Definition of anatomical zero positions for assessing shoulder pose with 3D motion capture during bilateral abduction of the arms

Orthopedic interventions at the shoulder typically aim to prevent or resolve pain and restore or establish shoulder function with respect to range of motion (ROM) and strength. In osteoarthritic shoulders, function can be improved significantly by total shoulder arthroplasty, which is an established procedure [1]. Soft tissue procedures may improve function in rotator cuff tears [2]. Clinically, a typical functional test is to assess the patient’s active shoulder abduction aiming for a lateral overhead movement. The function then is documented via scores such as the Constant score [3] or via planar angle measurements with a manual goniometer (Neutral-0-Method [4]) quantifying typically the abduction of the arm relative to the thorax. Patients surgically treated for shoulder instability or rotator cuff tears show limited ROM in this movement and they often compensate with increased shoulder girdle and thorax motion [5]. More than 50 years ago planar measurements showed that about one third of the humerus-thoracic movements originate from the closed chain mechanism of thorax, clavicle and scapula (altogether the shoulder) and only two-thirds from gleno-humeral rotations [6].

In the last decade, various methods have been used to objectively analyze the complex arm and/or shoulder motion in vivo, using MRI [7], biplanar X-ray [8], noninvasive marker-less fluoroscopic imaging [9], 3D electromagnetic tracking [10], inertial sensing [11], ultrasound-based methods [12], and optical marker tracking. The last technique relies on kinematic models which relate positions of skin placed markers to joint kinematics. Up to now, the majority of the models are based on anthropometric reference data and linear regression for determining joint parameters [13, 14]. Other approaches make use of functional joint center and axes estimation for individual joint parameter determination [15] as also recommended by the International Shoulder Group of the International Society of Biomechanics (ISB) [16]. A major drawback in all skin- based methods, including sensor techniques, is that the orientation of the scapula and also rotations about the clavicle’s longitudinal axis are most difficult to access [17]. Some approaches estimate the orientation of the scapula in a limited ROM due to skin motion artefacts [11, 14, 18]. In the case of a marker based system this is called “acromion marker based method” [19, 20], whereas others reduce the description of shoulder kinematics to the two angles elevation/depression and pro-/retraction corresponding to the two rotational degrees of freedom (DoF) associated with the orientation of the clavicle’s long axis [21, 22]. In order to objectively assess shoulder function and, more specifically, shoulder pose and motion in the clinical context of orthopedic interventions by using a 3D marker tracking procedure, the following aspects should be considered: 1) feasibility, i.e., a small set of markers is quick to prepare, and disturb only little; 2) reliability and 3) validity in an adequate range of arm abduction angles and shoulder poses. The first two of these aspects are widely discussed in the literature for several measurement procedures and joint angles. However validity is difficult to prove because there is no gold standard available. In the clinical setting a seemingly intuitive assessment of planar angles in static shoulder poses is performed by universal goniometry. The Neutral-0-Method describes the examination details primarily for the large joints of the lower and upper extremity. This includes the definition of the plane in which the angle is measured and also the anatomical zero position in which the two adjacent segments typically are aligned in parallel (e.g. elbow flexion is zero when the elbow is fully extended and the forearm is aligned in parallel to the humerus). However, a rigorous definition for shoulder elevation/depression and pro-/retraction for shoulder elevation/depression and pro-/retraction is missing in this clinical context. Likewise 3D motion capture is lacking of a definition of a neutral shoulder position. Hence, the comparison of data obtained by universal goniometry and 3D motion capture for proving at least face validity in static cases may be misleading. The typical approach of 3D motion capture to quantify a static reference is not helpful since a “relaxed” position of the shoulder and shoulder girdle may not represent a zero position.

The aim of the presented work was therefore to establish an anatomical zero position of shoulder pose to be included in an objective 3D measurement procedure and to check face validity by universal goniometry in the determination of shoulder pose in a sequence of arm abduction and anteversion positions. This can be regarded as an confirmative test despite the well-known limitations of universal goniometry e.g. in repeatability and accuracy as discussed in [23].

The proposed anatomical zero definition is implemented as an extension to the Heidelberg Upper Extremity (HUX) model [24], to obtain absolute angles for shoulder elevation/depression and pro-/retraction and is represented by projection angles to directly be compared with angles obtained by universal manual goniometry. Projection angles have been used in preference to Euler/Cardan angles as they promise a more intuitive understanding and handling when taking into account anatomic offset angles.

Feasibility, reliability and face validity is tested in a group of young and healthy volunteers performing arm ab-/adduction movements. For monitoring effects of shoulder dysfunction, a subject with gleno-humeral osteoarthritis was assessed in the same manner as a sample case.

In the shoulder position trials, the subjects chose to position their shoulders roughly between 15° and 25° both for shoulder elevation and protraction, with self-selected positions for “little” and “extensively” being typically only about 10° apart.

The degree of agreement between measuring shoulder elevation/depression and pro-/retraction by 3D motion capture and by universal goniometry based on the described custom prepared goniometer is shown by Bland-Altman diagrams [29] in Fig. 3.

With differences between the groups mean values of less than 2° shown by the first three Bland-Altmann diagrams (a-c) the methods are in good agreement for shoulder elevation. The agreement is poorer for shoulder pro-/retraction (d-f) compared to elevation due to a negative shift with systematic differences of up to 5° to the inter-subject data group (f), where universal goniometry always determines larger protraction values than motion capture – a discrepancy which is not present in shoulder elevation measurements (a-c).

In the shoulder movement trials, maximum shoulder protraction of 34.7° (SD 7.4°), maximum shoulder retraction of 8.3° (SD 4.3°), maximum shoulder elevation of 32° (SD 6°), and minimum shoulder elevation of 4.5° (SD 3.3)° was found on average in the reference group of eight healthy subjects. In the nomenclature of the Neutral-0-Method, these values correspond to a shoulder pro-/retraction ROM of (35|0|8); (max. pro-| 0| max. retraction) and to elevation/depression ROM of (32|5|0); (max. elev-| min. elevation| 0), indicating a “shoulder depression deficit” (no depressed position reached) found in all subjects.

A typical repetitive shoulder ab-/adduction pattern is illustrated in Fig. 4 (left), where shoulder elevation is plotted against arm abduction.

The arm is abducted up to about 160°. Shoulder elevation increases with arm abduction and shows a characteristic maximum of 35° elevation at about 130° arm abduction. All healthy subjects abducted their arm by at least 150°. Inter-subject means for elevation are shown in 30° intervals in the range between 0 and 150° abduction in Fig. 4 (right). Overall this reference group shows little shoulder elevation in the first interval 0 - 30°, continuous increase in elevation in the interval 30 - 120°, and again little change in the interval 120 - 150° arm abduction. Characteristic maxima in shoulder elevation as illustrated in Fig. 5b were found in three out of eight subjects in whom arm abductions larger than 150° were observed.

The advised bilateral arm abduction movement was performed in a very similar and characteristic way by all volunteers, as reflected in the large CMC values (Table 1).

In the time plots in Fig. 5, the gray bands illustrate in detail how the movement is performed. The arms are harmonically abducted with a maximum of typically 150° abduction at 50 % movement cycle (MC) (compare Fig. 5a), i.e., the arms are raised and lowered at the same speed. Maximum shoulder elevation is reached earlier, at about 30 % MC (compare Fig. 5b), and kept rather constant until 70 % MC with a slight dip at 50 % MC, i.e., at the position of maximum arm abduction. Furthermore, with increasing arm abduction the arms are continuously anteverted from typically 10° retroversion with hanging arms to about 30° anteversion at maximum abduction, i.e., at 50 % MC (compare Fig. 5c). Conversely, the shoulder moves into retraction, however, with a smaller ROM (pro-/retraction: (0|0|20); compare Fig. 5d). Throughout abduction until 90°, the arm is continuously externally rotated from 5° internal to 45° external rotation (compare Fig. 5e). For abduction angles larger than 90° the arm still rotates externally but the joint position is rated differently according to the concept of conjunct rotation [33].

In contrast to the healthy volunteers, the 72 year old patient with gleno-humeral osteoarthritis could only raise his arm up to 80° arm abduction (compared to 150° in the reference group; see Fig. 5a) but with increased shoulder elevation (35° compared to 30°; see Fig. 5b). Furthermore, the ROM in arm ante-/retroversion (40|0|20) is increased compared to (30|0|10); (see Fig. 5c) and shoulder pro-/retraction ROM (0|10|45) increased compared to (0|0|20); (see Fig. 5d). Similar to the reference subjects, the subject with gleno-humeral osteoarthritis continuously externally rotates his arm from 5° internal to 45° external rotation (see Fig. 5e).

Surgical interventions at the shoulder are performed as routine procedures to treat shoulder dysfunction of traumatic or degenerative causes. In recent years a number of in vivo methods have been developed to monitor motion of the shoulder complex. However, the direct comparison to the outcome assessed by universal goniometry is difficult. In the case of the scapula, there are well accepted standards in biomechanics to report its motion by Euler-Cardan angles but there is no standard in describing the motion of the shoulder as a whole [22]. In addition to the quasi 2D description of the shoulder in [28, 34] which describe the shoulder as a linear segment \( \overrightarrow{l} shoulder \) mainly following the motion of the clavicle this work is focused to add the missing anatomical zero position definitions of the shoulder with respect to elevation and protraction. To define the required reference coordinate system of the thorax, markers placed on the pelvis are used to reduce artifacts based on individual differences mainly in the shape of the anterior thorax. The implementation is done as an extension of the HUX model [24] but the anatomical zero positions can also be used for any other measurement procedure. Despite its clinical importance, we know of only one publication by Nadeau and co-authors [35] taking an anatomical zero position for shoulder elevation into account but not for pro-/retraction. The specificity of this anatomical zero position definition is that it closely resembles the intuitive measurement process of planar angles at the shoulder complex with universal goniometry to tie in with the experiences of orthopedists.

The results of the shoulder position trials with respect to elevation (compare Fig. 3) show a comparatively good agreement between marker tracking and manual goniometer readings, with a mean difference less than 2° in the intra-tester and the inter-subject data group carried out from the same tester. In the inter-tester data set the mean difference is even less than 0.5°. This indicates that systematic differences are mainly influenced by the tester and not by the measured subject and in consequence should be regarded as a limitation of the universal goniometry. However this effect is not significant for the data groups presented. We are aware that manual goniometer readings cannot serve as a golden standard due to possible inaccuracies of 5° and more [23]. This matches with the limits of agreement in the Bland-Altmann diagrams. However, the cross-checking will at least confirm that data obtained by the optical method determine the same quantity compared to using a manual goniometer and hence prove at least face validity. The results of our model with respect to shoulder elevation are also confirmed in a dynamic situation by measurements with MRI [7] resembling the “shoulder rhythm” of clavicula-thoracic motion as illustrated in Fig. 5. However, in the MRI study shoulder elevation is found to be smaller by 4° throughout abduction which may be explained by the different definition of the anatomic zero position in our approach. In addition, measurements based on electromagnetic tracking of trans-cortical pins [36] showed the same pattern but with less ROM which may be explained by the fact that sterno-clavicular motion in a standing position instead of thoracic-clavicular motion was measured. To our knowledge, the only other study using optical motion capture with a similar approach to monitor shoulder elevation during arm abduction is the work by Garofalo et al. [22]. Their results showed a similar ROM in elevation but shoulder abduction was only monitored up to 120° and no explicit anatomical zero position was taken into account at all.

The fact that for shoulder protraction we do not obtain equally satisfying results as for shoulder elevation, may be a consequence of the limitation that clavicle motion is modeled by only two instead of three DoF or simply reflect the situation that the clinical examiner can assess frontal plane (facing) angles more easily compared to assessing the persons shoulder position from above. However, “Pitching” of the shoulder, i.e. rotation about the long axis of the clavicle cannot be monitored by this model approach (and neither by universal goniometry), which does not take any information of the scapula into account. Hence shoulder motion can involve subject specific relevant shoulder pitching and may therefore influence the data on shoulder protraction. Similarly, the goniometer measurements are more difficult to be performed for shoulder pro-/retraction than for elevation because the alignment of the goniometer to the medio-lateral reference must rely on a much shorter lever arm.

The shoulder function primarily monitored in this work with lateral overhead movements of both arms, was proven to be performed in a very specific and repetitive pattern quantified by large CMC values (compare Table 1) and including out-of-plane motion for shoulder and arm (compare Fig. 5). The definition of the anatomical zero position allows to directly compare individual resting positions of the shoulder as well as maximum shoulder depression and elevation angles. None of the subjects studied in this work was able to actively depress (i.e. lower) the shoulder beneath anatomical zero position. All of the subjects have an individual but always elevated shoulder in their most relaxed position. The measured value of 5° agrees with the value of 5.9 ± 1° for the bony clavicle position given in [36].

To demonstrate the feasibility of the presented method in a clinical context and to illustrate possible shoulder dysfunction effects, a 72-year old man with gleno-humeral osteoarthritis was also monitored in the same manner as the young group of volunteers. The test took about 30 min with this very compliant subject reporting no negative effects. He showed overall less ROM in arm abduction which is a typical finding in these subjects. In contrast, when taking into account this small ROM, in relation he showed noticeably larger shoulder elevation motion. Further, he demonstrated an increased out-of-plane movement i.e. the arm performed also ante-/retroversion of 40° and the shoulder performed also a pro-/retraction of 35° during this lateral arm overhead reach. These findings may be explained as compensation mechanisms in order to reduce motion of the degenerated gleno-humeral joint. However, as the reference group is much younger this reasoning is preliminary. Also age dependent reduced proprioception as found in [37] may be responsible for increased out-of-plane motion. Nevertheless, the clear difference in pattern found in this subject in relation to the reference group gives rise to the hope that further studies using this marker-based model for monitoring degenerative or traumatic shoulder dysfunction will give more insight into biomechanical and functional changes induced by surgical interventions at the shoulder.

The missing link to determine absolute angles is filled by the proposed feature of an anatomical zero position definition. This allows for determining absolute joint angles independent from a static reference position. The feature was implemented successfully as an extension to the HUX model [24] to determine absolute shoulder elevation/depression and pro-/retraction angles. The definition of zero positions on the basis of correspondence with universal goniometry for static positions facilitates to create confidence in the method by orthopedists because the angles definitions are known and hence self-explanatory. Hence, the method is suitable for objectively and quantitatively assessing shoulder dysfunction in frontal plane movements as illustrated for eight volunteers and in the case with gleno-humeral osteoarthritis. In contrast, shoulder pro-/retraction shows systematic differences between the different data groups (inter-tester, intra-tester, inter-subject) for the mean differences between marker-based angle determination and manual goniometer measurements. Accuracy in sagittal plane shoulder motion may be lower for specific subjects.

The strategy to define anatomical zero positions in correspondence with universal goniometry has been found very helpful despite its obvious limitations to static measurements and its low accuracy.