Abstract
Moment arms for eight pairs of torso muscles were estimated based on data obtained from 19 sets of computed tomography (CT) scans.  Muscle centroid locations of the rectus abdominis, external oblique, internal oblique, transversus abdominis, latissimus dorsi, psoas, quadratus lumborum, and the erector spinae mass were identified and digitized relative to vertebral body centers, which were also determined from the scans of the eight females and eleven males.  Muscle moment arms were then calculated as the distance between the muscle and vertebral body centroids.  The centroids for each torso muscle were plotted at each intervertebral level from T10-11 to L4-L5.  When these sections were ordered in a cephalocaudad manner, the centroid-line paths, essential to the determination of muscle force lines-of-action, could be traced.  Finally, anthropometric variables such as height, weight, torso depth, breadth, and age were regressed agains the moment arm groups means to develop several prediction equations that would determine moment arm lengths based on these anthropometric variables.

Abstract
A model employing artificial neural networks (ANNs) is developed for the prediction of lumbar muscle activity in response to steady-state static external moment loads. The model is constructed using standard feedforward networks and trained with available data using the standard backpropagation algorithm. Training with a limited set of exemplars allowed accurate prediction of muscle activity for novel moment loads (generalization). Sensitivity analyses during training and testing phases showed that the choice of specific network parameters was not critical except at extreme values of those parameters. Model predictions were better correlated with experimental data than predictions made using two optimization-based methods (average r² = 0.83 using ANNs and 0.65 using optimization). The results suggest that lumbar muscle response varies smoothly and consistently with respect to the magnitude and orientation of external moments, and they also imply an upper limit on the accuracy of muscle activity prediction using only moment loads as input. ANNs present a useful alternative to EMG- and optimization- based approaches by being both reality-based and predictive.

Abstract
This study describes the effects of varied torso muscle geometries commonly assumed in optimization-based muscle force prediction models. Specifically, the sensitivity of predicted muscle and spinal forces to assumed muscle lines-of-action (LOA) is systematically examined. The practical significance of varied muscle LOAs is addressed by determining the relative precision needed for individual muscle LOAs and assessing which muscles are most critical to accurate prediction of spinal forces. To perform this analysis a non-linear optimization model was used to generate muscle force predictions during combined frontal and sagittal plane moment loadings with an assumed erect posture. The LOAs of the Erector Spinae, Rectus Abdominus, Internal and External Oblique, and Latissimus Dorsi were systematically varied in the frontal and sagittal planes over an anatomically feasible range. The results indicated that moderate changes in the assumed LOA could substantially alter the magnitudes of predicted muscle and spinal forces. The estimated activity level of a muscle, as well as the predicted active/silent state could be affected by the LOA of that muscle and others. The patterns of predicted muscle activity, with respect to load orientation, underwent only minor alterations with changing LOA. The relative activation of several muscles, however, was dependent on LOA, and frequently led to variations in predicted spinal compression (>100N change) and shear forces (>50N change). This dependence of estimated spinal forces on assumed muscle geometry was most pronounced for the Obliques and minimal for the more vertically oriented muscles and when loads were sagittally symmetric. This study suggests that muscle LOAs are critical inputs when interpreting absolute muscle and spinal force values predicted by models of physical exertions.

Summary
A model is developed to provide a geometric representation of the human spine including thoracic and lumbar motion segments, lumbar muscles, ribs, sternum, sacrum, and pelvis. An existing model was modified in order to allow for scaling using standard anthropometric measures, deformation to specific 3-dimensional postures using surface markers, and incorporation of muscle length-tension and motion segment passive bending properties. Experiments were performed to evaluate the accuracy of model postural predictions. Analysis of surface marker displacements demonstrated that the thoracic spine deforms only minimally over a range of flexion, extension, and lateral bending torso postures, suggesting that it can be treated as essentially rigid during low-weight lifting over the range of passive flexibility. Locations of bony landmarks were accurately reproduced (mean errors 2.9-6.8 mm) as were several body dimensions (mean differences 2.6-15.4 mm). It is concluded that linear scaling to subject-specific anthropometry and the use of specific surface markers provides an accurate and direct technique for describing spinal geometry. Predicted passive spinal moments were found to be comparable to those required to support body weight in different extreme postures. It is recommended that data obtained from this type of model be incorporated in future investigations of spinal loading.

Relevance
Accurate representations of muscle and spinal geometry are essential components towards understanding the physical consequences of human exertion. Tractable experimental procedures and a valid geometric model allow for study of different sized patients in either a clinical or industrial setting. The deformation of the spine in various extreme postures can be thereby described and provide the means for more accurate biomechanical modeling of imposed tissue loads.

Abstract
Due to the complexities involved in obtaining direct measures of in vivo muscle forces, validation of predictive models of muscle activity has been difficult. An artificial neural network (ANN) model had been previously developed for the estimation of lumbar muscle activity during moderate levels of static exertions. The predictive ability of this model is evaluated in this study using several techniques, including comparison of response surfaces and composite statistical tests of values derived from model output, with multiple EMG experimental datasets. ANN-predicted activation levels were accurately modeled to within 3% across a range of experiments and levels of combined flexion/extension and lateroflexion loadings. The results indicate both a high degree of consistency in the averaged muscle activity measured in several different experiments, and substantiate the ability of the ANN model to predict generalized recruitment patterns. It also is suggested that the use of multiple comparison methods provides a better indication of model behavior and prediction accuracy than a single evaluation criterion.

Abstract
Objective: This paper examines inter-individual differences in the patterns of torso muscle recruitment during 3-dimensional static moment loading of the lumbar spine.

Design: A mathematical model (artificial neural network) was used to differentiate individual patterns of muscle response.

Background: Traditionally, experimental myoelectric data is averaged over subjects, assuming an ideal mean response to a given loading. However, averaging may overlook important information and implications associated with inter-individual variability.

Methods: In this study, a simple classification tool in the form of a competitive neural network model is developed and used to evaluate lumbar muscle recruitment patterns. Results: Subjects formed consistent and denumerable clusters, and could be categorized as either majority or minority type responders, based on their individual muscle response patterns as discerned from the output of the competitive network model. The practical significance of these differences is shown by comparison of muscle activity with more established optimization-based force predictions. Those subjects categorized as majority-type responders had muscle activity in better correspondence with optimization-based predicted forces. Subjects in minority categories displayed more variance in their response patterns and larger degrees of antagonistic co- contraction.

Conclusions: The implications for deterministic (e.g. optimization-based methods) biomechanical modeling are discussed. It is speculated that inter-individual muscle recruitment differences may be important for assessing individual musculoskeletal risk.

Relevance
Substantial muscle forces, and subsequent spinal forces, are developed during a range of physical exertions. However, some individuals will become injured while others will remain healthy despite performing the same general tasks. Differences in muscle recruitment might be an important source of this differential by causing different internal loads for identical external conditions. Identification of at-risk individuals may be facilitated by detailed examination of muscle response patterns.

Abstract
An artificial neural network (ANN) was created to simulate lumbar muscle response to static moment loads. The network model was based on an abstract representation of a motor control system in which muscle activity is driven primarily to maintain moment equilibrium. The network model parameters were obtained by an iterative method (trained), using a modification of the standard backpropagation algorithm and moment equilibrium constraints. In contrast to previous ANN models of muscle activity, patterns of muscle activity are not target (training) values, but rather emerge as a result of moment equilibrium constraints. Assumptions regarding the moment generating capacity of muscles and competitive interactions between muscles were employed and enabled the prediction of realistic patterns of muscle activity upon comparison with experimental electromyographic (EMG) datasets (r²: 0.4-0.9). The success of the simulation model suggests that a motor recruitment plan can be mimicked with relatively simple systems and that 'competition' between responsive units (muscles) may be intrinsic to the learning process. Prediction of alternative recruitment patterns and differing magnitudes of co- contractile activity were achieved by varying competition parameters within and between units.

Abstract
The use of electromyographic measures, in concert with modeled or empirical representations of muscle physiology, is a common approach for estimation of muscle force.   Existing models of the lumbar musculature have allowed model parameters to vary for an individual subject.  While this approach improves apparent predictive ability, it loses some degree of construct validity since parameter variability may not be physiologically justifiable.  An EMG-based 5-parameter model, adapted and generalized from earlier reports, is presented here.  Inherent in the model is the requirement of subject-invariant modeling parameters.  As a practical analysis tool was desired, the model relies on relatively few calibration constants whose determination is described.  Empirical evaluation was undertaken using a database of 398 experimental trials involving lifting and transferring objects of moderate mass.  Model performance, evaluated by comparison of measured and predicted lumbar moments, was comparable to earlier models, with r ² mean (sd) values of 0.76(0.15) for sagittal plane moments, and rms mean (sd) errors of 14.1(7.4), 9.7(5.3), and 8.6(3.6) Nm in the sagittal, frontal, and horizontal planes respectively.  These empirical results and the argument of physiological veracity support the use of a subject-invariant model.

Abstract
This study examines the potential effect of short-term practice on low-back stresses during manual lifting and lowering of a 15 kg load, and while using two different types of materials handling devices (MHDs) to lift and lower a 40 Kg load.  The two MHDs used were an articulated balance arm and a pneumatic hoist.  The expectation was that low-back dynamic moments, EMG measured torso muscle antagonism, and EMG predicted L4/L5 disc compression forces would rapidly decrease with practice, and that the manual lift-lower activities would be learned faster than the MHD assisted exertions.  Four naive male college age subjects performed 40 lift and lower exertions, both manually and with the two MHDs for a total of 24 experiments.  Non-linear regressions of the peak and average low back moments, EMGs and disc compression values revealed only small decreases in the values (from 2% to 14%) over the 40 trials, and it was only statistically significant for five of the 48 regressions.  This would seem to indicate that if learning is present in these tasks it is going to be very slow learning, and thus future studies will need to include a much larger number of trials.  The effects of MHDs on the learning rates when compared to manual lifting learning rates was not statistically significant.  It was shown, however, that MHDs had a particularly beneficial effect on reducing L4/L5 compression forces during load lowering activities despite the MHD load being much heavier than the manual load.  It was also found that the level of torso muscle co-contraction increased significantly ( 2 to 4 times) when MHD handling was involved compared to manual lifting and lowering.

Abstract
Common manipulator-assisted materials handling tasks were performed in a laboratory simulation at self-selected and faster (paced) speeds.  The effects of pacing on peak hand forces, torso kinematics, spine moments and forces, and muscle antagonism were determined, along with any influences of several task variables on these effects.  The faster trials were performed 20% more rapidly than the self-paced trials.  It was found that (a) achieving this level of performance required ~10% higher hand forces and 5-10% higher torso moments, (b) consistent torso postures and motions were used for both speed conditions, and (c) the faster trials resulted in ~10% higher spine forces and ~15% higher levels of lumbar muscle antagonism.  On whole, these results suggest a higher risk of musculoskeletal injury associated with performance of object transfers at faster than self-selected speeds with and without a manipulator.  Further analysis provided evidence that the use of manipulators involves higher levels of motor coordination than do manual tasks.  Several implications regarding the use of material handling manipulators in paced operations are discussed.  Results from this investigation can be used in the design, evaluation, and selection of material handling manipulators.

---

11. Nussbaum, M.A., Chaffin, D.B. and Baker, G. (1999) Biomechanical analysis of materials handling manipulators in short distance transfers of moderate mass objects: joint strength, spine forces and muscular antagonism. Ergonomics, 42, 1597-1618.

Abstract
Although often suggested as a control measure to alleviate musculoskeletal stresses, the use of mechanical assistance devices (i.e. manipulators) in load transfers has not been extensively studied.  Without data describing the biomechanical effects of such devices, justification for decisions regarding implementation of such tools is difficult.  An experimental study of two types of mechanical manipulators (articulated arm and overhead hoist) was conducted to determine whether biomechanical stresses, and hence injury risk, would be alleviated.  Short distance transfers of loads with moderate mass were performed both manually and with manipulator assistance under a variety of task conditions.  Using analysis and output from new dynamic torso models, strength demands at the shoulders and low back, lumbar spine forces, and lumbar muscle antagonism were determined.  Strength requirements decreased significantly at both the shoulders and low back when using either manipulator in comparison with similar transfers performed manually.  Peak spine compression and a-p shear forces were reduced by about 40% on average, and these reductions are shown to be primarily caused by decreases in hand forces and resultant spinal moments.  Two metrics of muscular antagonism were defined, and analysis showed that torso muscle antagonism was largest overall when using the hoist.  The results overall suggest that hoist-assisted transfers, though best in reducing spine compression forces, may impose relatively higher demands on coordination and/or stability especially at extreme heights or with torso twisting motions.  The relatively higher strength requirements and spine compression associated with the articulated arm may be a result of the high inertia of the system.  Potential benefits of practice and training are discussed, and conclusions regarding implementation of mechanical manipulators are given.

Abstract
The risk of musculoskeletal injury associated with manual materials handling tasks has led in part to the use of material handling manipulators, yet there is limited empirical data to facilitate selection, design, and evaluation of these devices.  A laboratory study of two types of mechanical manipulators (articulated arm and overhead hoist) was conducted of short distance transfers of moderate loads, and the influence of various task parameters and transfer method on motion times, peak hand forces, and torso kinematics were obtained.  Use of manipulators increased elemental motion times for symmetric sagittal plane transfers by 36 - 63%, and asymmetric transfers (in the frontal plane) by 62 - 115%, compared to similar transfers performed manually.  Peak hand forces were significantly lower with both manipulators (40-50%), and approximately 10% higher for asymmetric versus symmetric transfers.  Overall torso kinematics were grossly similar with and without a manipulator.  These results suggest that for self-paced job tasks, moderate mass objects will be transferred slower over short distances and with lower levels of external (hand) forces when using mechanical aids.  These simple effects, however, were influenced by object mass and transfer height.

Abstract
The description of a lifting strategy is typically provided in qualitative terms.  A quantitative static descriptor or index differentiates the starting postures but not the primary moving segments.  This technical note proposes an index that quantitatively characterizes different dynamic postural strategies employed during sagittal plane lifting.  Dynamic lifting strategies are modeled in the velocity domain as different schemes of partitioning postural changes between the torso and leg segments.  The index consists of two parameters, assigned to two leg segments, which quantify their contributions relative to the torso.  Given a measured lifting movement, its index parameter values, ranging from 0.1 to 10, are estimated through an enumeration search process with the objective of minimizing the fitting error.  The use of this index is illustrated by applying it to 24 lifting movements performed by six subjects assuming either a back-lift or a leg-lift strategy.  Results indicate that a lifting strategy, in terms of whether the leg or the back is generally the prime mover, can be differentiated and visualized using this simple two-parameter index.  In addition, indistinct intermediate strategies are also discerned, as the involvement of each segment in a lifting movement is quantified.  The index is however limited in that it does not accommodate arm motion contributions to a lift nor possible time-dependent strategic changes during a lift.  Potential future applications include time-efficient movement prediction and simulation for computerized biomechanical or ergonomic analysis.

This article presents a set of heuristic algorithms (computational procedures) for determining upper extremity joint centers from a reduced set of surface markers (as compared to the conventional scenario of three markers per segment) during three-dimensional (3-D) motions.  Two experiments were conducted to empirically derive and evaluate the algorithms.  The first experiment characterized the geometric relations between the instantaneous helical axes and corresponding surface markers at the shoulder and elbow.  These relations, along with anatomic data available in the literature, were employed to derive the geometric heuristics.  The heuristics were then enhanced through a solidification optimization procedure with the objective function of minimizing segment length variability over time.  This latter variability was also used as the error measure in evaluation of the algorithms.  The second experiment, incorporating a wide range of upper extremity motions in 3-D, showed the following: (1) the geometric heuristics alone yielded an average error of approximately 7.5 mm; (2) with optimization-based enhancement, the average error was reduced to 3.7 mm; (3) further improvement of the algorithm performance is achievable by modifying the parameter setting of the optimization procedure.  While more extensive evaluation is needed before the proposed approach can be generalized, the current work demonstrates the viability of heuristic algorithms for estimating upper extremity joint centers based on a reduced set of surface markers and in an efficient manner.

Abstract
It is commonly recommended that those who must perform strenuous physical exertions be given training in proper techniques, and that musculoskeletal risks are thereby minimized.  In the present study, the effects of training were examined at a behavioral level, and it was assumed that any long-term reduction in injuries must be preceded by measurable changes in how exertions are performed.  Because of high injury incidence rates among nurses, common patient handling tasks were the basis of the study.  Participants performed several such tasks and sub-groups received training that consisted of either a commercial video or combined lecture and practice sessions.  Compared to a control group, several perceptual, postural, and biomechanical measures were significantly altered following training.  Specifically, training was associated with the adoption of a more upright lifting posture, and this change was retained in follow-up measures obtained after 4-6 weeks.  While substantial inter-subject variability was present, the results suggest that training can modify behaviors in an intended direction.  This observation provides support for the use of training as potential control measure, however more information is needed regarding retention and the relationship between trained behaviors and long-term injury risk.

Abstract
Despite extensive research on muscular fatigue during prolonged static efforts, there have been relatively few studies of more complex tasks (dynamic and intermittent).  A laboratory study of overhead work tasks was conducted to investigate whether electromyographic (EMG) measures can potentially serve as indicators of fatigue, particularly for ergonomic tasks analysis.  Sixteen participants performed the tasks until they either developed substantial discomfort or reached a three-hour limit.  EMG signals were obtained at intervals throughout the experiment from four shoulder muscles, both statically (during fixed-level test contractions) and dynamically (during task performance).  Both EMG RMS amplitude and spectral content (mean and median power frequencies) were examined and compared in terms of their variability and sensitivity.  In addition, a new fatigue index was developed to allow for estimation of substantial fatigue onset.  Variability was found to differ significantly between muscles and EMG measures, and was generally lowest for mean power frequencies obtained during static test contractions.  Sensitivity was typically greatest for RMS versus spectral measures, and slightly higher for median than mean power frequencies.  The results suggest that fatigue during dynamic tasks, while a complex phenomenon, can be monitored and quantified using EMG.

Abstract
Shoulder problems are prevalent in industrial work, particularly when tasks require the hands be used at or above shoulder level.  While extensive research has been conducted on prolonged static exertions, and several guidelines for such efforts exist, there is insufficient information for ergonomic evaluation of tasks that are intermittent and/or dynamic.  A laboratory simulation was conducted of overhead assembly work that was both intermittent and dynamic, and which varied the duty cycle (work/rest ratio), arm reach, and hand orientation of a tapping task.  Results consisted of endurance times and also the times of fatigue onset as indicated by perceived discomfort and declines in muscle strength.  Females exhibited longer (22%) endurance times, delayed reports of discomfort, and slower declines in strength.  Significant influences of duty cycle were found on both endurance and fatigue times, yet arm reach and hand orientation did not have consistent effects.  Distributions of endurance and fatigue times are presented as criteria for preliminary evaluation of overhead work.  Endurance times could be predicted with only moderate accuracy from earlier indicators of fatigue onset.  Existing guidelines, albeit developed for static tasks, appeared unsuitable for the simulated overhead assembly efforts examined.  Furthermore, such guidelines may fail to capture the substantial inter-individual variability that was observed in the experiment.