Journal of Biomechanics 112 (2020) 110044

Contents lists available at ScienceDirect

Journal of Biomechanics
journal homepage: www.elsevier.com/locate/jbiomech
www.JBiomech.com

Between-session reliability of subject-specific musculoskeletal models

of the spine derived from optoelectronic motion capture data
Katelyn Burkhart a,b,c, Daniel Grindle b,d, Mary L. Bouxsein a,b,c, Dennis E. Anderson b,c,⇑
a
Harvard-MIT Health Sciences and Technology Program, Massachusetts Institute of Technology, Cambridge 02139, MA, United States
b
Center for Advanced Orthopaedic Studies, Beth Israel Deaconess Medical Center, 330 Brookline Ave., Boston 02215, MA, United States
c
Department of Orthopaedic Surgery, Harvard Medical School, Boston 02115, MA, United States
d
Division of Engineering Mechanics, Department of Biomedical Engineering and Mechanics, Virginia Polytechnic Institute and State University, Blacksburg, VA, United States

a r t i c l e i n f o a b s t r a c t

Article history: This study evaluated the between-session reliability of creating subject-specific musculoskeletal models
Accepted 1 September 2020 with optoelectronic motion capture data, and using them to estimate spine loading. Nineteen healthy
participants aged 24–74 years underwent the same set of measurements on two separate occasions.
Retroreflective markers were placed on anatomical regions, including C7, T1, T4, T5, T8, T9, T12 and
Keywords: L1 spinous processes, pelvis, upper and lower limbs, and head. We created full-body musculoskeletal
Spine loading models with detailed thoracolumbar spines, and scaled these to create subject-specific models for each
Motion analysis
individual and each session. Models were scaled from distances between markers, and spine curvature
Repeatability
Musculoskeletal model
was adjusted according to marker-estimated measurements. Using these models, we estimated verte-
Model scaling bral compressive loading for five different standardized postures: neutral standing, 45trunk flexion, 15
trunk extension, 20lateral bend to the right, and 45axial rotation to the right. Intraclass correlation
coefficients (ICCs) and standard error of measurement were calculated as measures of between-
session reliability and measurement error, respectively. Spine curvature measures showed excellent
reliability (ICC = 0.79–0.91) and body scaling segments showed fair to excellent reliability (ICC = 0.4
6–0.95). We found that musculoskeletal models showed mostly excellent between-session reliability
to estimate spine loading, with 91% of ICC values > 0.75 for all activities. This information is a necessary
precursor for using motion capture data to estimate spine loading from subject-specific musculoskele-
tal models, and suggests that marker data will deliver reproducible subject-specific models and esti-
mates of spine loading.
Ó 2020 Elsevier Ltd. All rights reserved.

1. Introduction Krumm et al., 2016; Rast et al., 2016). While these concerns have
been well studied in gait models, there has been limited prior
Musculoskeletal models simulating complex spinal anatomy investigation into how this variation might affect marker-
have been developed to better understand internal spinal loading estimated spine models and spinal loading predictions.
conditions during daily living activities (Beaucage-Gauvreau While several detailed full-body lumbar spine models exist in
et al., 2019; Bruno et al., 2017a; Ignasiak et al., 2018). Optoelec- the OpenSim community (Actis et al., 2018; Beaucage-Gauvreau
tronic motion capture systems are commonly used to provide data et al., 2019; Raabe and Chaudhari, 2016), the only detailed thora-
input for scaling of musculoskeletal models to an individual’s columbar spine model (Bruno et al., 2015) does not incorporate
specific anthropometry. However, there is a known error in opto- lower extremities. A full-body model is necessary for use in typical
electronic motion capture due to inherent system performance motion analysis applications so that ground reaction forces can be
(Richards, 1999) and more notably due to variation in marker applied to the model and whole-body motion can be analyzed. A
placement on anatomical landmarks (Della Croce et al., 1999; detailed full-body musculoskeletal model would allow for predic-
tion of thoracolumbar spine loading and trunk muscular activa-
tions during different activities, and could help inform future
⇑ Corresponding author at: Center for Advanced Orthopaedic Studies, Beth Israel
studies investigating biomechanical mechanisms behind multiple
Deaconess Medical Center, RN115, 330 Brookline Ave., Boston 02215, MA, United
States. diseases and conditions including vertebral fracture, scoliosis, back
E-mail address: [email protected] (D.E. Anderson). pain, and hyperkyphosis. Therefore, the aim of this study was to

https://doi.org/10.1016/j.jbiomech.2020.110044
0021-9290/Ó 2020 Elsevier Ltd. All rights reserved.
K. Burkhart et al. Journal of Biomechanics 112 (2020) 110044

investigate how variation in the placement of anatomical markers

affects the reliability of body segment scaling, estimated spine
curvature, and resulting compressive spine loading outcomes
when creating subject-specific musculoskeletal models.

2. Methods

2.1. Creation of full-body musculoskeletal model

We created a full-body model of the thoracolumbar spine by

combining our previously validated male and female thoracolum-
bar spine models (Bruno et al., 2017a, 2015) with a published gait
model (Gait2354) (Anderson and Pandy, 2001, 1999; Carhart,
2000; Delp et al., 1990; Yamaguchi and Zajac, 1989). These models
were developed using the OpenSim musculoskeletal modeling
software (Delp et al., 2007) and include 552 and 54 individual mus-
cle fascicles, respectively. During model combination, the pelvis
and sacrum bodies were kept from the spine models, while lower
extremity segments (femur, tibia/fibula, talus, calcaneus and toes)
were introduced from the gait model. Pelvis and sacrum center of
mass (CoM) locations were used from the Gait2354 model, as these
parameters were based on more comprehensive anatomical
sources. The hip joint, formerly connected to ground, was modified
to connect the pelvis body to the femur (on the right and left) with
three degrees-of-freedom. The pelvis-ground joint was changed to
a kinematic joint that positions the model in space relative to the
ground and allows for 6 degrees of freedom (3 rotational and 3
translational). The ankle joint was modified to be a pin joint (with
one degree-of-freedom), allowing only dorsiflexion and plantar
flexion, and the toe joint was modified to be a weld joint, allowing
no movement at the midfoot. Markers were attached to the model
bodies corresponding to the experimental marker set used in data
collections.
To match the anthropometry of the male and female thora- Fig. 1. Male (left) and female (right) full body thoracolumbar spine model created
columbar models, we scaled the mass and length properties of in OpenSim with 598 individual muscle fascicles. The male model is 175 cm and
78 kg. The female model is 163 cm and 61 kg.
the legs to that of a 78 kg, 1.75 m male and 61 kg, 1.63 m female,
respectively (De Leva, 1996; Liu et al., 1971; Pearsall et al., 1996).
Inertial properties of the lower extremity bodies, as incorporated Institutional Review Board of Beth Israel Deaconess Medical
from the gait model, were scaled according to each body’s corre- Center. All participants provided written informed consent before
sponding mass and length changes. Inertial properties for the their study session.
arms, and thoracic and lumbar segments (represented as a slice
through the trunk at a spinal level) were added from literature,
2.3. Procedure
after scaling for height and weight of our male and female models
(De Leva, 1996; Pearsall et al., 1996). The psoas muscle spans
Subjects participated in the same set of measurements on two
from the thoracolumbar spine to the femur. As such, the psoas
separate occasions, an average of 8 days apart (range: 2–17 days).
fascicles were modified from the thoracolumbar models to have
The data collection procedure was previously published (Mousavi
their distal fixed attachments occur on the lesser trochanter of
et al., 2018). At the start of each session, height and weight were
the femur and a conditional via point on the pelvis, as adapted
documented. Anatomical landmarks were palpated and retrore-
from the Gait2354 model. Other lower limb musculature was
flective markers were attached to the skin using double-adhesive
incorporated directly from the gait model, while any fascicles that
tape. Rigid clusters with four markers each were attached over
connected the pelvis to the torso were excluded as these muscles
the palpated T1, T4, T5, T8, T9, T12 and L1 spinous processes. Mark-
already exist in the thoracolumbar models. The resulting full-
ers were placed over anatomic landmarks on the pelvis, including
body models include 598 Hill-type muscle fascicles and 108
the posterior (PSIS) and anterior (ASIS) superior iliac spines and
degrees-of-freedom (Fig. 1).
iliac crests. Additional markers were placed on C7 and bilaterally
on the acromion (shoulder), lateral epicondyle of the humerus (el-
2.2. Participants bow), radial styloid process (wrist), greater trochanter of the
femur, lateral and medial aspects of the knee joint, lateral and
Nineteen healthy adult volunteers (11 male, 8 female) were medial aspects of the ankle joint, posterior heel and first metatar-
recruited for this study. The mean ± SD (range) age, height, weight sophalangeal (MTP) joint (big toe). A headband with four markers
and BMI of the participants were: 47 ± 17 (24–74) years, attached was used to track head motion, and 41 other markers
172 ± 7 (162–184) cm, 71 ± 14 (45–98) kg, and 24.0 ± 3.3 were placed on the sternum, clavicles and extremities. A 10-
(16.8–30.9) kg/m2 respectively. Individuals were excluded if they camera motion analysis system (Vicon Motion Systems, Oxford,
experienced recent back pain, or had a history of spinal surgery UK) was used to collect marker data in a neutral upright standing
or traumatic fracture, thoracic deformity, or conditions that affect position. Subjects were instructed to stand upright with one foot
balance and movement. This study was approved by the on each force plate and keep their arms at their sides.
2
K. Burkhart et al. Journal of Biomechanics 112 (2020) 110044

2.4. Data processing the right. Trunk motions were distributed through the interverte-
bral joints and pelvis from reported literature ratios for flexion/ex-
Marker positions were averaged over each neutral standing trial tension, lateral bending and axial rotation, as previously described
to obtain static marker positions. Additional estimated virtual (Bruno et al., 2015), and identical kinematics were applied to all
markers and joint centers were calculated from existing measured subjects.
marker data. Hip joint centers, and intervertebral joints from C7/T1 Static optimization and joint reaction analyses were performed
to L5/S1 were estimated from marker data using published meth- with OpenSim (version 3.3) via custom Matlab scripts to estimate
ods (Nerot et al., 2018; Peng et al., 2015). The knee and ankle joint muscle and joint reaction forces and simulated ground reaction
centers were calculated as the midpoint between the medial and forces. Force and torque residual actuators were applied between
lateral joint landmarks. A centralized ‘head’ marker was estimated the ground and left/right calcanei to simulate ground reaction
as the centroid of the 4 external headband markers. Lastly, mid- forces between the feet and floor. Our optimization routine mini-
PSIS and mid-hip joint center markers were created as the mid- mized the sum of cubed muscle activations, and joint reaction cal-
point between these existing markers. culations were used to estimate the compressive force applied to
each vertebral body during activities (Bruno et al., 2015).
2.5. Subject-specific model creation
2.7. Statistical analyses
The male or female full-body thoracolumbar spine model, as
appropriate, was scaled to the anthropometry of each participant, Primary outcomes were magnitude and reliability of segment
using height, weight, and distance between anatomical landmarks scaling factors, marker-estimated thoracic and lumbar spine curva-
or marker-derived joint centers measured during the static neutral tures, and spine loading. Reliability of each outcome was examined
standing trial of each session. Intervertebral joint centers, as using intraclass correlation coefficients (ICC) and classified as
derived from marker data, were used to set spine curvature for excellent (ICC > 0.75), fair (0.4 < ICC < 0.75) or poor (ICC < 0.4)
each participant during each session. Height and weight were (Shrout and Fleiss, 1979). Standard error of measurement (SEM)
scaled using previously described methods (Bruno et al., 2017b). as a parameter of absolute reliability indicates magnitude of error
Overall thoracic kyphosis was calculated from the model as the and within-subject variability across repeated trials, and was cal-
angle between T4 and T12 vertebral bodies, similar to the classic culated as:
Cobb angle approach. Overall lumbar lordosis was calculated from pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
the model as the angle between L1 and L5 vertebral bodies. Body SEM ¼ SD 1  ICC
segment scaling factors were applied as described in Table 1. A
height scale factor was used if a body segment was not scaled from where SD is standard deviation of the measurement. Analyses were
marker-measured distances. Absolute body segment measure- performed with MATLAB (The Mathworks Inc., Natick, MA) and
ments were calculated by multiplying the body scale factor with Prism (GraphPad Software Inc., CA, USA).
the original body segment length in the musculoskeletal model.
3. Results
2.6. Spine loading simulations
Of the nineteen healthy subjects, two subjects were excluded
We used the same prescribed motions for all subjects to inves- from analysis due to missing or hidden retroreflective markers
tigate the reliability of spine loading between sessions evaluated on the ASIS of the pelvis or on the L1 vertebral body. For body seg-
by subject-specific thoracolumbar spine models created with ment scaling, mean differences between sessions were not differ-
marker-based motion capture data. For each subject-specific ent than zero (p > 0.05) (Figs. 2 and 3). The measurements with
model, we estimated vertebral compressive loading for five stan- excellent reliability (ICC > 0.75) were the head and neck length
dardized postures: neutral standing, 45 trunk flexion, 15 trunk and width, humeri and radii, spine and foot length and width
extension, 20lateral bend to the right, and 45axial rotation to (Table 2). All other scaling measurements had fair reliability with
ICCs ranging from 0.46 to 0.75. For the full 95% confidence interval

Table 1
Segment scaling from marker data.

Body Segment Axis Scaled to distance between:

Head/Neck S-I C7 marker to head center
M-L, Diagonal distance between headband
A-P markers
Humerus All Acromion marker to lateral epicondyle of
the humerus
Radius/Ulna All Lateral epicondyle of humerus to radial
styloid process
Spine (T1-L5 vertebral S-I C7 marker to mid-PSIS
bodies) M-L, Height scale factor
A-P
Pelvis/Sacrum M-L ASIS-ASIS and PSIS-PSIS
S-I, A- Mid - hip joint center and L5/S1 joint
P
Femur All Hip joint center to knee joint center
Tibia/Fibula All Knee joint center to ankle joint center
Foot (Talus, Calcaneus, M-L, Heel marker to 1st MTP joint (big toe)
Toes) A-P
S-I Height scale factor
All other bodies All Height scale factor
Fig. 2. Absolute difference (cm) in segment scaling of marker-estimated anthro-
S-I: Superior-inferior (height), M-L: Medio-lateral (width), A-P: Anterior-posterior pometry between sessions. Box plots show median and inter-quartile range, with
(depth) black dots representing outliers.

3
K. Burkhart et al. Journal of Biomechanics 112 (2020) 110044

Table 3
Mean of Session 1 (SD), between-session ICCs (95% CI), and SEM of marker-estimated
spine curvature measures. Kyphosis is calculated as T4-T12 angle, and lordosis is
calculated as L1-L5 angle. Units are in degrees.

Measurement Mean (SD) ICC (95% CI) SEM (degrees)

Kyphosis 40.3 (7.1) 0.91 (0.77–0.96) 2.2
Lordosis 50.5 (15.2) 0.79 (0.52–0.92) 7.0

The distribution of spine loading for neutral standing across T1

to L5 spine levels is shown in Fig. 5. Subject-specific spine loading
between sessions was not significantly different for all activities.
Interquartile ranges and total range of mean difference were sim-
ilar across all levels of the spine for most simulated activities
(Fig. 6, Supplemental Fig. 3). The ICC values of spine loading from
T1 to L5 were mostly excellent, with 91% of ICC point estimates
being greater than 0.75 for all activities (Supplement Table 1). ICCs
with point estimates  0.75 occurred in the upper thoracic spine,
Fig. 3. Mean difference in percentage in segment scaling of marker-estimated specifically T1 – T7 for extension and T1 – T3 for lateral bending
anthropometry between sessions. Box plots show median and inter-quartile range, (Supplemental Table 1).
with black dots representing outliers.

4. Discussion

Table 2
Mean of Session 1 (SD), between-session ICCs (95% CI), and SEM of body segment
Models created from optoelectronic motion capture data
scaling measures (in cm). showed excellent between-session reliability for thoracic and lum-
bar spine loading for the activities investigated in this study. Tho-
Measurement Mean (SD) ICC (95% CI) SEM
racic and lumbar spine curvature measures also had excellent
Head/Neck (Width, Depth) 20.9 (0.9) 0.89 (0.71–0.96) 0.26 reliability, and the scaled body segments showed fair to excellent
Head/Neck (Height) 18.2 (1.7) 0.75 (0.42–0.90) 0.90
reliability between sessions. This is the first study to determine
Humerus 32.4 (1.8) 0.81 (0.55–0.92) 0.81
Radius 25.3 (1.8) 0.86 (0.66–0.95) 0.67 the reliability of marker-derived spine loading from marker-
Spine (Height) 48.4 (3.7) 0.85 (0.63–0.94) 1.44 based subject-specific whole-body musculoskeletal models.
Pelvis (Height, Depth) 10.3 (0.6) 0.46 (-0.005–0.76) 0.43 We found a mean difference in pelvis width, measured as the
Pelvis (Width) 19.2 (1.1) 0.65 (0.28–0.86) 0.63
distance between hip joint centers, ranging from 1.7 cm and
Femur 43.8 (2.0) 0.56 (0.14–0.81) 1.30
Tibia 39.3 (2.3) 0.68 (0.33–0.87) 1.30 1.3 cm. Another study in adolescents found that pelvis width mean
Foot (Width, Depth) 20.6 (1.2) 0.95 (0.87–0.98) 0.26 difference ranged from 2.5 cm to 1.7 cm between sessions (Kainz
et al., 2017). These differences are similar to the inter-examiner
precision reported by Della Croce et al., where precision of pelvis
marker placement was between 1.5 and 2.5 cm, and precision of
range of the ICC, all scaling measures had fair to excellent ranges hip joint center (using Bell’s method) was 1.8 cm (Della Croce
except for the pelvis (height and depth, width), femur and tibia, et al., 1999). In the femur and tibia, we found that mean differences
which ranged from poor to excellent reliability (Table 2). Bland Alt- ranged from 3.8 to 1.7 cm and 3.2 to 3.2 cm respectively. These
man plots of between session measures showed no systematic dif- ranges are slightly larger than those reported in adolescents by
ferences or proportional biases (Supplemental Fig. 1). Kainz et al., where mean difference ranged from 2.3 cm to
For spine curvature between sessions, mean differences were 2.6 cm for femur and 2.3 to 2.0 cm in the tibia, however this
not significantly different from zero, however the variability was may be partially explained by a larger amount of soft tissue over
larger for lordosis than kyphosis (Fig. 4, Table 3). Both measures bony landmarks in adults, which would affect palpation repro-
had excellent reliability with ICCs > 0.75. Bland Altman plots of ducibility (Kainz et al., 2017). Our study’s pelvis height and depth,
between session measures showed no systematic differences or scaled from the distance between mid-hip joint center and the L5/
proportional biases (Supplemental Fig. 2). S1 joint, was one of the least reliable measurements, with a 95%

Fig. 4. Mean (%) (left) and absolute (degrees) (right) difference in marker-estimated spine curvature between sessions. Box plots show median and inter-quartile range, with
black dots representing individual data points.

4
K. Burkhart et al. Journal of Biomechanics 112 (2020) 110044

technique will help assess accuracy, and while our current study
did not investigate accuracy of marker placement, prior studies
in the lumbopelvic region have shown that subject gender and
BMI can affect placement accuracy, as well as experience level of
the examiner (Cooper et al., 2013; Ferreira et al., 2017; Snider
et al., 2011).
Our results suggest that compressive loading estimates have
excellent between-session reliability across all levels of the spine
and in different activities. The lowest ICCs were observed in the
upper thoracic spine, perhaps because overall loading is lower in
this region, making it relatively more sensitive to changes in the
model that affect spinal loading outcomes. Mean differences
between sessions show that 82% of our compressive spine loads
Fig. 5. Mean (blue line) of compressive spine load (N) during neutral standing in had less than 50N of variation between sessions across all spine
session 1, with box plots showing quartiles of data estimated at each spinal level levels and most activities. The absolute minimum and maximum
(T1-L5). Black dots are representative of outliers.
of mean differences in spine loading was 132N to 175N across
all activities and levels. As we previously reported, spine curvature
is an important determinant of spine loading (Bruno et al., 2017b).
confidence interval ranging from poor to excellent reliability. This As such, across all activities except trunk extension, if spine loading
reliability variable is difficult to compare with other studies, as it between sessions changed by more than 15% at any level, this is
incorporates elements of both pelvis height and depth in one mea- largely explained by differences (>10%) in lordosis or kyphosis
sure. However, it does enforce the notion that careful pelvis and between sessions. This indicates the majority of our spine loading
lower limb marker placement is critical, as these errors may prop- differences can be described, at least in part, by the differences in
agate to estimations of hip joint centers and the L5/S1 joint. spine curvature, and that an accurate assessment and implementa-
Skin-based marker data from the spine and torso can be used to tion of spine curvature is crucial for creating subject-specific mus-
estimate internal spine curvature, as would be measured from a culoskeletal models of the spine.
sagittal radiographic image. We used previously published and val- This study has several limitations, including a limited sample
idated methods to estimate thoracic kyphosis and lumbar lordosis size of 17 individuals. However, the sample includes men and
from marker data and external spine curvature, and these methods women across an age range of 24–74 years, allowing the reliability
reported larger errors in the lumbar spine than the thoracic spine results to be applicable across the majority of the adult population
between referenced and estimated joint centers (Nerot et al., for both sexes. Markers were placed on subjects by multiple exam-
2018). This may provide rationale for the larger between-session iners, which may have introduced additional precision error into
variation in lumbar lordosis than thoracic kyphosis that was our study compared to if only one examiner placed markers on
observed in this study. Other studies estimating spine curvature each subject for each session (Della Croce et al., 1999). Addition-
from marker data with different methods also found that lumbar ally, subject-specific spinal kinematics were not used for this
lordosis reliability was equal or lower to that of thoracic kyphosis, study, as we sought to isolate the effects of marker placement on
and error (or coefficient of variation) was higher when comparing spine modeling. Moreover, dynamic activities were not examined
lordosis to kyphosis (Dunk et al., 2005; Mousavi et al., 2018; Muyor and differences in subject spinal motions between sessions would
et al., 2017; Severijns et al., 2020). Prior studies comparing skin- also contribute to inter-session differences in spine loading. For
based marker-derived spine curvature to radiographic curvature instance, if different bending or lifting techniques were used, these
show moderate to good correlation (Grindle et al., 2020; Schmid motion differences will alter muscle activation and spine load
et al., 2015; Zemp et al., 2014). In addition to reliability and accu- (Bazrgari et al., 2007). Additionally, inherent trial-to-trial motion
racy of spine curvature, accuracy of marker placement on bony variability has been found to be responsible for 8–14% of the total
anatomical landmarks is important for assessment of spine curva- variation in spine range of motion, and 14–33% of the total varia-
ture. Confirmation of marker placement with an objective imaging tion in spinal loads (Granata et al., 1999).

Fig. 6. Mean difference in compressive spine load across spinal levels during neutral standing, 45 degrees of trunk flexion, 15 degrees of trunk extension, 20 degrees of lateral
bending and 45 degrees of axial rotation. Box plots show median, interquartile range and range. Outliers are represented by dots.

5
K. Burkhart et al. Journal of Biomechanics 112 (2020) 110044

In conclusion, this is the first study to investigate how variation Vertebral Fractures in the Thoracolumbar Region. J. Bone Miner. Res. 32, 1282–
1290. https://doi.org/10.1002/jbmr.3113.
in the placement of anatomical markers affects the reliability of
Bruno, A.G., Mokhtarzadeh, H., Allaire, B.T., Velie, K.R., De Paolis Kaluza, M.C.,
body segment scaling, estimated spine curvature, and resulting Anderson, D.E., Bouxsein, M.L., 2017b. Incorporation of CT-based measurements
compressive spine loading outcomes when creating subject- of trunk anatomy into subject-specific musculoskeletal models of the spine
specific musculoskeletal models. Importantly, the majority of influences vertebral loading predictions. J. Orthop. Res. 35, 2164–2173. https://
doi.org/10.1002/jor.23524.
between-session differences in spine loading in this study may Carhart, M.R., 2000. Biomechanical Analysis of Compensatory Stepping:
be explained by differences in spine curvature applied to models, Implications for Paraplegics Standing Via FNS. Arizona State University.
especially in lumbar lordosis. Thus accurate assessment of spine Cooper, K., Alexander, L., Hancock, E., Smith, F.W., 2013. The use of pMRI to validate
the identification of palpated bony landmarks. Man. Ther. 18, 289–293. https://
curvature is crucial for subject-specific modeling. Overall, this doi.org/10.1016/j.math.2012.10.005.
study supports the use of optoelectronic motion capture to create De Leva, P., 1996. Adjustments to zatsiorsky-seluyanov’s segment inertia
full-body musculoskeletal models that can estimate subject- parameters. J. Biomech. 29, 1223–1230. https://doi.org/10.1016/0021-9290
(95)00178-6.
specific spine loading. This information informs future studies on Della Croce, U., Cappozzo, A., Kerrigan, D.C., 1999. Pelvis and lower limb anatomical
dynamic spine loading, which are important for gaining insight landmark calibration precision and its propagation to bone geometry and joint
into mechanisms contributing to back pain, vertebral fractures angles. Med. Biol. Eng. Comput. 37, 155–161. https://doi.org/10.1007/
BF02513282.
and other musculoskeletal injuries. Delp, S.L., Anderson, F.C., Arnold, A.S., Loan, P., Habib, A., John, C.T., Guendelman, E.,
Thelen, D.G., 2007. OpenSim: Open-source software to create and analyze
Declaration of Competing Interest dynamic simulations of movement. IEEE Trans. Biomed. Eng. 54, 1940–1950.
https://doi.org/10.1109/TBME.2007.901024.
Delp, S.L., Loan, J.P., Hoy, M.G., Zajac, F.E., Topp, E.L., Rosen, J.M., 1990. An interactive
The authors declare that they have no known competing finan- graphics-based model of the lower extremity to study orthopaedic surgical
cial interests or personal relationships that could have appeared procedures. IEEE Trans. Biomed. Eng. 37, 757–767. https://doi.org/10.1109/
to influence the work reported in this paper. 10.102791.
Dunk, N.M., Lalonde, J., Callaghan, J.P., 2005. Implications for the use of postural
analysis as a clinical diagnostic tool: Reliability of quantifying upright standing
Acknowledgements spinal postures from photographic images. J. Manipulative Physiol. Ther. 28,
386–392. https://doi.org/10.1016/j.jmpt.2005.06.006.
Ferreira, A.P.A., Póvoa, L.C., Zanier, J.F.C., Machado, D.C., Ferreira, A.S., 2017.
The authors would like to acknowledge Rebecca Tromp and
Sensitivity for palpating lumbopelvic soft- tissues and bony landmarks and its
Javad Mousavi for their role in subject recruitment and data collec- associated factors : A single-blinded diagnostic accuracy study. J. Back
tion. This work was conducted with support from the National Musculoskelet. Rehabil. 30, 735–744. https://doi.org/10.3233/BMR-150356.
Institute on Aging (R00AG042458) and National Institute of Arthri- Granata, K.P., Marras, W.S., Davis, K.G., 1999. Variation in spinal load and trunk
dynamics during repeated lifting exertions. Clin. Biomech. 14, 367–375. https://
tis, Musculoskeletal and Skin Diseases (R01AR073019) of the doi.org/10.1016/S0268-0033(99)00004-2.
National Institutes of Health, the Department of Orthopaedic Sur- Grindle, D., Mousavi, S.J., Allaire, B., White, A., Anderson, D.E., 2020. Validity of
gery at Beth Israel Deaconess Medical Center, and Harvard Catalyst flexicurve and motion capture for measurements of thoracic kyphosis versus
standing radiographic measurements. JOR Spine (in press).
| The Harvard Clinical and Translational Science Center (National Ignasiak, D., Rüeger, A., Sperr, R., Ferguson, S.J., 2018. Thoracolumbar spine loading
Center for Advancing Translational Sciences, National Institutes associated with kinematics of the young and the elderly during activities of
of Health Award UL1 TR001102) and financial contributions from daily living. J. Biomech. 70, 175–184. https://doi.org/10.1016/j.
jbiomech.2017.11.033.
Harvard University and its affiliated academic healthcare centers. Kainz, H., Hoang, H.X., Stockton, C., Boyd, R.R., Lloyd, D.G., Carty, C.P., 2017. Accuracy
The content is solely the responsibility of the authors and does and Reliability of Marker-Based Approaches to Scale the Pelvis, Thigh, and
not necessarily represent the official views of Harvard Catalyst, Shank Segments in Musculoskeletal Models. J. Appl. Biomech. 33, 354–360.
https://doi.org/10.1123/jab.2016-0282.
Harvard University and its affiliated academic healthcare centers, Krumm, D., Cockcroft, J., Zaumseil, F., Odenwald, S., Milani, T.L., Louw, Q., 2016.
or the National Institutes of Health. The study sponsors had no role Analytical evaluation of the effects of inconsistent anthropometric
in the study design, data collection, analysis, manuscript prepara- measurements on joint kinematics in motion capturing. Gait Posture 46, 1–4.
https://doi.org/10.1016/j.gaitpost.2016.01.024.
tion, or the decision to submit the manuscript for publication.
Liu, Y.K., Laborde, J.M., Van Buskirk, W.C., 1971. Inertial properties of a segmented
cadaver trunk: their implications in acceleration injuries. Aerosp. Med. 42, 650–
Appendix A. Supplementary material 657. https://doi.org/10.1080/10643389.2012.728825.
Mousavi, S.J., Swann, M.C., White, A.P., Tromp, R., Anderson, D.E., 2018. Between-
session reliability of opto-electronic motion capture in measuring sagittal
Supplementary data to this article can be found online at posture and 3-D ranges of motion of the thoracolumbar spine. J. Biomech. 79,
https://doi.org/10.1016/j.jbiomech.2020.110044. 248–252. https://doi.org/10.1016/j.jbiomech.2018.08.033.
Muyor, J.M., Arrabal-Campos, F.M., Martínez-Aparicio, C., Sánchez-Crespo, A., Villa-
Pérez, M., 2017. Test-retest reliability and validity of a motion capture (MOCAP)
References system for measuring thoracic and lumbar spinal curvatures and sacral
inclination in the sagittal plane. J. Back Musculoskelet. Rehabil. 30, 1319–
Actis, J.A., Honegger, J.D., Gates, D.H., Petrella, A.J., Nolasco, L.A., Silverman, A.K., 1325. https://doi.org/10.3233/BMR-169750.
2018. Validation of lumbar spine loading from a musculoskeletal model Nerot, A., Skalli, W., Wang, X., 2018. Estimation of spinal joint centers from external
including the lower limbs and lumbar spine. J. Biomech. 68, 107–114. https:// back profile and anatomical landmarks. J. Biomech. 70, 96–101. https://doi.org/
doi.org/10.1016/j.jbiomech.2017.12.001. 10.1016/j.jbiomech.2017.11.013.
Anderson, F.C., Pandy, M.G., 2001. Dynamic optimization of human walking. J. Pearsall, D.J., Reid, J.G., Livingston, L.A., 1996. Segmental inertial parameters of the
Biomech. Eng. 123, 381–390. human trunk as determined from computed tomography. Ann. Biomed. Eng. 24,
Anderson, F.C., Pandy, M.G., 1999. A dynamic optimization solution for vertical 198–210. https://doi.org/10.1007/BF02667349.
jumping in three dimensions. Comput. Methods Biomech. Biomed. Engin. 2, Peng, J., Panda, J., Van Sint Jan, S., Wang, X., 2015. Methods for determining hip and
201–231. https://doi.org/10.1080/10255849908907988. lumbosacral joint centers in a seated position from external anatomical
Bazrgari, B., Shirazi-Adl, A., Arjmand, N., 2007. Analysis of squat and stoop dynamic landmarks. J. Biomech. 48, 396–400. https://doi.org/10.1016/j.
liftings: Muscle forces and internal spinal loads. Eur. Spine J. 16, 687–699. jbiomech.2014.11.040.
https://doi.org/10.1007/s00586-006-0240-7. Raabe, M.E., Chaudhari, A.M.W., 2016. An investigation of jogging
Beaucage-Gauvreau, E., Robertson, W.S.P., Brandon, S.C.E., Fraser, R., Freeman, B.J.C., biomechanics using the full-body lumbar spine model: Model
Graham, R.B., Thewlis, D., Jones, C.F., 2019. Validation of an OpenSim full-body development and validation. J. Biomech. 49, 1238–1243. https://doi.org/
model with detailed lumbar spine for estimating lower lumbar spine loads 10.1016/j.jbiomech.2016.02.046.
during symmetric and asymmetric lifting tasks. Comput. Meth. Biomech. Rast, F.M., Graf, E.S., Meichtry, A., Kool, J., Bauer, C.M., 2016. Between-day reliability
Biomed. Engin. 22, 451–464. https://doi.org/10.1080/10255842.2018.1564819. of three-dimensional motion analysis of the trunk: A comparison of marker
Bruno, A.G., Bouxsein, M.L., Anderson, D.E., 2015. Development and Validation of a based protocols. J. Biomech. 49, 807–811. https://doi.org/10.1016/j.
Musculoskeletal Model of the Fully Articulated Thoracolumbar Spine and Rib jbiomech.2016.02.030.
Cage. J. Biomech. Eng. 137, 1–10. https://doi.org/10.1115/1.4030408. Richards, J.G., 1999. The measurement of human motion: A comparison of
Bruno, A.G., Burkhart, K., Allaire, B., Anderson, D.E., Bouxsein, M.L., 2017a. Spinal commercially available systems. Hum. Mov. Sci. 18, 589–602. https://doi.org/
Loading Patterns from Biomechanical Modeling Explain the High Incidence of 10.1016/S0167-9457(99)00023-8.

6
K. Burkhart et al. Journal of Biomechanics 112 (2020) 110044

Schmid, S., Studer, D., Hasler, C.C., Romkes, J., Taylor, W.R., Brunner, R., Lorenzetti, S., Snider, K.T., Snider, E.J., Degenhardt, B.F., Johnson, J.C., Kribs, J.W., 2011. Palpatory
2015. Using skin markers for spinal curvature quantification in main thoracic accuracy of lumbar spinous processes using multiple bony landmarks. J.
adolescent idiopathic scoliosis: An explorative radiographic study. PLoS One 10. Manipulative Physiol. Ther. 34, 306–313. https://doi.org/10.1016/j.
https://doi.org/10.1371/journal.pone.0135689. jmpt.2011.04.006.
Severijns, P., Overbergh, T., Thauvoye, A., Baudewijns, J., Monari, D., Moke, L., Yamaguchi, G.T., Zajac, F.E., 1989. A planar model of the knee joint to characterize
Desloovere, K., Scheys, L., 2020. A subject-specific method to measure dynamic the knee extensor mechanism. J. Biomech. 22, 1–10.
spinal alignment in adult spinal deformity. Spine J. 20, 934–946. https://doi.org/ Zemp, R., List, R., Guïay, T., Elsig, J.P., Naxera, J., Taylor, W.R., Lorenzetti, S., 2014. Soft
10.1016/j.spinee.2020.02.004. tissue artefacts of the human back: Comparison of the sagittal curvature of the
Shrout, P.E., Fleiss, J.L., 1979. Intraclass correlations: Uses in assessing rater reliability. spine measured using skin markers and an open upright MRI. PLoS One 9, 1–8.
Psychol. Bull. 86, 420–428. https://doi.org/10.1037/0033-2909.86.2.420. https://doi.org/10.1371/journal.pone.0095426.