trial from dual fluoroscopy, capture a standing trial for motion capture.
   2. Label all markers. If any markers are not visible by at least three infrared cameras during the acquired static activity, reacquire a static image to ensure that all markers are visible.
   3. Remove the calibration-only markers and have the participant don a thyroid collar to provide radiation protection during the remainder of the data collection.
5. Dynamic trials
   1. For each of the dynamic trials captured with the dual fluoroscopy system, acquire motion capture video, ensuring that the entirety of each dual fluoroscopy video is within the bounds of the motion capture acquisition.
   2. Ensure that the break in the 5 V signal from the electronic trigger of the dual fluoroscopy system is captured within each trial.

4. Image preprocessing

1. CT-based model
   1. Segment the proximal and distal femur of the side of interest and the entire pelvis, as these bones are used for tracking and/or coordinate system generation.
   2. Ensure that the segmentations are representative of the bone shape in all three imaging planes and appear relatively smooth.
      NOTE: The ability to analyze arthrokinematics is dependent on obtaining high-quality reconstructions through careful segmentation.
   3. Convert the image data to Unsigned char (8 bit) and adjust as necessary with offset and scaling to produce an image with a range of 0 to 255.
   4. Isolate only the bone region in the converted image and crop around the bounds of the bone. Record the dimensions of the cropped images.
   5. Save as 2D TIFF format.
   6. Open the image, change the type to 16-bit, and save it as a single 3D TIFF file.
2. Surface reconstruction
   1. Generate surfaces from the segmentation labels, smooth and decimate the surfaces iteratively, ensuring that the faces are never reduced by more than half in any single iteration.
      NOTE: Using the described process, the target number of faces is approximately 30,000 for each proximal and distal femur surface and 70,000 for each hemi-pelvis surface.
   2. Export each surface as a surface mesh in *.vtk format for use as a model file for landmark identification.
3. Landmark identification for the coordinate system
   1. Identify landmarks of the femur for generation of the femoral coordinate system (Figure 5).
      NOTE: The parameters provided below are specific to the referenced dataset and imaging protocols; values may need to be altered to select the landmarks appropriately.
      1. Open the proximal femur as a model file. Open the Post toolbar and Data panel to add a standard field of 1-Princ Curvature, select a smoothness of 10 and then visualize the result. Over-select the faces of the femoral head and use the select range option from the Edit panel to include only negative curvature. Unselect any selected faces that do not belong to the femoral head. Export this femoral head surface as a surface mesh in *.k format for a sphere fit to determine the center of the femoral head.
      2. Using a similar process, apply 1-Princ Curvature to the distal femur with the smoothness of 5 and again select range to include only the faces with negative curvature. Export this femoral condyle surface for a cylinder fit to determine the medial-lateral axis.
      3. Apply 2-Princ Curvature to the distal femur, using a smoothness of 3. Highlight the ridges of the epicondyles and select range using an upper cut-off of -0.1. Export these faces to generate a plane and use it to isolate the faces of the posterior condyles for the cylinder fit.
   2. Identify landmarks of the pelvis for generation of the pelvic coordinate system (Figure 5).
      NOTE: The parameters provided below are specific to the referenced dataset and imaging protocols; values may need to be altered to select the landmarks appropriately.
      1. For each hemi-pelvis, apply 2-Princ Curvature with a smoothness of 5 and select range to include only positive faces to isolate the lunate surface of the acetabulum. Export the lunate surface and use a sphere fit to determine the center of the acetabulum.
      2. Re-apply 2-Princ Curvature with a smoothness of 2 and select all faces with curvature less than -0.15 to highlight the spines of the pelvis. Choose points on the edge of these spines that best represent the ASIS and PSIS as landmarks and record them.

5. Bone motion tracking

1. Calibration
   1. Identify 12 beads within each of the cube images from the dual fluoroscopy cameras (collected in step 2.2.6). Based on the calibrated distances between each of the beads of the cube and the measurements of the location of the cube within the dual fluoroscopy system, determine the spatial orientation of each fluoroscope through minimization of the sum-of-squares projection error between the projected and known bead locations.
   2. Use the grid images to correct for image distortion and apply the correction to all images associated with that grid image.
   3. Use the motion images to quantify the dynamic accuracy of the system and use marker-based tracking to track it.
2. Markerless tracking
   1. Add the location of the selected landmarks to the bone-specific parameters file and collect the dynamic position of these landmarks in the dual fluoroscopy system as output for all tracked frames.
   2. Determine the frames that will be tracked (based on the kinematic data from motion capture, see step 6.1.2) and open the markerless tracking software with the associated bone-specific parameters file.
   3. Select a frame within the desired range with good visualization of the bone, and manually orient the CT-based digitally reconstructed radiograph (DRR) of the bone of interest (either the proximal femur or hemi-pelvis) using the six degrees of freedom available in the software (Figure 6).
      NOTE: As most trials begin in a position similar to standing, this initial position can likely be used as an initial starting point for all trials.
   4. Once the DRR of the bone appears well-aligned in both views, save the solution by clicking the Manual button in the Solutions panel.
      NOTE: Every time a solution is saved, the orientation parameters and the normalized cross-correlation coefficient are plotted for reference. The normalized cross-correlation coefficient is calculated based on all pixels with non-zero values for both the fluoroscope and bone DRRs.
   5. Apply the Diagonal Hessian Search (DHS) optimization step by clicking the DHS button within the Solutions panel and review the result. If the optimized result is preferred, move onto the next frame; otherwise, make any necessary adjustments, and resave by clicking the Manual button within the Solutions panel. Repeat this step until a satisfactory solution is found.
      NOTE: In the case of poor image contrast, the optimization algorithm may not always produce a satisfactory result.
   6. For every fifth frame, repeat this process, using the solution for the previous frame as a starting point. Use the DHS optimization to automate the process.
   7. To complete the first pass of tracking, use another tool that interpolates via linear projection (LP) and optimizes solutions between the tracked frames by clicking the Range of LP + DHS button within the Solutions panel. In the window, enter the set of frames to be tracked and the two frames to be used for reference.
      NOTE: The two reference frames can be any frames within the identified set of frames. However, the use of the first and last frames provides bounds for the orientation of the bones within the frame range, which can be beneficial when contrast is low.
   8. Review and refine each frame of the trial, using both Manual and DHS-based solutions. Use the plot of parameters to ensure that the correlation coefficient is sufficiently high and that the orientation of the bone does not have sudden jumps in any parameter.
   9. To ensure accurate tracking, have another researcher review the solution for each frame and make any necessary modifications to the solutions.
   10. Repeat steps 5.2.1-5.2.9 for each bone.
3. Visualization of motion
   1. Open the femur and pelvis surfaces in the software for kinematic visualization. If necessary, convert the surfaces to meshes using the convert to mesh function. Select both surfaces and export as a surface mesh in *.k format.
   2. Using the output from tracking, generate a text file with the coordinate transformations for each bone and frame.
      NOTE: The order of the surfaces must match the order of the transformations.
   3. For visualization of kinematics, use the kinemat tool and the above two files from steps 5.3.1 and 5.3.2 to animate the kinematics. Verify that the animated kinematics look reasonable and that the surfaces have appropriate distance between them using either a semi-transparent surface or the surface distance tool. If necessary, return to step 5.2.8.

6. Data analysis

1. Skin marker kinematics
   1. Within the motion capture software, batch-process all files to apply the static model and label markers. Once the trial is complete, remove any unlabeled trajectories.
      NOTE: Due to the obstructions of the dual fluoroscopy system, more manual gap filling than usual may be required.
   2. Use the kinematic and force plate data to identify dynamic events, such as toe-off or heel-strike during gait or maximum range of motion for pivoting activities. Determine the frames of interest for tracking of dual fluoroscopy data.
   3. Export all trial data for kinematic processing in *.c3d format, including both analog data (i.e., trigger and force plate data) and marker trajectories.
   4. Apply the desired model template file (saved as *.mdh file format) to the static trial, then assign this model to the motion files.
      NOTE: For analysis, a lower limb model with a generalized International Society of Biomechanics (ISB) head-abdomen-thorax (HAT) segment and the CODA pelvis, a pelvis segment model defined by the two ASIS and the center of the PSIS landmarks, was used.
2. Dual fluoroscopy kinematics
   1. Isolate frames of interest, ensuring that only contiguous frames that are tracked for both the femur and pelvis are included.
   2. Filter landmark positions using a lowpass Butterworth filter (0.12 normalized cut-off frequency from residual analysis and 4^th order filter).
   3. Use the filtered positions of the landmarks throughout each motion trial to track the dynamic position of the femoral coordinate system (Figure 5).
      1. Define the femur origin as the sphere-fit center of the femoral head.
      2. Define the femur z-axis (inferior-superior axis) between the center of the knee and the origin, pointing superiorly.
      3. Define the femur x-axis (medial-lateral axis) as the long axis of a cylinder fitted to the femoral condyles, pointing to the left. To isolate the region of the condyles to be represented with a cylinder, fit a plane to the epicondyle surfaces and isolate the posterior portion of the femoral condyles.
      4. Define the femur y-axis (anterior-posterior) as the cross-product of the defined z- and x-axes, pointing posteriorly. Correct the orientation of the x-axis to create an orthogonal coordinate system.
   4. Use the filtered positions of the landmarks throughout each motion trial to track the dynamic position of the pelvic coordinate system (Figure 5).
      1. Define the pelvis origin as the center of the two ASIS landmarks.
      2. Define the pelvis y-axis (anterior-posterior axis) between the center of the two PSIS landmarks and the origin, pointing anteriorly.
      3. Define the pelvis x-axis (medial-lateral axis) between the origin and the right-side ASIS landmark, pointing to the right.
      4. Define the pelvis z-axis (inferior-superior axis) as the cross-product of the defined x- and y-axes, pointing superiorly. Correct the orientation of the x-axis to create an orthogonal coordinate system.
   5. Generate the rotation matrix between the coordinate systems and calculate joint kinematics per MacWilliams and colleagues' Equation 11 (Figure 7)²¹.
   6. Calculate joint translations by transforming the vector distance between the sphere fit centers of the femoral head and the lunate surface of the acetabulum into the pelvis coordinate system.
      NOTE: This provides a single vector to represent joint translation for each image frame.
3. Arthrokinematics
   1. Visualize the kinematics as described in step 5.3 to animate subject-specific arthrokinematics (Figure 8).
   2. Apply the surface distance data field to measure distances between the femur and pelvis surfaces during each dynamic activity (Figure 8).
      NOTE: These data also provide quantification of the relative distance between joint surfaces but require interpretation to quantify joint translation.
   3. Export surface-to-surface distances using the surface distance tool to quantify data across all participants.
4. Comparison with skin marker motion capture
   1. Using the cube images and trigger from each motion trial, spatially and temporally sync the dual fluoroscopy and motion capture systems.
   2. Transform the landmark locations used for skin marker motion capture (i.e., ASIS, PSIS, condyles) from the markerless tracking coordinate system to the motion capture coordinate system.
   3. Combine these data with the marker locations from skin marker motion capture and import for kinematic and kinetic analysis and reporting. Adjust the analysis to utilize either dual fluoroscopy or skin marker locations for each landmark and compare landmark locations and kinematics between the two systems.

Results

Using dual fluoroscopy as a reference standard, the accuracy of skin-marker-based estimates of the hip joint center and the effect of soft-tissue artifact on kinematic and kinetic measurements were quantified^22,23,24. The superior accuracy of dual fluoroscopy was then used to identify subtle differences in pelvic and hip joint kinematics between patients with FAIS and asymptomatic control participants²⁵. Dual-fluoroscopy-based arthrokinematics were analyzed to quantify hip joint coverage, the relationship between morphology and kinematics, and bone-to-bone distances during dynamic motions^26,27,28,29.

Before developing a protocol to investigate weight-bearing hip joint kinematics, the system was validated in cadaveric specimens with implanted metal beads during supine clinical exams to an accuracy within 0.5 mm and 0.6°³⁰. Once validated, kinematics during clinical exams were measured using dual fluoroscopy in patients with FAIS and asymptomatic control participants. The results demonstrated that patients had altered motion in both internal rotation and adduction³¹.

Using weight-bearing dual fluoroscopy as a reference standard, the error in identifying the location of the hip joint center as well as the errors caused by soft-tissue artifact were then directly analyzed. Functional methods of identifying the hip joint center, i.e., the star-arc motion, were identified to be more accurate than predictive, landmark-based methods with errors of 11.0 and 18.1 mm, respectively³². Dynamic errors in the hip joint center were similar to those from standing; however, an additional 2.2 mm of spurious hip joint center movement was attributed to soft tissue artifact, with errors of more than 5 cm during dynamic movement for the greater trochanter marker²³.

In addition to the errors in identification of the hip joint center, joint angles were underestimated by greater than 20° in internal-external rotation pivots²³. While the underestimation of kinematics is cause for concern in itself, these errors reduced the measured range of motion and calculated kinetic variables during even a low range of motion activities, such as gait²⁴. However, accurate dual fluoroscopy kinematic data can be difficult to incorporate into musculoskeletal models. Specifically, model marker errors were approximately 1 cm when running inverse kinematics with dual fluoroscopy-based landmark locations. While this error is relatively small compared to the 5 cm errors due to soft tissue artifact found for skin marker motion capture data, such error is an order of magnitude larger than that of bone positions measured by dual fluoroscopy.

In addition to the quantification of errors in traditional skin marker motion capture, the accuracy and methodology behind dual fluoroscopy provide the capability to evaluate even subtle differences in kinematics between cohorts, which may otherwise be hidden by the errors of the measurement technique. While differences in hip joint kinematics were not observed between patients with cam FAIS and asymptomatic control participants, differences in pelvic kinematics that would have been difficult to detect in the presence of soft-tissue artifact were identified²⁵. This assessment required direct comparison between cohorts. Moreover, the potential relationship between kinematic variation and bone morphology, such as femoral anteversion, was also investigated²⁷. These findings indicated the need for consideration of both morphology and biomechanics in the diagnosis of hip pathologies and the planning of conservative or surgical treatments.

A major hurdle in the use of biomechanical data in a clinical care setting is the difference in coordinate systems used by biomechanists and clinicians. In a biomechanics lab, the landmarks used to define coordinate systems of the femur and pelvis are driven by the ability to identify and track the landmarks from the skin surface during dynamic motion. In contrast, surgical coordinate systems are defined using bony landmarks identifiable during surgery with a patient supine or prone. The direct tracking of the femur and pelvis in dual fluoroscopy allowed for the evaluation of the influence of various coordinate system definitions on kinematic output²⁹. The differences between coordinate system definitions resulted in kinematic offsets greater than 5°. However, these offsets were relatively consistent during motion and could be accounted for through bony landmark identification.

The combination of subject-specific bone morphology and kinematics — arthrokinematics — provides a joint-level assessment of form and function. For patients with DDH, femoral under-coverage is thought to be the cause of degeneration, and therefore, measurements of coverage are used heavily in diagnosis and surgical planning. Unfortunately, these measurements are often limited to static images, obtained with an individual supine, and only in two dimensions. Dual fluoroscopy-derived arthrokinematics were used to directly measure the variability in femoral coverage during dynamic activities²⁶. Importantly, strong correlations between coverage in standing and coverage during gait when evaluated in entirety were found. Yet, regionalized coverage varied for both anterior and posterior regions of the femoral head even during the stance phase of gait.

Extra-articular impingement is a cause of pain at the hip and surrounding region and describes abnormal contact between the femur and regions of the pelvis outside the acetabulum, including the ischium and anterior inferior iliac spine. The dynamic nature of ischiofemoral impingement was evaluated through the comparison of clinical MRI-based measurements of ischiofemoral space and those during dynamic activities²⁸. Therein, decreased space was observed dynamically in comparison to the standard clinical measures; sex-based differences, which could not be attributed to kinematic differences, were also identified. These methods could also be applied to evaluate joint space dynamically, providing insight into the variability of the position of the femoral head within the acetabulum and the variability across patient cohorts (Figure 8).

Figure 1: Overhead view of the dual fluoroscopy system positioned over the instrumented treadmill for a left hip. The system is positioned to minimize the effect of scatter and maximize the field of view. The image intensifiers are positioned approximately 100-110 cm from the source of the emitter and angled 50° from one another. Please click here to view a larger version of this figure.

Figure 2: View from the contralateral (right) side of a participant during dynamic activities. The participant is positioned between the two image intensifiers (II) such that the field of view of the dual fluoroscopy system is centered over the left hip joint. Level and incline walking, internal and external rotation pivots, and range of motion activities are performed on a treadmill platform. Abbreviation: FHJC = functional hip joint center. Please click here to view a larger version of this figure.

Figure 3: Overhead view of the motion capture system relative to the dual fluoroscopy system. The optical motion capture system includes 10 infrared cameras and a single video-based camera and is positioned on a frame hanging from the ceiling. Please click here to view a larger version of this figure.

Figure 4: Anterior and posterior view of the marker set used for skin marker motion capture. There are five plates with four markers each, which are positioned on the back, thighs, and shanks of the participants; all other markers are applied directly to the skin. Calibration markers are removed for dynamic motion capture. Marker labels prefaced with an R or L indicate markers on the right or left side of the body; marker labels suffixed with S, L, R, I, A, or P indicate marker locations on a marker plate, specifically superior, left, right, inferior, anterior, or posterior, respectively. Abbreviations: *SHO = shoulder; CLAV = center of clavicles; STRN = bottom of sternum; BACK_* = markers of plate placed on the lower back; *ILC = iliac crest; *ASI = anterior superior iliac spine; *PSI = posterior superior iliac spine; GRT_TRO = greater trochanter; *THI_* = markers of the respective plates placed on the thigh; *KNE_M = medial femoral condyle (knee); *KNE_L = lateral femoral condyle (knee); *TIB_* = markers of the respective plates placed on the shank (tibia); *ANK_M = medial malleolus (ankle); *ANK_L = lateral malleolus; *5TH = fifth metatarsophalangeal joint; *TOE = first metatarsophalangeal joint; *HEE = calcaneus (heel). Please click here to view a larger version of this figure.

Figure 5: Landmarks and coordinate systems of the femur and pelvis. Landmarks of bilateral anterior superior iliac spine (ASIS; magenta) and posterior superior iliac spine (PSIS; cyan) and their mid-points are used to define the coordinate system of the pelvis. The center of the femoral head (orange) and bilateral femoral condyles (green), their mid-point, and a cylinder fit of the condyles are used to define the coordinate system of the femur (shown for left femur). The third axis of each bone is determined from the cross-product of the two displayed axes. Please click here to view a larger version of this figure.

Figure 6: Dual fluoroscopy images and associated markerless tracking of a left hip. Images are shown for maximum rotation of the external and internal rotation pivots (center), with the image from the anterior fluoroscope (left) and the posterior fluoroscope (right). Markerless tracking solutions for the pelvis (top) and femur (bottom) for each dual fluoroscopy image. Please click here to view a larger version of this figure.

Figure 7: Dual fluoroscopy measured kinematics. Kinematics for 100 frames surrounding the maximum rotation (vertical dotted line) of external and internal rotation pivots for a representative participant. Please click here to view a larger version of this figure.

Figure 8: Arthrokinematics-based surface distance between a left hemi-pelvis and femur. Arthrokinematics are shown for maximum rotation of the external and internal rotation pivot (center) with respective bone models measured with dual fluoroscopy (outer). Please click here to view a larger version of this figure.

Discussion

Dual fluoroscopy is a powerful tool for the investigation of in vivo kinematics, especially for the hip, which is difficult to accurately measure using traditional optical motion capture. However, fluoroscopy equipment is specialized, wherein a unique system setup may be required when imaging other joints of the human body. For example, several modifications were made to the mounting of the image intensifiers, positioning of the system, and settings of the beam energy in the application of dual fluoroscopy to the study of ankle kinematics^32,33,34,35. In addition to requiring considerable study preparation, dual fluoroscopy requires the acquisition of additional data, including 3D medical imaging and potentially traditional skin marker motion capture to track whole-body kinematics, as well as lengthy post-processing, including CT image segmentation and markerless tracking of the acquired images. Fortunately, fully processed data from dual fluoroscopy can be used in various applications with capabilities reaching far beyond those available with traditional motion capture.

Optical motion capture utilizes the motion of markers on the skin to estimate body segment positions, while radiation-based dual fluoroscopy allows for direct measurement of only the bone positions. While significant effort has been dedicated to quantifying soft tissue dynamics relative to bone motion^{36, 37}, it is inherently difficult to measure the motion patterns of the large mass of soft tissue between the outer layer of skin and the bones. However, for thinner tissues in direct contact with the bones, such as the cartilage and labrum of the hip, the combination of dual fluoroscopy and CT arthrogram imaging provides the ability to dynamically evaluate their spatial relationship. The data collected during supine clinical exams were used to show that the location of clinically observed damage to the acetabular labrum aligned with the position of contact between the femur and labrum during supine impingement exams³⁸. Importantly, this analysis identified that the region of initial and greatest contact between the femur and labrum did not align with the location of the smallest distance between the bones.

Individuals with hip pathoanatomy are at risk of damage to the cartilage and labrum. However, the mechanisms responsible for chondrolabral injuries are not well understood. Conceivably, arthrokinematics data built from CT arthrogram data could be analyzed to study the mechanics of the cartilage and labrum. For example, the observed penetration between surface reconstructions representing soft tissue (e.g., labrum, cartilage) and bone could be analyzed and interpreted to approximate the strain experienced by these tissues. However, even slight errors in the tracking of kinematics or reconstruction of surfaces could result in drastic differences in estimated strains and joint loads. Thus, more advanced modeling methods, such as the FE method, may be required to comprehensively evaluate chondrolabral mechanics in the hip. Data from dual fluoroscopy, traditional skin marker motion capture of whole-body kinematics, and the instrumented treadmill can serve as input for models that estimate muscle forces and joint reaction loads and torques. These kinetic data can then serve as loading conditions to FE models that estimate chondrolabral stresses and strains.

In addition to the specific steps involved in the protocol, the scheduling of different aspects of the study is also relevant to successful data acquisition. First, in studies using arthrogram imaging, which is inherently invasive due to the injection of contrast into the hip capsule, the arthrogram must be performed either several days before or any time after the completion of motion capture experiments to avoid any effect on patient motion patterns. Second, all calibration must be performed prior to, but just before, the arrival of the participant to ensure that the system configuration is not altered between calibration and image acquisition. Third, the participant should be instructed to perform dynamic trials in a random order to eliminate any effect of ordering on the performance of tasks.

Another major consideration for the use of dual fluoroscopy for the measurement of hip kinematics is radiation exposure. It is important to note, however, that 80% of the estimated dose equivalent of radiation in the described protocol is from the CT scan. One solution to reduce exposure is the substitution of magnetic resonance imaging (MRI) for CT imaging. While MRI can be used for surface reconstruction, the tracking of dual fluoroscopy images also relies on the projection of bone densities from the digitally reconstructed radiographs. Although MRI cannot directly measure bone density, specific sequences, such as the dual echo steady state (DESS), provide some differentiation between the denser cortical bone and the less dense cancellous bone. These images can be transformed to have a similar appearance to CT images and could potentially reduce the radiation exposure of participants in dual fluoroscopy studies.

Owing to the large amount of soft tissue surrounding the hip joint, the specific positioning of the dual fluoroscopy system must be optimized to reduce X-ray scatter. The position of the participant relative to the X-ray emitters and the angle between the image intensifiers were found to be important factors. This protocol indicates the positioning of the dual fluoroscopy system used to study hip motion in participants during weight-bearing activities. It is, however, also relevant to note that the participant cohort was limited to individuals with a BMI less than 30 kg/m². A similar BMI limit is recommended when capturing dual fluoroscopy images of joints surrounded by large masses of soft tissue.

The protocol described herein can be applied to various dual fluoroscopy system configurations and joints, including supine and weight-bearing hip kinematics, both treadmill and overground weight-bearing ankle kinematics, and sitting shoulder kinematics^{16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35}. Owing to the minimal global motion of the hip joint during treadmill gait, an instrumented treadmill was used for the assessment of weight-bearing kinematics of the hip joint. Without a treadmill or a moving fluoroscope system, it would only be possible to capture the hip joint during activities performed in a confined field of view. However, the use of a treadmill is not appropriate for all joints. As an example, application of this protocol to the investigation of ankle kinematics during treadmill walking captured only a small portion of gait due to the inherent motion of the treadmill^32,35, while overground gait was able to capture a larger portion of gait, spanning from prior to heel-strike to after toe-off^33,40,41.

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
Amira Software	ThermoFisher Scientific	Version 6.0
Calibration Cube		Custom	36 steel beads (3 mm diameter, spacing 6.35 cm, uncertainty 0.0036 mm)
Calibration Wand	Vicon	Active Wand
CT Scanner	Siemens AG	SOMATOM Definition 128 CT
Distortion Correction Grid		Custom	Acrylic plate with a grid of steel beads spaced 10 mm and 31 beads across the diameter (2 mm diameter)
Dynamic Calibration Plate		Custom	Acrylic plate with 3 steel beads spaced 30 mm (2 mm diameter, uncertainty 0.0013 mm)
Emitter (2)	Varian Interay; remanufactured by Radiological Imaging Services	Housing B-100/Tube A-142
Epinephrine	Hospira	Injection, USP 10 mg/mL
FEBioStudio Software	FEBio.org	Version 1.3	Mesh processing and kinematic visualization
Graphical Processing Unit	Nvidia	Tesla
Hare Traction Splint	DynaMed	Trac-III, Model No. 95201
High-speed Camera (2)	Vision Research, Inc.	Phantom Micro 3
Image Intensifier (2)	Dunlee, Inc.; remanufactured by Radiological Imaging Services	T12964P/S
Iohexol injection	GE Healthcare	Omnipaque 240 mgI/mL	517.7 mg iohexol, 1.21 mg tromethamine, 0.1 mg edetate calcium disodium per mL
ImageJ	National Institutes of Health and Laboratory for Optical and Computational Instrumentation
Lidocaine HCl	Hospira	Injection, USP 10 mg/mL
Laser and Mirror Alignment System		Custom	Three lasers adhered to acrylic plate that attaches to emitter, mirror attaches to face of image intensifier
Markless Tracking Workbench	Henry Ford Hospital, Custom Software	Custom
MATLAB Software	Mathworks, Inc.	Version R2017b
Motion Capture Camera (10)	Vicon	Vantage
Nexus Software	Vicon	Version 2.8	Motion capture
Phantom Camera Control (PCC) Software	Vision Research, Inc.	Version 1.3
Pre-tape Spray Glue	Mueller Sport Care	Tuffner
Retroreflective Spherical Skin Markers			14 mm
Split Belt Fully Instrumented Treadmill	Bertec Corporation	Custom
Visual3D Software	C-Motion Inc.	Version 6.01	Kinematic processing