Annals of Biomedical Engineering ( 2018)
https://doi.org/10.1007/s10439-018-2078-7
Trunk Hybrid Passive–Active Musculoskeletal Modeling to Determine
the Detailed T12–S1 Response Under In Vivo Loads
P. KHODDAM-KHORASANI,1 N. ARJMAND ,1 and A. SHIRAZI-ADL2
1
Department of Mechanical Engineering, Sharif University of Technology, Tehran 11155-9567, Iran; and 2Division of Applied
Mechanics, Department of Mechanical Engineering, École Polytechnique, Montreal, QC, Canada
(Received 9 December 2017; accepted 18 June 2018)
Associate Editor Thurmon E. Lockhart oversaw the review of this article.
Abstract—Biomechanical models of the spine either simplify
intervertebral joints (using spherical joints or deformable
beams) in musculoskeletal (MS) or overlook musculature in
geometrically-detailed passive finite element (FE) models.
These distinct active and passive models therefore fail to
determine in vivo stresses and strains within and load-sharing
among the joint structures (discs, ligaments, and facets). A
novel hybrid active–passive spine model is therefore devel-
oped in which estimated trunk muscle forces from a MS
model for in vivo activities drive a mechanically-equivalent
passive FE model to quantify in vivo T12–S1 compression/
shear loads, intradiscal pressures (IDP), centers of rotation
(CoR), ligament/facet forces, and annulus fiber strains. The
predicted and in vivo L4–L5 IDP and L1–S1 CoRs showed
satisfactory agreements. The FE model under commonly-
used in vitro loading (pure moments and follower loads)
predicted different kinetics from those of the hybrid model
under in vivo loads (muscle exertions and gravity loads)
contributing to suggest the inadequacy of such in vitro loads
when simulating in vivo tasks. For an improved assessment of
the injury risk, evaluation of the internal loads, and design of
implants, such hybrid models should therefore be used.
Keywords—Spine, Finite element, Musculoskeletal, Model,
Disc fiber strain, Center of rotation, Intradiscal pressure,
Muscles, Facets, Ligaments.
INTRODUCTION
Quantification of load sharing among different
passive (discs, ligaments and facets) and active (mus-
cles) structures of the spine during in vivo regular, sport
and workplace activities is essential for the improved
development of injury prevention, treatment strategies
and implants of the spine. In vivo measurements of the
Address correspondence to N. Arjmand, Department of Mechanical
Engineering, Sharif University of Technology, Tehran 11155-9567,
Iran. Electronic mail: arjmand@sharif.edu
spine loads are invasive, complex, costly, and limited
with low number of volunteers. For instance, loads
transmitted through instrumented vertebral cages have
been measured in patients.54,56 Intradiscal pressure
(IDP) at lumbar34,35,42,62,75,76 and thoracic50,61 inter-
vertebral discs has also been measured as an indirect
estimate of the spine compressive loads. Such pressure
measurements remain, however, sensitive to pressure
transducer technology (e.g., liquid-filled or piezo-re-
sistive transducers),21 needle/transducer size,10 and its
self-heating.32,33 On the other hand, there exists no
direct in vivo measurement of forces in ligaments,
contact facet joints, and disc collagen fibers. Similarly,
muscle forces can only be indirectly estimated via
electromyography (EMG); an approach that suffers
from several simplifications/assumptions in the collec-
tion and post-processing of EMG signals and the
muscle EMG–force relationship.21
In the absence of satisfactory non-invasive mea-
surement techniques, computational biomechanical
models have emerged to evaluate internal loads within
the passive–active structures of the spine. These mod-
els, however, often tend to focus on and more accu-
rately represent either active or passive structures.
Musculoskeletal (MS) models, that are primary
developed to compute forces in trunk muscles, incor-
porate often a rather detailed musculature but a sim-
plified passive (ligamentous) spine.3,4,6,16,17,25,41,70
Intervertebral joints are regularly simulated as (fric-
tional or frictionless) pivots16,17,25 or deformable
beams3,4,6 (with no facets or ligaments) that allow for
the estimation of the overall joint loads but fail to
evaluate internal stresses and strains within and load
sharing between the joint structures. Such MS models
do neither allow for the direct estimation of IDP nor
 2018 Biomedical Engineering Society
 KHODDAM-KHORASANI et al.
do they account for the load-dependant motion of the
centers of rotation (CoRs) at different levels.
On the other hand, finite element (FE) models of the
passive spine incorporate accurate detailed geometries
and nonlinear mechanical properties of the joint
components (i.e., ligaments, facet articulations, disc
annulus and nucleus, vertebrae, and endplates).
Therefore, these FE models allow for the prediction of
stresses and strains within and load sharing among
these structures under given load or displacement
conditions.22 Excluding muscles, these models are of-
ten used as complementary tools along with in vitro
measurements. Consequently, the important role of
muscle exertions is either fully overlooked or simplified
by a single follower load (FL)37,55,65,67; a virtual
compression force with varying line of action that
follows the lumbar lordosis curve while passing
through the vertebral body centroids. As a pure com-
pressive force, this FL fails to mimic the shear com-
ponents of trunk muscle forces and gravity/inertia
loads.26 Moreover, a single constant FL cannot rep-
resent the varying compressive force, due to distributed
gravity as well as the complex muscle recruitments,
along the spinal height.67 These passive models are
therefore limited when simulating in vivo activities.
To circumvent these shortcomings, a coupled hybrid
model simulating in sufficient details both active
(muscles) and passive components is required. Due to
the computational effort necessary in such coupled
models, we recently proposed a staggered solution in
which gravity loads and muscle forces computed by a
validated MS model during in vivo activities were
prescribed into a passive L4–L5 FE model.9 This study
demonstrated the feasibility of combining two distinct
(MS and FE) models to simulate in vivo activities. The
passive model however was limited only to one lumbar
segment (L4–L5). Moreover, due to different passive
joint properties used in the MS and passive FE models,
dissimilar kinematics between the two models were
computed under identical tasks. In addition, the MS
model, unlike the passive FE model, neglected the
passive disc stiffening under increasing compressive
loads.44,65
The present study therefore aims to perform a
combined MS-FE model study of the whole T12–S1
spine in which the in vivo active–passive load sharing,
internal stresses and strains, and vertebral CoRs are
determined in an iterative procedure that guarantees
identical spine kinematics and load-dependent nonlin-
ear passive properties in both models. Trunk gravity
loads and predicted muscle forces by our MS
model3,4,6 during in vivo activities are iteratively used
to drive a mechanically-equivalent passive FE model of
the T12–S1 that computes disc stresses and strains,
IDP, ligament forces, facet joint forces (FJFs) and
CoRs. For validation, predictions of the hybrid model
for L4–L5 IDP and L1–S1 CoRs are compared with
available in vivo data. It is hypothesized that (1) a MS
model (with motion segments as deformable beams)
with nonlinear passive resistance and kinematics
identical to those in the detailed FE model can itera-
tively be developed, and (2) in vitro loading (pure
moment plus FL) in passive FE models cannot gen-
erate accurate in vivo load-sharing among different
joint structures.
MATERIALS AND METHODS
T12–S1 Passive FE Model
This model is an updated version of a previously-
developed L1–S1 lumbar spine model12,64 based on CT
images of the lumbar spine of a cadaver (65 years-old
male). Six vertebrae (L1 through S1 as rigid bodies),
five discs (annulus and nucleus), 10 contact facet sur-
faces, and nine sets of ligaments (anterior longitudinal
(ALL), posterior longitudinal (PLL), capsular (CL),
intertransverse (ITL), ligamentum flavum (LF), su-
praspinous (SSL), interspinous (ISL), fascia (L4 and
L5 to ilium) and iliolumbar (IL, L5 to ilium) liga-
ments) were included. Annulus layers and disc nucle-
uses were considered as, respectively, isotropic annulus
bulk (reinforced by membranes of collagen fibres) and
fluid-filled cavities. For the current study, this model
was updated (Figs. 1a and 1b) by the (1) addition of
the T12–L1 motion segment (assuming a segmental
lordosis similar to that of the L1–L2 motion segment
and a mean disc height identical to that of the T12–L1
in the MS model), (2) substantial refinement of the
mesh (~ 1730 solid brick elements for each disc annu-
lus) to improve model predictions, (3) use of hypere-
lastic Mooney–Rivlin material properties (c1 = 0.42,
c2 = 0.10515) for the ground substance of the annulus
bulk, (4) representation of 14 distinct collagen fiber
layers using membrane layers reinforced by rebar ele-
ments (with criss-cross patterns at ± 3031 and non-
linear stress–strain data for material properties66,68)
distributed in the foregoing annulus ground substance,
and (5) simulation of facet articulations by frictionless
surface-to-surface contacts with average gap limit of
1.25 mm (simulating compliant articular cartilage
layers). Similar to the original model, bodies and
posterior bony elements of the vertebrae were distinct
rigid elements attached together via two deformable
beams at pedicles.64 Ligaments were modeled by mul-
tiple axial tension-only elements with nonlinear prop-
erties.66 Each nucleus was an incompressible fluid-filled
cavity.43
 Trunk Hybrid Passive–Active Model for T12–S1 Responses

FIGURE 1. The T12–S1 passive finite element (FE) model: (a) 3D mesh generation and (b) detailed view of the L4–L5 disc
including its ground substance, nucleus cavity and 14 circumferential rebar layers of disc annulus. The musculoskeletal (MS)
model and its musculature in (c) sagittal and (d) frontal planes that include 46 local (originating from the pelvis and inserting into
lumbar vertebrae) and 10 global (originating from the pelvis and inserting into the thoracic spine) fascicles of muscles (ICPL:
iliocostalis lumborum pars lumborum, ICPT: iliocostalis lumborum pars thoracic; IP, iliopsoas; LGPL, longissimus thoracis pars
lumborum; LGPT, longissimus thoracis; MF, multifidus; QL, quadratus lumborum; IO, internal oblique; EO, external oblique; RA,
rectus abdominus). (e) Sagittal and (f) frontal view of the FE model with the axial connector elements used to apply muscle forces.

MS Model incorporated (Figs. 1c and 1d).71 Curved lines of ac-

Our previously-developed MS model3,4,6,8 evaluated
forces in trunk muscles during eight in vivo activities
using an optimization algorithm described below. The
geometry of the spine (dimensions and lumbar lordo-
sis) in the MS and FE models was identical. The pelvis,
T1–T12 thorax and lumbar vertebrae were rigid
(Figs. 1c and 1d). T12–L1 through L5–S1 segments
were simulated by 3-node nonlinear shear-deformable
beams located 4 mm posterior to the disc centers to
account for the shift in the disc CoR observed in the
passive FE model under different loads.66 The trunk
weight of ~ 344 N (~ 52%51 of the total body weight of
the male subject in our earlier studies with a body mass
of ~ 68 kg) was partitioned among upper arms
(~ 36 N), forearms/hands (~ 29 N), head (46 N) (ap-
plied via rigid elements at their centers of mass on the
rigid thorax recorded in our in vivo measurements),3,4
and T1–L5 segments (~ 233 N between the T1 through
L5 segments and applied via rigid elements at their
centers of mass located off center at each vertebra.
Segmental gravity forces and their corresponding lever
arms in the sagittal plane are provided in supplemen-
tary data).5,49 For tasks with a hand load, a force in the
gravity direction was also applied via a rigid element
attached to the rigid thorax at the center of mass of the
hand load measured in vivo. In total, 56 muscles were
tion of back muscles (i.e., their wrapping around and
contact forces at bony vertebrae) were considered.8
Based on our earlier in vivo measurements for each
task,3,4 the rotation of thorax (T), pelvis (P), and
lumbar (L = T 2 P; partitioned between T12 and L5
vertebrae6) as well as gravity/hand loads were pre-
scribed into the model. To determine muscle forces, an
optimization algorithm minimizing sum of cubed
muscle stresses was used. Overall joint loads (local
compression and shear loads at the beams) were sub-
sequently evaluated.
Distinct Validation of the MS and FE Models
The MS model was previously validated by com-
paring its predicted muscle forces and compressive
loads vs. measured muscle EMG and IDPs, respec-
tively.3,4,6,52 As to the passive FE model, range of
motions (RoMs), IDPs, and FJFs predicted under
single/combined moments and FL38 (in vitro loading,
Table 1) were compared with in vitro and in silico data.
The L5 vertebra was fixed in these simulations.
Moreover, a loading condition suggested elsewhere to
simulate forward flexion (i.e., 7.5 Nm flexion plus a FL
of 1175 N)22,60 was also considered in the FE model
and its predictions were compared with those of the
 KHODDAM-KHORASANI et al.
TABLE 1. Combined moment and follower load
conditions considered for the passive FE model.
Loading FL (N) Moment (Nm) References
Flexion 1175 7.5 Rohlmann et
Extension 500 7.5 Rohlmann et
Lateral bending 700 7.8 Dreischarf et
Axial rotation 720 5.5 Dreischarf et
hybrid model under in vivo loading (muscle exertions).
For this latter simulation, the S1 translational degrees-
of-freedom were constrained.
Simulated Tasks
Eight symmetric static tasks including the neutral
upright standing posture, forward flexion of 20, 40,
60, 80 and 107 (full flexion) with arms in the gravity
direction, and upright two-handed lifting of 19.8 kg at
25 cm (close) and 55 cm (away) anterior distances to
the L5–S1 disc were considered.3,4,24 These tasks were
chosen due to the availability of the measured L4–L5
IDP76 for subsequent comparisons and the required
kinematics3,4 to drive simulations.
Hybrid Model
For each task, the associated pelvic rotation mea-
sured in vivo and muscle forces estimated in the MS
model (via connector elements in between insertion
points as in Figs. 1e and 1f) at the deformed configu-
rations were prescribed into the passive FE model.
Each uniaxial connector element ran between the up-
per insertion point of a muscle to its lower insertion
point. These insertion points were either located on
bony vertebrae in the FE model or attached via rigid
elements to the bony structures (Figs. 1e and 1f). The
force–elongation property of each connector element
was defined so that it generated a constant axial force
(equal to the muscle force estimated from the MS
model) regardless of the elongation in the connector
element. As for the curved (wrapping) global thoracic
muscles (LGPT and ICPT), multi-segment connectors
with a unique force in between insertions and wrapping
points on vertebrae in contact were considered. In
addition, gravity loads and weight in hands were
applied at their locations measured in vivo. Despite
identical loading conditions, different deformed con-
figurations (between MS and passive FE models) were
computed due to their distinct simulations (e.g., com-
pression-dependent stiffness) of motion segments. To
assure identical final deformed T12–S1 kinematics
between the MS and FE models (the T1-T12 segment is
considered as a single rigid body in both models), the
(FL) analysis of MS model was repeated after adjustments
in its passive, compression-dependent stiffness (mo-
ment-rotation) properties based on those in the passive
FE model under gravity loads, weight held in hands
al.60 and muscle exertions. The updated muscle forces were
al.60
al.20
again prescribed into the passive FE model and this
al.19 iterative procedure was continued until convergence,
i.e., the two models had almost similar kinematics for
the simulated task (average difference of < 1 mm
between the MS and FE models for the final positions
of the T12–S1 centroids). In all simulations, the S1
translational degrees-of-freedom were constrained,
pelvis rotation was prescribed based on the measured
value in vivo while remaining rotational and transla-
tional degrees-of-freedom at the vertebral levels were
left unconstrained (i.e., predicted by the hybrid model).
A flowchart representing the method was provided in
supplementary data.
Model Outputs
T12–S1 IDPs, CoR of each vertebra with respect to
its lower vertebra,37,46 ligament forces, FJFs, and
maximal disc fiber strains were computed for each
in vivo activity. Moreover, based on the free body
diagram at the disc mid-height, ligaments, and facet
joints at each level of the passive FE model, and con-
sideration of the force/moment equilibrium require-
ments at the disc centroid, the resultant spinal loads
(local compression, shear, and passive moment) as well
as force/moment sharing among discs, ligaments and
facet joint were determined37 for each activity. For this
purpose, force-moment vectors in ligaments and con-
tacting facet joints were directly calculated from nodal
forces in the FE model while those at the disc were
computed by solving the force equilibrium equations in
axial and shear directions as well as the moment
equilibrium equations in the sagittal plane.
Validations of the Hybrid Model
The predicted L4–L5 IDP in each task and L1–S1
CoRs in flexion tasks were compared, respectively,
with the corresponding in vivo IDP in one healthy male
subject76 and in vivo image-based (upright-to-flexion)
CoR in ten male subjects.46 The body weight (~ 68 kg)
considered in the hybrid model and that of the subject
(70 kg) who participated in the IDP measurements76
were similar thus making such comparisons more rel-
evant (body weight has a great impact on spinal
loads2,24,25). Moreover, L4–L5 IDP was also estimated
in the MS model using an equation that relates IDP to
axial compression force and intersegmental flexion
rotation.24
 Trunk Hybrid Passive–Active Model for T12–S1 Responses

RESULTS

Validations of the Passive FE Model

RoM of the L4–L5 disc-vertebra unit model alone
(without ligaments and facets) loaded by a 10 Nm pure
moment in different planes fell within the in vitro range
for the human disc specimens27 (Fig. 2a). The L1-L5
intact model (with ligaments and facets) under a 7.5
Nm moment in different directions, yielded RoMs in
agreement with in vitro57 and in silico22 data (Fig. 2b).
The model loaded with a FL up to 1000 N predicted
L4–L5 IDPs within the in vitro13 and other FE model22
ranges (Fig. 2c). In combined FL and moment in dif-
ferent anatomical planes,58–60 the model also predicted
intervertebral rotations, IDP, and FJFs (Figs. 3a–3k)
within the range of numerical predictions22 though
generally in disagreement with in vivo data for IDP76
and intervertebral rotations.45,47,48

Hybrid Model Predictions

Convergence
Depending on the simulated task and after 3-4
iterations (20-30 min computational time), similar
kinematics were found between the FE and MS models
(differences of < 1 mm and 0.6 in, respectively, dis-
placements and rotations of the T12–S1 vertebral
centroids) under identical passive properties and
muscle forces (Fig. 4). For different tasks, the com-
pression-induced increase in disc stiffness remained, in
agreement with in vitro measurements,44 below 20%.

L5–S1 Load Sharing

Resultant spinal loads (total joint compression and
shear forces) and passive joint moments were almost
identical in both models. Contributions of various
joint structures to resist the overall spinal forces/mo-
ments could however be estimated only by the hybrid
model, i.e., the passive FE model under muscle forces
and gravity loads (Fig. 5). Generally, the disc had the
greatest contribution to resist compression and shear
loads (Figs. 5a and 5b). Ligaments resisted greater
shear forces at larger trunk flexion angles while facets
resisted greater shear and compression loads in tasks
involving larger extension (or less flexion) angles
(Figs. 5a and 5b). Predicted IDPs, mean (left and
right) FJFs, and maximal disc fiber strains were found
to be disc-level and task dependent (Fig. 6). Regardless
of the simulated task, the maximal IDP occurred either
at the upper (L1–L2) or lowermost (L5–S1) disc level
(Fig. 6a). Moreover, IDP increased at all lumbar levels
with the trunk flexion and extension from the relaxed
upright posture. FJFs were relatively small in the up-
FIGURE 2. Comparisons between the passive FE model
predictions and in vitro/numerical data obtained under (a)
moment of up to 10 Nm to the L4–L5 disc-alone model
(without ligaments and facets), (b) moment of up to 7.5 Nm
to the L1–L5 intact model (with ligaments and facets), and (c)
a follower load (FL) of up to 1000 N. Numerical ranges are
provided based on eight different validated FE models
available in the literature.22 In vitro data for the disc-alone
moment-RoM, intact L1–L5 moment-RoM, and FL-IDP graphs
are based on the work of Heuer et al.,27 Rohlmann et al.57 and
Brinckmann and Grootenboer,13 respectively. For the L1–L5
RoM–moment curve in lateral bending, our model being in
agreement with several others22 shows a rather linear
behavior in contrast to in vitro data. The moment–RoM
response in lateral bending is influenced by the facet
contact surfaces that vary from an individual to another.
right and flexed postures and considerably increased,
especially at lower lumbar levels, in tasks involving
trunk extension (Fig. 6b). Disc fiber strains were lar-
 KHODDAM-KHORASANI et al.

FIGURE 3. Passive FE model predictions for intervertebral rotations, intradiscal pressure (IDP), and facet joint forces (FJF)
(average of left and right contact forces) are compared with experimental/numerical data obtained under (a and b) 7.5 Nm flexion
moment and 1175 N follower load, (c–e) 7.5 Nm extension moment and 500 N follower load, (f–h) 7.8 Nm lateral bending moment
and 700 N follower load, and (i–k) 5.5 Nm axial moment and 720 N follower load. Numerical ranges are provided based on eight
different validated FE models available in the literature.22 Experimental data for the intervertebral rotations are based on the in vivo
three-dimensional radiographic images of lumbar spine (Pearcy and Tibrewal46; Pearcy et al.45; Pearcy47) while those for the IDP
data are based on the measured in vivo values using a pressure transducer inserted into the L4–L5 disc.75 Note that these loading
conditions (combined moment and follower load) do not accurately represent any specific in vivo loading task but, similar to
previous model investigations, such crude comparisons with in vivo data are nevertheless made here. Moreover, the geometry of
our FE model does not necessarily match that of the subjects participated in the foregoing experimental studies. In vivo IDPs were
collected in only one subject thus neglecting the inter-individual variabilities.

gest either at the upper (L1–L2) or lowermost (L5–S1)
disc and increased with the trunk flexion or extension
from the upright posture (Fig. 6c). Radial location of
the maximum fiber strain was generally in the inner-
most layer while its circumferential position changed
depending on the task; in posterolateral regions (at the
upper annulus-endplate junction) in neutral upright
and flexed tasks63 and in lateral regions (at the lower
annulus-endplate junction) in extended (i.e., load-
holding) tasks (Fig. 7). Maximal fiber strain for the
outermost layer occurred in the anterolateral regions
(at disc mid-height) for all the simulated tasks. Fiber
strains in the innermost layers were mainly influenced
by intervertebral rotations while in the outermost
layers were primarily affected by the disc bulging (i.e.,
IDP). For the two load-holding tasks, segments in
extension (L2–S1) had their maximum fiber strains at
the innermost layer in lateral regions (at the lower
annulus-endplate junction) and for the outermost layer
in the anterolateral regions (at the disc mid-height).
 Trunk Hybrid Passive–Active Model for T12–S1 Responses
FIGURE 4. Schematic of the passive FE model in the undeformed (supine posture without any loading) and deformed (full flexion
under gravity loads and muscle exertions) configurations at different iterations. Symbols indicate position of geometric center of
vertebrae.
In Vivo vs. In Vitro Loading Conditions
The in vitro loading proposed elsewhere to simulate
most realistically in vivo forward flexion (7.5 Nm
flexion plus a FL of 1175 N),22,60 when applied to the
passive FE model resulted in different patterns of IDPs
(e.g., decreasing from upper toward lower levels) and
disc fiber strains as compared to those predicted by the
hybrid model under the in vivo full flexion task (Fig. 6).
Validation of the Hybrid Model for the CoR and IDP
For the tasks involving trunk flexion and in general
agreement with in vivo measurements,46 the T12–S1
CoRs for both FE and MS models were located ~ 6–
8 mm posterior to the disc centroid (Fig. 8a). In cor-
roboration, location of the L2–L3 and L3–L4 CoRs in
trunk flexion of 45 (with pelvic being constrained) has
also been measured (radiographic imaging) to be at
~ 5.0 mm posterior to vertebral body centroid.77
Moreover, predicted and in vivo76 L4–L5 IDPs were in
satisfactory agreement (R2 = 0.99 and RMSE= 0.38
MPa) (Fig. 8b).
DISCUSSION
This study presents a novel hybrid passive–active
model of the T12–S1 spine to determine tissue stresses
and strains, joint load sharing, and CoRs during in vivo
activities. To do this, trunk muscle forces predicted by
a musculoskeletal (MS) model of the thoracolumbar
spine were used to drive a mechanically-equivalent,
geometrically-detailed, passive FE model of the T12–
S1 spine. For a given simulated task and after an
iterative approach, the MS and passive FE models
reached similar geometries (height, lordosis, and ver-
tebral orientations), deformed configurations, gravity
and muscle loadings, boundary conditions, and passive
stiffness thus confirming our first hypothesis that an
equivalent detailed passive FE model of a MS model
can be used to predict joint load sharing under similar
kinematics, gravity and muscle forces. It is to be noted
that the iterative interaction between the two models
that results in the accurate estimation of muscle forces
accounting for the complex nonlinear passive response
and of detailed passive structure load-sharing under
realistic external loads is original. Predictions of the
hybrid model for the L4–L5 IDPs and L1–S1 CoRs
showed adequate agreements with measured in vivo
data. Results also indicated that pure moments com-
bined with a single-valued FL fail to mimic the in vivo
loading generated by muscle exertions and gravity
loads thus confirming our second hypothesis.
Limitations
The optimization algorithm used in the MS model
to estimate muscle forces, i.e., minimizing sum of cu-
 KHODDAM-KHORASANI et al.

FIGURE 5. Comparison of (a) compression, (b) shear and (c) moment sharing of different L5–S1 joint structures (disc, ligaments
and facets) under the simulated in vivo tasks by the hybrid-model (for the sake of clarity, compression sharing on the vertical axis
was shown from 75% to 100%). For facets, sum of bilateral contact forces is considered. For all the simulated tasks facets resist
extension moment (i.e., generate flexion moment). As ALL and PLL are attached to the disc, their forces were included in
calculation of disc forces.

bed muscle stresses, overlooked antagonistic (mainly
abdominal) muscle forces. It has been shown that this
cost function generates back muscle activities in best
qualitative agreement, as compared to seven other cost
functions, with EMG data for a number of loaded and
unloaded tasks.7 The other cost functions that also
estimate muscle activities in qualitative agreement with
EMG data generate similar lumbar compression and
shear loads with differences less than 10%. Moreover,
only major local and global muscles attaching the
 Trunk Hybrid Passive–Active Model for T12–S1 Responses

FIGURE 6. (a) Intradiscal pressure values (IDP), (b) mean (left and right) facet joint forces (FJFs), and (c) maximal disc fiber
strains at all lumbar levels during different simulated in vivo tasks by the hybrid model. Results of the passive FE model under the
proposed in vitro loading condition for flexion (7.5 Nm and 1175 N FL)22,60 are also depicted.

pelvis-sacrum, respectively, to the lumbar and thoracic
spines were considered in the present study while
neglecting the smaller intersegmental muscles; a sim-
plification whose effect on the predicted joint load
sharing remains to be investigated. Neglecting antag-
onistic muscle forces in the present study is also sug-
gested to underestimate spine loads.72 Predicted muscle
activities in agonistic (back) muscles, however, re-
mained in qualitative agreement with our measured
EMG data3,4,6 even after the iterative procedure that
resulted in small to moderate changes (< 20%) in the
estimated muscle forces.
Although the passive FE model was validated by
comparing its predictions for RoMs, IDPs, and FJFs
under various loadings with available measured data
(Figs. 2 and 3), it is however to be noted that such
agreements could have also been obtained with some-
what alternative sets of material properties for the discs
and ligaments.22,36 In a recent study, we showed that
changes in the stiffness of passive tissues (ligament
properties and their initial laxities) could affect the
predicted joint load sharing under moment/FL loading
conditions.36 For this reason, a model should prefer-
ably be validated as much as possible in different
loading conditions (magnitudes, directions and com-
binations). It is also to be noted that while wrapping of
back muscles around bony structures was accounted
for in the FE model, the inter-muscular forces were
neglected. Muscles were simulated by 1D, rather than
3D, elements in the current FE model. Moreover,
estimation of annulus fiber strains were made assuming
constant annulus fibrosus criss-cross angles at ± 30,
 KHODDAM-KHORASANI et al.

FIGURE 7. Radial and circumferential locations (shown by thick solid red and blue lines for, respectively, the innermost and
outermost layer) and magnitudes of the maximal fiber strain in different levels and tasks.

FIGURE 8. Predictions of the hybrid FE/MS model as compared with experimental data obtained under in vivo loading conditions:
(a) sagittal T12–S1 center of rotations (CoRs) for tasks involving trunk flexion and (b) L4–L5 intradiscal pressure (IDP) during all
simulated tasks. IDP predictions based on the MS model for two different L4–L5 disc cross-sectional areas, CSA, (1455 mm2 similar
to the disc CSA in the MS and passive FE models, R2 = 0.88 and RMSE= 0.35 MPa, or 1800 mm2 similar to the disc CSA of the
subject that participated in the measurement study,75 R2 = 0.91 and RMSE= 0.34 MPa) are also depicted. Center of rotation of each
vertebra, relative to its inferior vertebra in different tasks was depicted on the schematic of the FE model in upright posture and
was compared with radiographic in vivo measurements48 (in vivo CoR for the T12–L1 segment was not available). In vivo
measurements of the L4–L5 IDPs are based on data reported by Wilke et al.75
 Trunk Hybrid Passive–Active Model for T12–S1 Responses

while the neglected additional complexity of the disc Analyses of Findings

fiber organization and likely inhomogeneous distri-
The in vitro loading proposed elsewhere to most
bution of these angles18,28,79 could influence predic-
realistically represent in vivo forward flexion22,60 failed,
tions.40
when applied to the passive FE model, to yield pat-
All three models (MS, FE, and hybrid) were vali-
terns of IDPs and disc fiber strains in agreement with
dated as much as possible by comparing their pre-
those predicted by the hybrid model under forward
dictions with available experimental data. For MS
flexion tasks. For instance, the suggested in vitro
model, estimated muscle activities and L4–L5 com-
loading (7.5 Nm flexion plus 1175 N FL) resulted in a
pressive loads were compared with, respectively,
monotonically decreasing IDP from the upper
EMG data3,4,6 and measured IDPs (Fig. 8b). The
(~ 1.4 MPa at the L1–L2) toward the lower
passive FE model was validated by the comparison of
(~ 1.0 MPa at the L5–S1) discs (Fig. 6a). This
its predictions under pure moment in different planes,
decreasing trend was in agreement with predictions of
pure follower load, and combinations thereof with
five other passive FE models under the proposed
measurements (Figs. 2 and 3). Finally, for the hybrid
in vitro loading condition.22 Our hybrid model under
model, that is built from these validated MS and FE
muscle exertions and gravity loads, however, predicted
models, predictions for the T12–S1 CoRs and L4–L5
relatively larger IDPs at the lower discs for all flexion
IDPs were compared with in vivo data (Fig. 8).
activities is expected under the additional segmental
Comparisons of the predicted and measured L4–L5
masses and local lumbar muscle forces acting from the
IDPs (Fig. 8b) should be carried out in the light of
upper toward the lower levels. The same discrepancy is
few facts. Our MS and FE models were different
found when comparing the maximum disc fiber strains
from the single subject that participated in the IDP
(Fig. 6c).
measurement study76 in terms of size, lordosis, mus-
The disagreement between predictions of the FE
culature, as well as geometry and structural-material
model (under in vitro loading) and measured in vivo
properties of the passive tissues. A more accurate,
data for IDP and intervertebral rotations is evident
though currently unavailable, validation procedure
under all loads considered (Fig. 3). Such findings fur-
involves modeling and measurements on the same
ther demonstrate that an in vivo loading condition
subjects. Nevertheless, a number of parameters
cannot accurately be represented by a constant varia-
including mechanical properties of the passive tissues,
tion of both moment and FL (i.e., in vitro loading)
accurate geometry of ligaments and discs, as well as
throughout the lumbar spine. In other words, the
detailed musculature cannot yet accurately be made
complex recruitment patterns of trunk muscles pre-
subject-specific. Note that the size of the subject
dicted by our MS model cannot be simply mimicked by
based on which the MS and FE models are devel-
a constant FL. In the full flexion task, for instance,
oped (68.4 kg and 174.5 cm) match closely that of
segmental loads at the T12–L1 through L5–S1 levels
the subject participated in the IDP measurement
were, respectively, ~ 1283, 1346, 1357, 1459, 1558,
study (70 kg and 173.9 cm).76 This concordance
1508 N in compression and ~ 100, 92, 39, 70, 79 and
nevertheless makes presented comparisons more rel-
304 N in anterior shear forces. Such proximal to distal
evant. The L4–L5 disc cross-sectional areas were
increases in compression forces, even when neglecting
however quite different between the FE model
accompanying shear forces, demonstrates the need for
(1455 mm2) and the subject that participated in the
a variable FL instead of a constant one.30,60 Multiple
in vivo measurement study (1800 mm2).76 Therefore,
FL applications67 along with flexion rotations or mo-
comparisons were made (Fig. 8b) for both values in
ments that both vary from a level to another are hence
the MS model. Differences in the disc heights also
essential when attempting to accurately simulate an
affect model predictions.22,39,78 Furthermore, in vivo
in vivo task.
IDPs recorded on one subject overlook the inherent
With the maximum fiber strains remained below
inter-individual variabilities. The equation relating
their failure tensile strain, i.e., ~ 15–25%,63,69 no disc
the L4–L5 IDP to joint compression and flexion in
failure is expected in the currently simulated tasks.
the MS model,24 that was used in this work, was
However, the posterolateral region (near annulus–
developed for rather small extension intersegmental
endplate junctions) could be at the risk of failure ini-
rotations (< 2) and thus may not be as accurate for
tiation under greater compression and flexion rota-
IDP predictions in tasks involving larger extension
tions.1,63,74 Our results are also corroborated by
rotations (load holding away from the chest). Finally,
findings of both in vivo (imaging and intraoperative
changes in the applied L1–L5 lumbar rotations in the
examinations) and in vitro studies that indicate lumbar
MS model when simulating a specific in vivo task
disc herniation commonly occurs at the endplate or
could also influence predictions and hence associated
annulus-endplate junctions.53,73,74
comparisons.
 KHODDAM-KHORASANI et al.
The MS model with T12–S1 segments simulated by
shear-deformable beam elements located at fixed pos-
terior distances (4 mm) to the disc geometric centers
predicted CoRs in flexion tasks that differed from those
of the FE model (under in vivo loading conditions) by
~ 3 mm in average for all levels/tasks. Compared to the
FE model, the MS model generally predicted more
posterior CoRs at the upper levels whereas more
anterior CoRs at the lower vertebrae. Nevertheless,
both MS and FE models predicted CoRs to be within
the range of those recorded in vivo. Our recent modeling
investigation has demonstrated the substantial effects
of the joint (simulated as beams or spherical pivots)
positioning on muscle forces and spinal loads.23 The
alterations in CoR position with loading at different
levels and their effects on spinal loads hence caution
against the use of fixed spherical joints regularly placed
at the joint geometric center.11,14,25,29
In conclusion, this study proposed a novel approach
to estimate the lumbar joint load sharing, internal
stresses and strains, and CoRs during various in vivo
activities by using a combined hybrid (validated MS
and passive FE) model. Predictions of this hybrid
model for the L4–L5 IDPs and L1–S1 CoRs compared
well with the measured in vivo values. It was also
shown that the loading conditions proposed in the
literature with fixed moment and FL magnitudes
throughout the lumbar spine cannot adequately rep-
resent in vivo loading conditions. Passive FE models,
when applied to assess the risk of injury or to design
implants, should therefore be used under loads closely
replicating in vivo loading conditions.
ELECTRONIC SUPPLEMENTARY MATERIAL
The online version of this article (https://doi.org/10.
1007/s10439-018-2078-7) contains supplementary
material, which is available to authorized users.
ACKNOWLEDGMENTS
This work was supported by grants from Sharif
University of Technology (Tehran, Iran). Assistance of
Mr. H. Asadi and Dr. S. Naserkhaki in the finite ele-
ment modeling and Mr. A.H. Eskandari in the mus-
culoskeletal modeling is greatly appreciated.
REFERENCES
1
Adams, M., and W. Hutton. Prolapsed intervertebral disc:
a hyperflexion injury. Spine 7:184–191, 1982.
2
Akhavanfar, M., H. Kazemi, A. Eskandari, and N. Arjmand.
Obesity and spinal loads: a combined MR imaging and
subject-specific modeling investigation. J. Biomech. 70:
102–112, 2018.
3
Arjmand, N., D. Gagnon, A. Plamondon, A. Shirazi-Adl,
and C. Lariviere. Comparison of trunk muscle forces and
spinal loads estimated by two biomechanical models. Clin.
Biomech. 24:533–541, 2009.
4
Arjmand, N., D. Gagnon, A. Plamondon, A. Shirazi-Adl,
and C. Lariviere. A comparative study of two trunk
biomechanical models under symmetric and asymmetric
loadings. J. Biomech. 43:485–491, 2010.
5
Arjmand N. Computational biomechanics of the human
spine in static lifting tasks. PhD Thesis, Ecole Polytech-
nique de Montreal, pp. 218-220, 2006.
6
Arjmand, N., and A. Shirazi-Adl. Model and in vivo
studies on human trunk load partitioning and stability in
isometric forward flexions. J. Biomech. 39:510–521,
2006.
7
Arjmand, N., and A. Shirazi-Adl. Sensitivity of kinematics-
based model predictions to optimization criteria in static
lifting tasks. Med. Eng. Phys. 28:504–514, 2006.
8
Arjmand, N., A. Shirazi-Adl, and B. Bazrgari. Wrapping
of trunk thoracic extensor muscles influences muscle forces
and spinal loads in lifting tasks. Clin. Biomech. 21:668–675,
2006.
9
Azari, F., N. Arjmand, A. Shirazi-Adl, and T. Rahimi-
Moghaddam. A combined passive and active muscu-
loskeletal model study to estimate L4–L5 load sharing. J.
Biomech. 70:157–165, 2018.
10
Bashkuev, M., P.-P. A. Vergroesen, M. Dreischarf, C.
Schilling, A. J. van der Veen, H. Schmidt, and I. Kingma.
Intradiscal pressure measurements: a challenge or a rou-
tine? J. Biomech. 49:864–868, 2016.
11
Bassani, T., E. Stucovitz, Z. Qian, M. Briguglio, and F.
Galbusera. Validation of the AnyBody full body muscu-
loskeletal model in computing lumbar spine loads at L4L5
level. J. Biomech. 58:89–96, 2017.
12
Breau, C., A. Shirazi-Adl, and J. De Guise. Reconstruction
of a human ligamentous lumbar spine using CT images—a
three-dimensional finite element mesh generation. Ann.
Biomed. Eng. 19:291–302, 1991.
13
Brinckmann, P., and H. Grootenboer. Change of disc
height, radial disc bulge, and intradiscal pressure from
discectomy an in vitro investigation on human lumbar
discs. Spine 16:641–646, 1991.
14
Bruno, A. G., M. L. Bouxsein, and D. E. Anderson.
Development and validation of a musculoskeletal model of
the fully articulated thoracolumbar spine and rib cage. J.
Biomech. Eng. 137:081003, 2015.
15
Chen, C.-S., C.-K. Cheng, C.-L. Liu, and W.-H. Lo. Stress
analysis of the disc adjacent to interbody fusion in lumbar
spine. Med. Eng. Phys. 23:485–493, 2001.
16
Cholewicki, J., and S. M. McGill. Mechanical stability of
the in vivo lumbar spine: implications for injury and
chronic low back pain. Clin. Biomech. 11:1–15, 1996.
17
Damsgaard, M., J. Rasmussen, S. T. Christensen, E. Sur-
ma, and M. de Zee. Analysis of musculoskeletal systems in
the anybody modeling system. Simul. Model. Pract. Theory
14:1100–1111, 2006.
18
Dittmar, R., M. M. van Rijsbergen, and K. Ito. Moder-
ately degenerated human intervertebral disks exhibit a less
geometrically specific collagen fiber orientation distribu-
tion. Global Spine J 6:439–446, 2016.
19
Dreischarf, M., A. Rohlmann, G. Bergmann, and T. Zan-
der. Optimised loads for the simulation of axial rotation in
the lumbar spine. J. Biomech. 44:2323–2327, 2011.
 Trunk Hybrid Passive–Active Model for T12–S1 Responses
20
Dreischarf, M., A. Rohlmann, G. Bergmann, and T. Zan-
der. Optimised in vitro applicable loads for the simulation
of lateral bending in the lumbar spine. Med. Eng. Phys.
34:777–780, 2012.
21
Dreischarf, M., A. Shirazi-Adl, N. Arjmand, A. Rohl-
mann, and H. Schmidt. Estimation of loads on human
lumbar spine: a review of in vivo and computational model
studies. J. Biomech. 49:833–845, 2016.
22
Dreischarf, M., T. Zander, A. Shirazi-Adl, C. Puttlitz, C.
Adam, C. Chen, V. Goel, A. Kiapour, Y. Kim, and K.
Labus. Comparison of eight published static finite element
models of the intact lumbar spine: Predictive power of
models improves when combined together. J. Biomech.
47(8):1757–1766, 2014.
23
Ghezelbash, F., A. Eskandari, A. Shirazi-Adl, N. Arjmand,
Z. El-Ouaaid, and A. Plamondon. Effects of motion seg-
ment simulation and joint positioning on spinal loads in
trunk musculoskeletal models. J. Biomech. 70:149–156,
2018.
24
Ghezelbash, F., A. Shirazi-Adl, N. Arjmand, Z. El-Ouaaid,
and A. Plamondon. Subject-specific biomechanics of trunk:
musculoskeletal scaling, internal loads and intradiscal
pressure estimation. Biomech. Model. Mechanobiol.
15:1699–1712, 2016.
25
Hajihosseinali, M., N. Arjmand, A. Shirazi-Adl, F.
Farahmand, and M. Ghiasi. A novel stability and kine-
matics-driven trunk biomechanical model to estimate
muscle and spinal forces. Med. Eng. Phys. 36:1296–1304,
2014.
26
Han, K.-S., A. Rohlmann, S.-J. Yang, B. S. Kim, and T.-
H. Lim. Spinal muscles can create compressive follower
loads in the lumbar spine in a neutral standing posture.
Med. Eng. Phys. 33:472–478, 2011.
27
Heuer, F., H. Schmidt, L. Claes, and H.-J. Wilke. Stepwise
reduction of functional spinal structures increase vertebral
translation and intradiscal pressure. J. Biomech. 40:795–
803, 2007.
28
Holzapfel, G. A., C. Schulze-Bauer, G. Feigl, and P.
Regitnig. Single lamellar mechanics of the human lumbar
anulus fibrosus. Biomech. Model. Mechanobiol. 3:125–140,
2005.
29
Ignasiak, D., S. J. Ferguson, and N. Arjmand. A rigid
thorax assumption affects model loading predictions at the
upper but not lower lumbar levels. J. Biomech. 49:3074–
3078, 2016.
30
Kim, K., Y. H. Kim, and S. Lee. Investigation of optimal
follower load path generated by trunk muscle coordination.
J. Biomech. 44:1614–1617, 2011.
31
Little, J. P. Finite element modelling of anular lesions in the
lumbar intervertebral disc. Brisbane: Queensland Univer-
sity of Technology, 2004.
32
Nachemson, A. Towards a better understanding of low-
back pain: a review of the mechanics of the lumbar disc.
Rheumatology 14:129–143, 1975.
33
Nachemson, A. L. Disc pressure measurements. Spine
6:93–97, 1981.
34
Nachemson, A., and G. Elfstrom. Intravital dynamic
pressure measurements in lumbar discs. Scand J Rehabil
2:1–40, 1970.
35
Nachemson, A., and J. M. Morris. In vivo measurements
of intradiscal pressure. J. Bone Joint Surg. 46:1077–1092,
1964.
36
Naserkhaki, S., N. Arjmand, A. Shirazi-Adl, F. Farah-
mand, and M. El-Rich. Effects of eight different ligament
property datasets on biomechanics of a lumbar L4–L5 fi-
nite element model. J. Biomech. 70:33–42, 2018.
37
Naserkhaki, S., and M. El-Rich. Sensitivity of lumbar spine
response to follower load and flexion moment: finite ele-
ment study. Comput. Methods Biomech. Biomed. 20:550–
557, 2017.
38
Naserkhaki, S., J. L. Jaremko, and M. El-Rich. Effects of
inter-individual lumbar spine geometry variation on load-
sharing: geometrically personalized finite element study. J.
Biomech. 49:2909–2917, 2016.
39
Natarajan, R. N., and G. B. Andersson. The influence of
lumbar disc height and cross-sectional area on the
mechanical response of the disc to physiologic loading.
Spine 24:1873, 1999.
40
Noailly, J., J. A. Planell, and D. Lacroix. On the collagen
criss-cross angles in the annuli fibrosi of lumbar spine finite
element models. Biomech. Model. Mechanobiol. 10:203–
219, 2011.
41
Noailly, J., H.-J. Wilke, J. A. Planell, and D. Lacroix. How
does the geometry affect the internal biomechanics of a
lumbar spine bi-segment finite element model? Conse-
quences on the validation process. J. Biomech. 40:2414–
2425, 2007.
42
Okushima, H. Study on hydrodynamic pressure of lumbar
intervertebral disc. Nihon geka hokan 39:45, 1970.
43
Park, W. M., K. Kim, and Y. H. Kim. Effects of degen-
erated intervertebral discs on intersegmental rotations,
intradiscal pressures, and facet joint forces of the whole
lumbar spine. Comput. Biol. 43:1234–1240, 2013.
44
Patwardhan, A. G., R. M. Havey, G. Carandang, J. Si-
monds, L. I. Voronov, A. J. Ghanayem, K. P. Meade, T.
M. Gavin, and O. Paxinos. Effect of compressive follower
preload on the flexion–extension response of the human
lumbar spine. J. Orthop. Res. 21:540–546, 2003.
45
Pearcy, M. J. Stereo radiography of lumbar spine motion.
Acta Orthop. Scand. 56:1–45, 1985.
46
Pearcy, M. J., and N. Bogduk. Instantaneous axes of
rotation of the lumbar intervertebral joints. Spine 13:1033–
1041, 1988.
47
Pearcy, M., I. Portek, and J. Shepherd. Three-dimensional
x-ray analysis of normal movement in the lumbar spine.
Spine 9:294–297, 1984.
48
Pearcy, M., and S. Tibrewal. Axial rotation and lateral
bending in the normal lumbar spine measured by three-
dimensional radiography. Spine 9:582–587, 1984.
49
Pearsall, D. J., J. G. Reid, and L. A. Livingston. Segmental
inertial parameters of the human trunk as determined from
computed tomography. Ann. Biomed. Eng. 24:198–210,
1996.
50
Polga, D. J., B. P. Beaubien, P. M. Kallemeier, K. P.
Schellhas, W. D. Lew, G. R. Buttermann, and K. B. Wood.
Measurement of in vivo intradiscal pressure in healthy
thoracic intervertebral discs. Spine 29:1320–1324, 2004.
51
Potvin, J. Use of NIOSH equation inputs to calculate
lumbosacral compression forces. Ergonomics 40:691–707,
1997.
52
Rajaee, M. A., N. Arjmand, A. Shirazi-Adl, A. Plamon-
don, and H. Schmidt. Comparative evaluation of six
quantitative lifting tools to estimate spine loads during
static activities. Appl. Ergon. 48:22–32, 2015.
53
Rajasekaran, S., N. Bajaj, V. Tubaki, R. M. Kanna, and A.
P. Shetty. ISSLS prize winner: the anatomy of failure in
lumbar disc herniation: an in vivo, multimodal, prospective
study of 181 subjects. Spine 38:1491–1500, 2013.
 KHODDAM-KHORASANI et al.
54
Rohlmann, A., U. Arntz, F. Graichen, and G. Bergmann.
Loads on an internal spinal fixation device during sitting. J.
Biomech. 34:989–993, 2001.
55
Rohlmann, A., L. Bauer, T. Zander, G. Bergmann, and H.-
J. Wilke. Determination of trunk muscle forces for flexion
and extension by using a validated finite element model of
the lumbar spine and measured in vivo data. J. Biomech.
39:981–989, 2006.
56
Rohlmann, A., G. Bergmann, and F. Graichen. Loads on
internal spinal fixators measured in different body posi-
tions. Eur. Spine J. 8:354–359, 1999.
57
Rohlmann, A., S. Neller, L. Claes, G. Bergmann, and H.-J.
Wilke. Influence of a follower load on intradiscal pressure
and intersegmental rotation of the lumbar spine. Spine
26:E557–E561, 2001.
58
Rohlmann, A., R. Petersen, V. Schwachmeyer, F. Grai-
chen, and G. Bergmann. Spinal loads during position
changes. Clin. Biomech. 27:754–758, 2012.
59
Rohlmann, A., T. Zander, F. Graichen, M. Dreischarf, and
G. Bergmann. Measured loads on a vertebral body
replacement during sitting. Spine J. 11:870–875, 2011.
60
Rohlmann, A., T. Zander, M. Rao, and G. Bergmann.
Realistic loading conditions for upper body bending. J.
Biomech. 42:884–890, 2009.
61
Sato, K., S. Kikuchi, and T. Yonezawa. In vivo intradiscal
pressure measurement in healthy individuals and in
patients with ongoing back problems. Spine 24:2468, 1999.
62
Schultz, A., G. Andersson, R. Ortengren, K. Haderspeck,
and A. Nachemson. Loads on the lumbar spine: validation
of a biomechanical analysis by measurements of intradiscal
pressures and myoelectric signals. JBJS 64:713–720, 1982.
63
Shiraz-Adl, A. Strain in fibers of a lumbar disc: analysis of
the role of lifting in producing disc prolapse. Spine 14:96–
103, 1989.
64
Shirazi-Adl, A. Biomechanics of the lumbar spine in
sagittal/lateral moments. Spine 19:2407–2414, 1994.
65
Shirazi-Adl, A. Analysis of large compression loads on
lumbar spine in flexion and in torsion using a novel
wrapping element. J. Biomech. 39:267–275, 2006.
66
Shirazi-Adl, A., A. M. Ahmed, and S. C. Shrivastava.
Mechanical response of a lumbar motion segment in axial
torque alone and combined with compression. Spine
11:914–927, 1986.
67
Shirazi-Adl, A., and M. Parnianpour. Load-bearing and
stress analysis of the human spine under a novel wrapping
compression loading. Clin. Biomech. 15:718–725, 2000.
68
Shirazi-adl, S. A., S. C. Shrivastava, and A. M. Ahmed.
Stress analysis of the lumbar disc-body unit in compression
a three-dimensional nonlinear finite element study. Spine
9:120–134, 1984.
69
Stemper, B. D., J. L. Baisden, N. Yoganandan, B. S.
Shender, and D. J. Maiman. Mechanical yield of the lum-
bar annulus: a possible contributor to instability. J. Neu-
rosurg. Spine 21:608–613, 2014.
70
Stokes, I. A., and M. Gardner-Morse. Lumbar spine
maximum efforts and muscle recruitment patterns pre-
dicted by a model with multijoint muscles and joints with
stiffness. J. Biomech. 28:173177–175186, 1995.
71
Stokes, I. A., and M. Gardner-Morse. Quantitative anat-
omy of the lumbar musculature. J. Biomech. 32:311–316,
1999.
72
Van Dieen, J., and I. Kingma. Effects of antagonistic co-
contraction on differences between electromyography
based and optimization based estimates of spinal forces.
Ergonomics 48:411–426, 2005.
73
Wade, K. R., P. A. Robertson, A. Thambyah, and N. D.
Broom. How healthy discs herniate: a biomechanical and
microstructural study investigating the combined effects of
compression rate and flexion. Spine 39:1018–1028, 2014.
74
Wade, K., P. Robertson, A. Thambyah, and N. Broom.
‘‘Surprise’’ loading in flexion increases the risk of disc
herniation due to annulus–endplate junction failure: a
mechanical and microstructural investigation. Spine
40(12):891–901, 2015.
75
Wilke, H. J., P. Neef, M. Caimi, T. Hoogland, and L. E.
Claes. New in vivo measurements of pressures in the
intervertebral disc in daily life. Spine 24:755–762, 1999.
76
Wilke, H.-J., P. Neef, B. Hinz, H. Seidel, and L. Claes.
Intradiscal pressure together with anthropometric data–a
data set for the validation of models. Clin. Biomech.
16:S111–S126, 2001.
77
Xia, Q., S. Wang, M. Kozanek, P. Passias, K. Wood, and
G. Li. In-vivo motion characteristics of lumbar vertebrae in
sagittal and transverse planes. J. Biomech. 43:1905–1909,
2010.
78
Zander, T., M. Dreischarf, A.-K. Timm, W. W. Baumann,
and H. Schmidt. Impact of material and morphological
parameters on the mechanical response of the lumbar spi-
ne—a finite element sensitivity study. J. Biomech. 53:185–
190, 2017.
79
Zhu, D., G. Gu, W. Wu, H. Gong, W. Zhu, T. Jiang, and
Z. Cao. Micro-structure and mechanical properties of
annulus fibrous of the L4–5 and L5–S1 intervertebral discs.
Clin. Biomech. 23:S74–S82, 2008.