Can optimal marker weightings improve thoracohumeral kinematics accuracy?
Introduction
Human movement kinematics is commonly assessed using stereophotogrammetry and skin-markers placed above bony landmarks. When skeleton kinematics is the subject of interest, the primary source of error in joint angles comes from the displacement of the skin-markers with respect to their underlying bones. This occurrence, termed soft tissue artifact (STA), is the consequence of muscle contraction, skin elasticity, impacts, etc. (Peters et al., 2010). Efforts have been made to reduce errors due to STA, which are usually assessed using invasive methods: e.g. intracortical pins (Andersen et al., 2010, Reinschmidt et al., 1997b) or fluoroscopy (Stagni et al., 2005), for a review see Leardini et al. (2005). Such method have been used to investigate the lower-limb STA (Akbarshahi et al., 2010, Cappozzo et al., 1996, Reinschmidt et al., 1997b, Tsai et al., 2009), but few investigations focused on upper-limb (Hamming et al., 2012b, Matsui et al., 2006). Since STA are different between segments, e.g. thigh vs shank (Benoit et al., 2006, Camomilla et al., 2009, Reinschmidt et al., 1997a, Stagni et al., 2005), further investigations are needed to identify suitable methods for reducing STA propagation to the upper-limb kinematics.
While marker sets exist for upper-limb use in conjunction with optoelectronics systems (Butler et al., 2010, Jackson et al., 2012), electromagnetic sensors are preferred in clinical studies for reasons of space and cost (Finley and Lee, 2003, Hamming et al., 2012a, Meskers et al., 1998, Stokdijk et al., 2003). Efforts have already been made to better track the scapula which slides under the skin (Lempereur et al., 2014). Regarding the humerus, errors up to 30° were reported in axial rotation due to STA (Hamming et al., 2012a) when using cuff mounted electromagnetic sensors. This error cannot be compensated for since one sensor on each segment does not provide any redundancy.
On the lower-limb, markers undergo different STA according to their location. On each marker, STA is composed of a rigid (or in-unison) component and a deformation (or own) component (Andersen et al., 2012, Grimpampi et al., 2014, Leardini et al., 2005). Some authors have proposed mathematical models representing STA (Camomilla et al., 2013, Dumas et al., 2014) and others used least squares algorithms to reduce STA (Cheze et al., 1995), especially the deformation component, in so-called local optimization algorithms. To reduce the rigid component and avoid joint dislocation problems, chain models with set degrees of freedom in combination with nonlinear least squares algorithms (Begon et al., 2009, Laitenberger et al., 2014, Lu and O’Connor, 1999) or extended Kalman filters (Fohanno et al., 2014, Halvorsen et al., 2004) (termed as global optimization) have emerged.
Since STA is not uniform within and between the body segments, these algorithms were improved by introducing weightings, in both global (Alonso et al., 2007, Ausejo et al., 2011, Begon et al., 2008) and local optimization (Andriacchi et al., 1998). Each marker weight can manually be adjusted in the musculoskeletal OpenSim software (Delp et al., 2007). Lu and O’Connor (1999) introduced a weighting matrix to reflect the error distribution among the markers. For simplicity, they chose equal weightings for all the markers at the same segments but smaller weightings to the thigh than the pelvis and shank. Indeed skin movement artefact is bigger on the thigh (Cappozzo et al., 1996). In their application to the upper-limb, Roux et al. (2002) refined the weightings with segmental residual errors given by the algorithm of Söderkvist and Wedin (1994). Unfortunately, to the best of our knowledge, weighting values, methods for their identification, and assessment of the gain in accuracy have never been provided for lower-limb or upper-limb.
The objective of this study was to assess the effect of skin marker weightings in a local optimization algorithm on arm orientation accuracy. First, optimal weightings for each skin marker were obtained based on a gold standard humeral orientation. Then optimal weightings obtained for each movement and each subject were applied to other movements and other subjects to determine if weightings are subject- and/or movement-specific.
Section snippets
Experiment
Four male subjects (age: 32, 41, 44 and 27 years, height: 1.72, 1.82, 1.77 and 1.65 m, mass: 80, 115, 82 and 57 kg, and BMI 27, 35, 26 and 21 kg m−2, for S1 to S4 respectively) volunteered after giving their informed consent. The protocol was approved by the ethic committees of both University of Montreal and Karolinska Institutet, where the experiment took place. As fully described in Dal Maso et al. (2014), an orthopaedic surgeon inserted a pin into the humerus under sterile surgery conditions.
Best scenario
Using seven skin-markers and no weighting, the orientation error ranged from 1.9° (S4 arm abduction) to 17.9° (S3, combing). The participant with the smallest BMI, S4, showed an average error of 5.4±2.3° (with peak value of 11.1±4.5, see Table 1). The average error in S1 was twice larger than in S4. The tasks subjected to the largest deviations were composed of large arm internal–external rotation (i.e., all movements of internal–external rotations, comb hair and reach back). The smallest
Discussion
The present study assessed the effect of skin marker weightings in local optimization on the accuracy of arm orientation. In line with the early works of Lundberg (1996), using qualitative validation, we explored a method to reduce STA. Based on a gold standard measure (i.e., intracortical pin with five markers reconstructed using 18 cameras), our main findings are that (i) the expected improvement varies between 1° and 5° in the best but unrealistic scenario; (ii) weightings are movement- and,
Conclusion
By comparison to an intracortical pin experiment, the average error of a skin marker based method to estimate humerus orientation was about 10° and could be reduced to 5° when applying optimal marker weightings in a least-square algorithm. Unfortunately, there is no generic set of weightings that will systematically improve the accuracy for all kinds of movements and subjects. While non-invasive techniques to personalize weighting do not exist, using a redundant marker set without weightings
Conflict of interest statement
The authors have no conflict of interest to declare.
References (49)
- et al.
Non-invasive assessment of soft-tissue artifact and its effect on knee joint kinematics during functional activity
J. Biomech.
(2010) - et al.
Motion data processing and wobbling mass modelling in the inverse dynamics of skeletal models
Mech. Mach. Theory
(2007) - et al.
Do kinematic models reduce the effects of soft tissue artefacts in skin marker-based motion analysis? An in vivo study of knee kinematics
J. Biomech.
(2010) - et al.
A linear soft tissue artefact model for human movement analysis: proof of concept using in vivo data
Gait Posture
(2012) - et al.
Kinematics estimation of straddled movements on high bar from a limited number of skin markers using a chain model
J. Biomech.
(2008) - et al.
Effect of skin movement artifact on knee kinematics during gait and cutting motions measured in vivo
Gait Posture
(2006) - et al.
Three-dimensional kinematics of the upper limb during a Reach and Grasp Cycle for children
Gait Posture
(2010) - et al.
A hip joint kinematics driven model for the generation of realistic thigh soft tissue artefacts
J. Biomech.
(2013) - et al.
Non-invasive assessment of soft tissue local displacement during movement: a feasibility study
J. Biomech.
(2009) - et al.
Position and orientation in space of bones during movement: experimental artefacts
Clin. Biomech. (Bristol, Avon)
(1996)
A procedure for determining rigid body transformation parameters
J. Biomech.
A solidification procedure to facilitate kinematic analyses based on video system data
J. Biomech.
Coupling between 3D displacements and rotations at the glenohumeral joint during dynamic tasks in healthy participants
Clin. Biomech.
Functionally interpretable local coordinate systems for the upper extremity using inertial & magnetic measurement systems
J. Biomech.
Generalized mathematical representation of the soft tissue artefact
J. Biomech.
Influence of joint constraints on lower limb kinematics estimation from skin markers using global optimization
J. Biomech.
Effect of sitting posture on 3-dimensional scapular kinematics measured by skin-mounted electromagnetic tracking sensors
Arch. Phys. Med. Rehabil.
The accuracy of measuring glenohumeral motion with a surface humeral cuff
J. Biomech.
The accuracy of measuring glenohumeral motion with a surface humeral cuff
J. Biomech.
Improvements in measuring shoulder joint kinematics
J. Biomech.
Human movement analysis using stereophotogrammetry. Part 3. Soft tissue artifact assessment and compensation
Gait Posture
Validity and reliability of 3D marker based scapular motion analysis: a systematic review
J. Biomech.
Bone position estimation from skin marker co-ordinates using global optimisation with joint constraints
J. Biomech.
On the use of bone and skin markers in kinematics research
Hum. Mov. Sci.
Cited by (15)
Influence of marker weights optimization on scapular kinematics estimated with a multibody kinematic optimization
2023, Journal of BiomechanicsWhich data should be tracked in forward-dynamic optimisation to best predict muscle forces in a pathological co-contraction case?
2018, Journal of BiomechanicsCitation Excerpt :Joint kinematics and skin marker STA were obtained from the experimental protocol described in Begon et al. (2015). Briefly, markers were placed on skin and on intracortical-pin clusters screwed in the left humerus of four healthy participants (Begon et al., 2015). A skeletal model was then created and joint kinematics was reconstructed from both skin and intracortical-pin markers (Begon et al., 2015; Laitenberger et al., 2015).
Effect of tibia marker placement on knee joint kinematic analysis
2018, Gait and PostureCitation Excerpt :These results were in consistent with previous studies which evidenced that changing the definition of marker sets caused only slight differences in the calculated kinematics [14,24]. However, some early studies focused on the number of markers per segment in least square algorithms and highlighted that increasing the number was a valid criterion, also, a cluster made up of four markers seemed a good practical compromise [18,21,32]. The phenomenon of error variability decreased in their studies may due to the markers further apart were affected by less mutually correlated STA [24].
Kinematic models of the upper limb joints for multibody kinematics optimisation: An overview
2017, Journal of BiomechanicsMultibody kinematics optimization with marker projection improves the accuracy of the humerus rotational kinematics
2017, Journal of BiomechanicsCitation Excerpt :In particular, the axial rotation tends to be underestimated (Cutti et al., 2005a; Ludewig et al., 2002; Schmidt et al., 1999) and errors up to 30° on the arm axial rotation were found due to STA, when using an electromagnetic sensor attached to the lateral aspect of the arm (Hamming et al., 2012). Using seven skin markers spread over the arm, this error can be reduced by about 10° (Begon et al., 2015). However, this last study has shown that single-body optimization with marker weightings fails to improve the arm kinematics, since optimal weightings are subject- and task-specific (Begon et al., 2015).
Main component of soft tissue artifact of the upper-limbs with respect to different functional, daily life and sports movements
2017, Journal of BiomechanicsCitation Excerpt :Nevertheless, regardless the task and participant, some markers such as CL5 (lateral part of the clavicle) and EpM (medial epicondyle of the humerus) were always few affected by STA. Consequently as suggested by Begon et al. (2015) these marker locations (lateral part of the clavicle and medial epicondyle of the humerus) may be relevant for a set using a limited number of markers or when the experimenter looks for a sensor location with small STA. Concerning the scapula, (Matsui et al., 2006) observed similar deviation for the two markers located near the acromion process, while the skin marker located on the inferior angle underwent the highest deviation.