Biplane videoradiography can quantify shoulder kinematics with a high degree of accuracy. The protocol described herein was specifically designed to track the scapula, humerus, and the ribs during planar humeral elevation, and outlines the procedures for data collection, processing, and analysis. Unique considerations for data collection are also described.
The shoulder is one of the human body’s most complex joint systems, with motion occurring through the coordinated actions of four individual joints, multiple ligaments, and approximately 20 muscles. Unfortunately, shoulder pathologies (e.g., rotator cuff tears, joint dislocations, arthritis) are common, resulting in substantial pain, disability, and decreased quality of life. The specific etiology for many of these pathologic conditions is not fully understood, but it is generally accepted that shoulder pathology is often associated with altered joint motion. Unfortunately, measuring shoulder motion with the necessary level of accuracy to investigate motion-based hypotheses is not trivial. However, radiographic-based motion measurement techniques have provided the advancement necessary to investigate motion-based hypotheses and provide a mechanistic understanding of shoulder function. Thus, the purpose of this article is to describe the approaches for measuring shoulder motion using a custom biplanar videoradiography system. The specific objectives of this article are to describe the protocols to acquire biplanar videoradiographic images of the shoulder complex, acquire CT scans, develop 3D bone models, locate anatomical landmarks, track the position and orientation of the humerus, scapula, and torso from the biplanar radiographic images, and calculate the kinematic outcome measures. In addition, the article will describe special considerations unique to the shoulder when measuring joint kinematics using this approach.
The shoulder is one of the human body's most complex joint systems, with motion occurring through the coordinated actions of four individual joints, multiple ligaments, and approximately 20 muscles. The shoulder also has the greatest range of motion of the body's major joints and is often described as a compromise between mobility and stability. Unfortunately, shoulder pathologies are common, resulting in substantial pain, disability, and decreased quality of life. For example, rotator cuff tears affect about 40% of the population over age 601,2,3, with approximately 250,000 rotator cuff repairs performed annually4, and an estimated economic burden of $3-5 billion per year in the United States5. Additionally, shoulder dislocations are common and are often associated with chronic dysfunction6. Lastly, glenohumeral joint osteoarthritis (OA) is another significant clinical problem involving the shoulder, with population studies indicating that roughly 15%-20% of adults over the age of 65 have radiographic evidence of glenohumeral OA7,8. These conditions are painful, impair activity levels, and decrease quality of life.
Although the pathogeneses of these conditions are not fully understood, it is generally accepted that altered shoulder motion is associated with many shoulder pathologies9,10,11. Specifically, abnormal joint motion may contribute to the pathology9,12, or that the pathology may lead to abnormal joint motion13,14. Relationships between joint motion and pathology are likely complex, and subtle alterations in joint motion may be important in the shoulder. For example, although angular motion is the predominant motion occurring at the glenohumeral joint, joint translations also occur during shoulder motion. Under normal conditions these translations likely do not exceed several millimeters15,16,17,18,19, and therefore may be below the level of in-vivo accuracy for some measurement techniques. While it may be tempting to assume that small deviations in joint motion may have little clinical impact, it is important to also recognize that the cumulative effect of subtle deviations over years of shoulder activity may exceed the individual's threshold for tissue healing and repair. Furthermore, in-vivo forces at the glenohumeral joint are not inconsequential. Using custom instrumented glenohumeral joint implants, previous studies have shown that raising a 2 kg weight to head height with an outstretched arm can result in glenohumeral joint forces that can range from 70% to 238% of body weight20,21,22. Consequently, the combination of subtle changes in joint motion and high forces concentrated over the glenoid's small load-bearing surface area may contribute to the development of degenerative shoulder pathologies.
Historically, the measurement of shoulder motion has been accomplished through a variety of experimental approaches. These approaches have included the use of complex cadaveric testing systems designed to simulate shoulder motion23,24,25,26,27, video-based motion capture systems with surface markers28,29,31, surface-mounted electromagnetic sensors32,33,34,35, bone pins with reflective markers or other sensors attached36,37,38, static two-dimensional medical imaging (i.e., fluoroscopy39,40,41 and radiographs17,42,43,44,45), static three-dimensional (3D) medical imaging using MRI46,47, computed tomography48, and dynamic, 3D single plane fluoroscopic imaging49,50,51. More recently, wearable sensors (e.g., inertial measurement units) have gained popularity for measuring shoulder motion outside the laboratory setting and in free-living conditions52,53,54,55,56,57.
In recent years, there has been a proliferation of biplane radiographic or fluoroscopic systems designed to accurately measure dynamic, 3D in-vivo motions of the shoulder58,59,60,61,62. The purpose of this article is to describe the authors' approach for measuring shoulder motion using a custom biplanar videoradiography system. The specific objectives of this article are to describe the protocols to acquire biplanar videoradiographic images of the shoulder complex, acquire CT scans, develop 3D bone models, locate anatomical landmarks, track the position and orientation of the humerus, scapula, and torso from the biplanar radiographic images, and calculate kinematic outcome measures.
Prior to data collection, the participant provided written informed consent. The investigation was approved by Henry Ford Health System's Institutional Review Board.
Protocols for acquiring, processing, and analyzing biplane radiographic motion data are highly dependent upon the imaging systems, data processing software, and outcome measures of interest. The following protocol was specifically designed to track the scapula, humerus, and the third and the fourth ribs during scapular-plane or coronal-plane abduction and to quantify glenohumeral, scapulothoracic, and humerothoracic kinematics.
1. CT imaging protocol
2. Biplane X-ray motion capture protocol
NOTE: The custom biplanar x-ray system used in this protocol is described in the Table of Materials. Data collection procedures will likely vary with different system components. The x-ray systems are arbitrarily termed "green" and "red" to distinguish procedures and resulting image sequences and are positioned with an approximately 50° inter-beam angle and a source-to-image distance (SID) of approximately 183 cm (Figure 2). A minimum of two research personnel are required for the data collection; one to operate the x-ray system and computer, and the other to instruct the research participant.
3. Data processing protocol
NOTE: Procedures for preparing the bony geometry, image pre-processing (i.e., distortion and non-uniformity correction and image calibration), and markerless tracking are highly variable and depend on the software used. The procedures described herein are specific to the proprietary software. However, the major data processing steps are likely translatable to any x-ray motion capture software package.
4. Data analysis protocol
NOTE: The proprietary markerless tracking software used in this protocol results in the raw and filtered trajectories of the anatomical landmarks that will be used to construct anatomical coordinate systems. These coordinates are expressed relative to the laboratory coordinate system defined by the calibration object during the calibration procedure. The following protocol describes, in general terms, the procedures for calculating kinematic outcome measures from these landmark trajectories such that they can be computed in any programming language (e.g., MATLAB). A second proprietary software is used to calculate kinematics and proximity statistics.
A 52-year-old asymptomatic female (BMI = 23.6 kg/m2) was recruited as part of a previous investigation and underwent motion testing (coronal plane abduction) on her dominant (right) shoulder65. Prior to data collection, the participant provided written informed consent. The investigation was approved by Henry Ford Health System's Institutional Review Board. Data collection was performed using the protocol previously described (Figure 3).
The participant's glenohumeral, scapulothoracic, and humerothoracic kinematics are presented in Figure 7, Figure 8, and Figure 9, respectively. Visual inspection of glenohumeral and scapulothoracic kinematics suggests the participant's shoulder motion was consistent with what is generally expected during coronal plane abduction66. Specifically, glenohumeral motion consisted of elevation and slight external rotation, and was generally in a plane posterior to the scapula (Figure 7), while scapulothoracic motion consisted of upward rotation, posterior tilt, and slight internal/external rotation (Figure 8).
During the motion trial, the minimum subacromial distance (i.e., narrowest width of the subacromial outlet for a given frame) ranged from 1.8 mm at 74.0° humerothoracic elevation (frame 45) to 8.3 mm at 134.0° humerothoracic elevation (frame 89) (Figure 10A, Figure 11A). The average subacromial distance (i.e., average width of the subacromial outlet within the specified 200 mm2 measurement area) tended to follow a similar trajectory as the minimum distance metric. For example, the average subacromial distance ranged from 4.2 mm at 75.4° humerothoracic elevation (frame 46) to 9.2 mm at 134.0° humerothoracic elevation (frame 89). Finally, the minimum subacromial distance tended to follow a complementary trajectory to the surface area metric (Figure 10B) such that the minimum distance tended to be smaller when the surface area is larger. Plotting the location of the minimum distance on the humeral head suggests the location closest to the acromion shifts laterally across the rotator cuff footprint as the humerothoracic elevation angle increases (Figure 11A). Across the motion trial, the contact path length measured 40.5 mm on the humeral head and 28.8 mm on the acromion.
During the motion trial, the minimum glenohumeral distance (i.e., narrowest width of the glenohumeral joint space) ranged from 1.0 mm at 137.9° humerothoracic elevation (frame 92) to 2.1 mm at 34.2° humerothoracic elevation (frame 21) (Figure 12A, Figure 11B). As with the subacromial distances, the average glenohumeral distance tended to follow a similar trajectory as the minimum distance metric, and these distances followed a complementary trajectory with the surface area metric (Figure 12B). For example, the average glenohumeral distance ranged from 1.4 mm at 137.9° humerothoracic elevation (frame 92) to 2.6 mm at 23.5° humerothoracic elevation (frame 12). Plotting the location of the glenohumeral contact center relative to the glenoid edge contours suggest that the participant's arthrokinematics included moderate surface interactions. Specifically, the humerus stayed relatively centered in the glenoid in the anterior/posterior direction but shifted superiorly and then inferiorly during the motion trial (Figure 11B). Across the motion trial, the contact path length measured 30.0 mm on the glenoid and 45.4 mm on the humeral head.
Figure 1: The CT field of view. (A) coronal, (B) sagittal, and (C) transverse planes. During acquisition, the CT technologist ensures the field of view includes the clavicle (superiorly), the distal humeral epicondyles (inferiorly), the entire glenohumeral joint (laterally), and the costovertebral and sternocostal joints (medially). Please click here to view a larger version of this figure.
Figure 2: Schematic of the biplane videoradiographic system. The x-ray systems are positioned with a 50° inter-beam angle and a source-to-image distance (SID) of 183 cm. Participants are positioned in the biplane volume such that their glenohumeral joint is located approximately at the intersection of the x-ray beams. Systems are termed "green" and "red" to distinguish the control panels and the filenames of the images. Please click here to view a larger version of this figure.
Figure 3: Biplane radiographic images from a representative subject during coronal plane abduction. Although the jaw appears in the images of the green system, care should be taken to avoid including the head in the field of view to minimize dose to this area. Please click here to view a larger version of this figure.
Figure 4: Definition of anatomical coordinate systems. (A) Humeral coordinate system defined by digitizing the geometric center of the humeral head, medial epicondyle, and lateral epicondyle. (B) Scapular coordinate system defined by digitizing the medial spine, inferior angle, and posterior aspect of the acromioclavicular joint. (C) Rib coordinate system defined by digitizing the posterior aspect of the costovertebral facet, the lateral-most aspect of the rib, and the lateral sternum at the level of the rib. Please click here to view a larger version of this figure.
Figure 5: Definition of regions of interest (ROI) for proximity statistics. (A) humeral head ROI, which is used to calculate acromiohumeral distance and glenohumeral joint contact patterns, (B) acromial and glenoid ROIs, which are used to calculate acromiohumeral distance and glenohumeral joint contact patterns, respectively. Please click here to view a larger version of this figure.
Figure 6: Screenshots of the proprietary markerless tracking software. The screenshot illustrates the optimized solutions of the humerus and scapula from a representative subject during coronal plane abduction. Please click here to view a larger version of this figure.
Figure 7: Glenohumeral kinematics from a representative subject during a single trial of coronal plane abduction. Note: Anterior position has been transformed to be a positive value. Abbreviations: med. = medial; lat. = lateral; sup. = superior; inf. = inferior; ant. = anterior; post. = posterior. Please click here to view a larger version of this figure.
Figure 8: Scapulothoracic kinematics from a representative subject during a single trial of coronal plane abduction. Note: Anterior position has been transformed to be a positive value. Abbreviations: IR = internal rotation; ER = external rotation; UR = upward rotation; DR = downward rotation; AT = anterior tilt; PT = posterior tilt; med. = medial; lat. = lateral; sup. = superior; inf. = inferior; ant. = anterior; post. = posterior. Please click here to view a larger version of this figure.
Figure 9: Humerothoracic kinematics from a representative subject during a single trial of coronal plane abduction. Note: Anterior position has been transformed to be a positive value. Abbreviations: med. = medial; lat. = lateral; sup. = superior; inf. = inferior; ant. = anterior; post. = posterior. Please click here to view a larger version of this figure.
Figure 10: Assessment of the subacromial space during a trial of coronal plane abduction in a representative subject. (A) Measures of acromiohumeral distance are displayed across frames along with the corresponding humerothoracic elevation angles. The minimum distance is calculated as the smallest distance between the centroids of the nearest-neighbor triangle between the humeral head and acromial ROIs. The average distance represents the area-weighted mean of the minimum distance, calculated over the triangles in the humeral head ROI that have the smallest gap to their nearest neighbor on the acromial ROI. (B) The surface area of the humeral head ROI that is within 10 mm of the acromial ROI is displayed across frames along with the corresponding humerothoracic elevation angles. Abbreviation: HT = humerothoracic. Please click here to view a larger version of this figure.
Figure 11: Proximity mapping. (A) subacromial space, (B) glenohumeral joint space. The subacromial proximity is mapped on the humeral head ROI using the minimum distance metric for the frame of data in which the minimum distance was smallest (i.e., frame #45). The contact path (black) represents the minimum distance trajectory between frames #1-45. The glenohumeral joint proximity is mapped using the weighted-average contact center for the frame of data in which the joint space was smallest (i.e., frame #92). The contact path (black) represents the centroid trajectory between frames #1-92. Please click here to view a larger version of this figure.
Figure 12: Assessment of the glenohumeral joint space during a trial of coronal plane abduction in a representative subject. (A) Measures of glenohumeral joint space are displayed across frames along with the corresponding humerothoracic elevation angles. The minimum distance is calculated as the smallest distance between the centroids of the nearest-neighbor triangle between the glenoid and humeral head ROIs. The average distance represents the area-weighted mean of the minimum distance, calculated over the triangles in the glenoid ROI that have the smallest gap to their nearest neighbor on the humeral head ROI. (B) The surface area of the glenoid ROI that is within 10 mm of the humeral head ROI is displayed across frames along with the corresponding humerothoracic elevation angles. Abbreviation: HT = humerothoracic. Please click here to view a larger version of this figure.
The technique described here overcomes several disadvantages associated with conventional techniques for assessing shoulder motion (i.e., cadaveric simulations, 2D imaging, static 3D imaging, video-based motion capture systems, wearable sensors, etc.) by providing accurate measures of 3D joint motion during dynamic activities. The accuracy of the protocol described herein was established for the glenohumeral joint against the gold standard of radiostereometric analysis (RSA) to be ±0.5° and ±0.4 mm67,68. Similar protocols have been developed for other joints such as the knee69, spine70, and foot/ankle71. Importantly, without a system that is sufficiently accurate, the sample size necessary to detect statistically significant and clinically potential meaningful differences in joint motion could be prohibitive. Furthermore, this level of accuracy affords the ability to describe potentially important outcome measures such as joint positions and/or translations62,72, arthrokinematics72,73,74,75, subacromial distances61,72,75, and instantaneous axes of motion76. Ultimately, accurately measuring in-vivo joint motion is essential for providing a mechanistic understanding of shoulder function under normal and pathologic conditions, and for assessing the effects of non-surgical and surgical clinical interventions.
The accuracy afforded by quantifying shoulder kinematics using biplane videoradiography comes with many challenges and limitations. The primary limitation associated with this technique is the radiation exposure to the participant as a result of the CT scan and biplane x-ray imaging. Consequently, the number of motion trials that can be acquired or follow-up sessions over time is limited. The effective dose corresponding with the protocol described here is approximately 10.5 mSv, with the majority (approximately 10 mSv) coming from the CT scan, which includes imaging of the distal humerus so that the epicondyles can be used to construct the humeral anatomical coordinate system64. For context, this dose corresponds to approximately 3 years of exposure to natural background sources of radiation. Recent recommendations of the National Council on Radiation Protection and Measurements suggest this dose can be classified as "minor" assuming a moderate expected benefit to the individual or society77. Consequently, it is imperative that motion analysis using biplane videoradiography be used in a well-designed study based on a solid scientific premise that has the potential to have an significant impact on public health.
Reducing the dose associated with biplane videoradiography is crucial to facilitate the broader use of this technology in research and clinical settings. Fortunately, recent advances in CT and MR imaging may substantially reduce the dose to the participant. For example, humeral and scapular bone models derived using MRI78,79 or lower dose CT80 have been shown to have acceptable accuracy for many research applications. Furthermore, redefining the humeral coordinate system in a manner that does not require the humeral epicondyles81 will decrease the dose by reducing the CT imaging volume. Careful practice of motion trials before acquiring any images is also crucial to ensure that each collected trial has value and does not unnecessarily add to the participant's total dose. Ultimately, carefully considering these factors, and many others, is critical when responsibly using biplane videoradiography to quantify 3D kinematics in human research participants.
The participant's body habitus and differences in tissue density (and therefore image brightness) between the central torso and the lateral aspect of the shoulder presents additional challenges when quantifying shoulder motion using biplane videoradiography. In particular, clear visualization of the scapula and ribs is often challenging using the radiographic technique described in this protocol (i.e., ~70 kVp, 320 mA, 2 ms pulsed exposure) in individuals with high BMI (>30 kg/m2) and women with large or dense breast tissue. Kinematic tracking accuracy likely deteriorates without clear visualization of bone edges. Consequently, careful selection of participants by restricting BMI can ameliorate many of these challenging imaging considerations. However, "washout" of the lateral acromion at lower angles of humeral elevation is common even in participants of healthy body habitus (Figure 2A, green system at Frame 1). This is because there is little tissue (and thus density) around the acromion when the humerus is at lower angles of elevation, and visibility of this region is conceded in order to visualize the scapula and ribs. However, once the humerus elevates and the bulk of the shoulder in projected onto itself (thus increasing radiographic density), the acromion becomes well-visualized. Therefore, the optimal radiographic technique for a motion trial does not necessarily guarantee visualization of all bones at all times, but allows for the clear visualization of enough bony anatomy to conduct markerless tracking.
Another challenge when using biplane videoradiography is the relatively small 3D imaging volume, which is predominantly defined by the image receptor size, the orientation of the two imaging systems, and the SID. Although limiting the 3D imaging volume helps control the radiation dose (i.e., through collimation), a small imaging volume may restrict the range over which joint motion can be acquired and/or the types of tasks being assessed. For example, tasks that require trunk motion (e.g., throwing) may be incompatible with biplane videoradiography motion analysis because the participant will likely move outside of the 3D imaging volume while performing the task. Patient movement outside the imaging volume is common even in simpler tasks such as raising the arm, especially in individuals whose humeral elevation range of motion is significantly impaired (e.g., due to massive rotator cuff tears, adhesive capsulitis, OA), because these individuals often compensate by leaning to the contralateral side. Consequently, careful positioning of the participant within the imaging volume and verbal cues to avoid leaning are crucial steps in the data collection process (section 2.4).
The small 3D imaging volume also limits the visualization of other segments that may be of interest. For example, tracking the torso is necessary to quantify scapulothoracic and humerothoracic kinematics. The protocol described in this article addresses this challenge by tracking the third and fourth ribs. However, other investigators have tracked the torso using an external surface-based tracking system synced with the radiographic system49,50,62. Each of these approaches has unique limitations. For example, tracking the ribs requires good visualization of the central torso, which is challenging in individuals with larger body habitus without washing out the lateral shoulder, as previously described. Furthermore, tracking the ribs may be challenging with a smaller image intensifier (i.e., less than 40 cm). In contrast, tracking torso motion using surface sensors introduces skin motion artifact. Regardless of the approach used, the limited 3D imaging volume remains a challenge when quantifying shoulder kinematics using biplane videoradiography.
In summary, biplane videoradiography allows for highly accurate quantification of shoulder kinematics. Variations in the protocol described herein has been used for numerous studies within the lab58,59,72,73,82, with each protocol variation carefully constructed based on the specific research aims in order to minimize dose, maximize image quality, and maximize segment visibility. Ultimately, accurately measuring in-vivo joint motion is important for providing a mechanistic understanding of shoulder function under normal and pathologic conditions, and for assessing the effects of non-surgical and surgical clinical interventions.
The authors have nothing to disclose.
Research reported in this publication was supported by the National Institute of Arthritis and Musculoskeletal and Skin Diseases under award number R01AR051912. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health (NIH).
Calibration cube | Built in-house | N/A | 10 cm Lucite box with a tantalum bead in each corner and four additional beads midway along the box’s vertical edges (12 beads total). The positions of each bead are precisely known relative to a corner of the box that serves as the origin of the laboratory coordinate system. |
Distortion correction grid | Built in-house | N/A | Lucite sheet that covers the entire face of the 16 inch image intensifier and contains an orthogonal array of tantalum beads spaced at 1 cm. |
ImageJ | National Institutes of Health | N/A | Image processing software used to prepare TIFF stack of bone volumes. |
Markerless Tracking Workbench | Custom, in house software | N/A | A workbench of custom software used to digitize anatomical landmarks on 3D bone models, constructs anatomical coordinate systems, uses intensity-based image registration to perform markerless tracking, and calculates and visualize kinematic outcomes measures. |
MATLAB | Mathworks, Inc | N/A | Computer programming software. For used to perform data processing and analysis. |
Mimics (version 20) | Materialise, Inc | N/A | Image processing software used to segment humerus, scapula, and ribs from CT scan. |
Open Inventor | Thermo Fisher Scientific | N/A | 3D graphics program used to visualize bones |
Phantom Camera Control (PCC) software (version 3.4) | N/A | Software for specifying camera parameters, and acquiring and saving radiographic images | |
Pulse generator (Model 9514) | Quantum Composers, Inc. | N/A | Syncs the x-ray and camera systems and specifies the exposure time |
Two 100 kW pulsed x-ray generators (Model CPX 3100CV) | EMD Technologies | N/A | Generates the x-rays used to produce radiographic images |
Two 40 cm image intensifiers (Model P9447H110) | North American Imaging | N/A | Converts x-rays into photons to produce visible image |
Two Phantom VEO 340 cameras | Vision Research | N/A | High speed cameras record the visible image created by the x-ray system |