Pilot Validation of a Computational Seated Human ModelUsing an Instrumented Chair
C.G. Olesen1, M. de Zee1 and J. Rasmussen2
1Department of Health Science and Technology, Aalborg University, Denmark
2Department of Mechanical Engineering, Aalborg University, Denmark
e-mail:
Abstract
Sitting-acquired deep tissue injuries (SADTI) are the most serious type of pressure ulcers. In order to investigate the ethiology of SADTI a new approach is under development. A musculo-skeletal model can predict forces at different seated positions and thereby find an optimum position where the risk of SADTI is minimized. This study focuses on validation of a musculo-skeletal model developed in the AnyBody Modeling System for predicting reaction forces on a wheelchair. A Wheelchair with force-measuring equipment was developed, an experiment was conducted, and the experimental results were compared with the predictions of the computational model. The results show that the model predicted changes in reaction forces due to changes in the chair posture well. It was concluded that the experimental approach is satisfactory and more experiments should be carried out to completely validate the computational model.
1
1.INTRODUCTION
Pressure ulcers, more commonly known as pressure sores, are a frequent complication to spinal cord injury (SCI) patients. The disease can have various ethiologies and is in general poorly understood. The type of pressure sore we are particularly interested in is the sitting acquired deep tissue injury (SADTI) that wheel chair users, i.e. paraplegic and quadriplegic patients are susceptible to.
Statistics show that 24 % of all patients with SCI experience a pressure ulcer during their rehabilitation hospital stay [1]. It is also estimated that 50-85% of all patients with SCI will experience a pressure ulcer during their life time [2]. These are very general prevalence’s that cover all kinds of pressure ulcers; the deep tissue injury is not the most common one, but certainly the worst. The problem is that they can spread underneath the skin and when they do so, they can be very difficult to treat and in some cases they are lethal. [3].
The ethiology behind pressure ulcers in general and specifically SADTI are poorly understood. A number of main risk factors have been identified, but to which degree these risk factors interact and how they cause pressure sores for wheel chair users should be investigated in greater detail.
It is well acknowledged that pressure sores are primarily caused by sustained mechanical loading of the soft tissues [4, 5]. The types of loading can be described as pressure, pressure gradients and shear forces [6]. These loads conspire in a complicated fashion to generate stress states varying from point to point in the soft tissues. An understanding of the input loads is therefore the first step towards a genuine understanding of pressure sore formation.
The magnitude and position of the loads is influenced by the patient’s posture and the support conditions of the wheelchair, for instance what cushion type is used, backrest angle, seat inclination angle, foot rest height, arm rest height and so on [6, 7].
Figure 1. The experimental setup with the wheelchair mounted with the force measuring equipment. The bright dots are the reflective markers.
A lot of the risk factors have been experimentally investigated in combination, but it is difficult to compare the results due to the variations between subjects and the differences between the protocols used. For instance, when Maurer and Sprigle (2004)[8] studied how the seat inclination angle and backrest angle affect the pressure in the buttock area, they found no significant differences in pressure. However, they did not look at the shear forces or the pressure gradients and according to [9] when shear is present it takes six times less pressure to cause ulcerations. Others have shown similar results on how pressure and shear causes blood occlusions in the superficial tissue [10-12]. They all showed that pressure can cause occlusion in the blood flow, but if shear is present it takes less pressure to lower the blood flow. So shear forces are important and perhaps even more so when looking at how paraplegic patients sit. Bennett et al. (1984)[13] saw higher shear stresses in paraplegic patients than normal subjects during sitting.
There are studies that describe experiments where subjects have been seated on chairs and wheelchairs with load cells and pressure mats. For instance, Gilsdorf et al. (1990)[14] performed an experiment where a force plate was used as a seat in a wheel chair, and normal and shear forces were measured for normal subjects while changing the backrest angle. Their results are difficult to compare with for example another result also done by Gilsdorf et al. (1991)[7] on paraplegic and quadriplegic patients, because the latter study unlike the former used a wheelchair equipped with armrests.
Figure 2. Recording and post processing of experimental data using a Matlab script, transfer of the motion data to the AnyBody model for simulation of forces and finally comparison between measured and simulated forces.
In some experiments conducted by Goossens et al. (1997)[15] on healthy subjects the maximum shear was found in areas where the pressure also was highest but no relation between seat angles and shear stress could be found, possibly because of the influence of body posture.
Bennett et al. (1984)[13] developed a shear stress sensor that also included pressure and blood flow but it was developed as a hard sensor which was not suitable for measuring in a soft cushion environment. Measurements were therefore taken on subjects sitting on wood surfaces. This study did see larger shear stresses in paraplegic and hospitalized geriatric patients than in normal subjects and they measure larger blood flow in the two latter groups. This indicates that paraplegic patients have a higher than normal need to relieve pressure from the buttock area.
All the aforementioned experiments contribute to the understanding of forces between a seated human and its environment, but differences in aims and experimental protocols make them impossible to compare. Consequently this study proposes that a more analytical approach could contribute further to understanding of how the seating posture affects the mechanical loading of the soft tissue in the buttock region. Such an understanding might also benefit designers of automotive, airline and office seats because of contribution of contact forces’ to the perceived discomfort.
A validated analytical model could be used to predict load values from different seating postures without the need for costly experiments. Furthermore, an analytical model is free from the inevitable experimental noise and statistical variations due to individual differences between subjects and differences in protocols that may otherwise occlude significant findings.
Before using any model it should be validated with respect to the predicted forces [16]. Therefore the objectives of this study is to validate a musculo-skeletal model with respect to its ability to predict reaction forces and how these change with a change in the seated posture. This is accomplished by comparing force predictions from a model with measurements from test subjects.
2. METHODS AND MATERIALS
The study design includes an experimental and a modeling part. The experimental setup includes measurements of reaction forces on a wheelchair measured in synchronization with motion capture data for recording of the sitting posture. A musculo-skeletal model was created to mimic the seated position and thereby estimate reaction forces between the chair and the model. These reaction forces could then be compared with the forces measured during the experiment and in that way validate the musculo-skeletal model.
Experimental Setup
Force measurements
A custom-built wheelchair, see Figure 1 (Wolturnus A/S, Nibe, Denmark) was mounted with force-measuring equipment (Advanced Mechanical Technology, Inc., Watertown, MA, US). The chair was constructed by mounting an OR6-7-1000 force plate as a seat, and as a footrest an OR6-7-2000 force plate mounted in the floor was used. The backrest consisted of two horizontal bars mounted in each side to two multi-axis force and torque transducers. In total, four multi-axis force transducers (2 x FS6-250 & 2 x FS6-500) were mounted in the two backrest bars. The backrest bars were covered with 4 mm foam and the seat was covered with a 2 mm rubber mat.
The wheelchair could be adjusted in a number of ways, the inclination angle of the seat and backrest. The seat could slide forward/backward. The backrest bars could slide up/down and forward/backwards. These adjustments meant that the wheelchair could be adjusted into any posture with respect to backrest height, seat dept, seat angle and backrest angle.
The forces applied to the force plates and force transducers were amplified using amplifiers from the same manufacturer as the force measuring equipment. The amplifiers were sampled using Texas Instrument 16 bit A/D converters, connected to a computer where sampled by Mr.Kick, a software for data collecting (Mr.Kick, Aalborg University, Aalborg, Denmark). Details about Mr.Kick can be found at [17]. All six channels on each force-measuring device were sampled giving a total of 36 channels. The sampling rate was 20 Hz over 10 seconds giving 200 samples per channel.
Posture Assessment
It is partially unknown how the chair adjustments influence the posture of the occupant. Therefore the posture of the subject had to be measured.
The chair positions were measured using a motion capture system (Qualisys Proreflex 240, Gothenburg, Sweden) with eight cameras. Passive reflective markers were placed in each corner of the two force plates and on both sides of the two backrest bars seen in Figure 1
The posture of the subject sitting in the wheelchair was also measured using passive reflective markers. The markers were placed at the following anatomical landmarks: Forehead, sternum, shoulder, elbow, hand, pelvis, greater trochanter, knee and ankles, which also can be seen in Figure 1. The positions of the markers were measured with a sampling rate of 20 Hz over 10 seconds.
The force measurements and the motion capture system were started by a trigger in order to synchronize the force and motion capture recordings.
Conducting the Experiment
One subject was included in the study. The subject was 179 cm high and weighed 68 kg.
The subject height, mass and distance from floor to rear side of the knee while seated were measured. The latter distance was used as a guide for setting of an appropriate seat height, i.e. the distance from the footrest force plate to the front top edge of the seat. This distance was kept constant throughout all experiments. The subject was seated in the chair and markers were placed at the anatomical landmarks. The subject was instructed to sit in as relaxed as possible with the arms crossed over the chest. The force plates and load cells were reset every time the posture of the chair was changed to make sure that the baseline was the same for all measurements. The experiment included 14 different seated postures each measured at least twice to be sure that at least one measurement was ok. The 14 different postures varied in seat angle, backrest height and backrest angle. Between each of the postures only one parameter was changed and the subject sat at the chair 4 min between each experiment, to ensure that the seated position had reached a steady state as described by Crawford, S.A. et. al. (2005) [18]. It was realized during the post processing that not all the 14 experiments were of any particular interest for the study because they included repetitions. Consequently only 11 of them were used for validation purposes.
Post Processing
Motion Capture
The motion capture data from the Qualisys cameras was first post processed in QTM (Qualisys Tracking Manager V.1.10.282) were the markers were identified and named. The markers were then exported to a TSV file (Tab Separated Values).
Table 1 shows the most important in- and output parameters to the AnyBody model
Input Parameters / Output ParametersSeat Angle / Backrest normal force
Backrest Angle / Seat normal force
Height of Backrest / Seat shear force
Distance between backrest bars / Footrest normal force
Subject posture / Footrest shear force
The TSV file was then processed using a custom made Matlab (MatLab R2007B, Mathworks Inc., Ma, US.) script that calculated the angles and positions between the seat, backrest bars, and footrest. A relative origin was also calculated as the where an extension to the line between the two backrests in the right side intersected with the seat. The markers placed at the bony landmarks at the subject were calculated relatively to the calculated origin. The output of the Matlab scripts was a text file that could be included in the AnyBody model. The text file consisted of all the marker positions for driving the model. For an overview of the process, see Figure 2.
Force Data
The collected force data was gathered using Mr. Kick which saves each data file as a Matlab data file. A script was used for averaging the values. A Matlab script combined the motion capture data with the measured forces, and the center of mass (COM) for the subject was calculated based on the normal and shear forces from the force plates and load cells.
AnyBody Modeling
The modeling was done using the AnyBody modeling system (AnyBody Technology A/S, Aalborg, Denmark). The model used was the “Seated Human” from the public domain AnyBody model repository [19], described in [20]. The AnyBody Modeling System is computer software designed for constructing musculo-skeletal models of the human body and its environment and for determining how interact. The mathematic and mechanical theories behind were described in [21].
Figure 3 shows to the left the anybody model in the seated position driven by the data recorded with the Qualisys motion capture system. To the right a close up of the backrest support, where shown a posture only supported by the upper backrest. The backrests are the two squares. The lines originating from the spine indicates the normal force.
The AnyBody model was altered to fit the experiment. First the model was scaled in weight to the body mass if the test subject, i.e. 68 kg. The subject was approximately the same anthropometrical size as the AnyBody model, which is based on a 50th percentile male.
The model consists of two parts: A human model and an environment model representing the wheelchair. The seated model relies on a set of assumptions described in [20]. In addition, a few assumptions were made about the interface between the chair and the human body:
Forces / 1.exp / 2.exp / 3.exp / Mean / StdBT(Fs) / 5,4 / 4,5 / 5,6 / 5,2 / 0,59
BT(Fn) / 18,2 / 21,3 / 22,5 / 20,7 / 2,2
BB(Fs) / 4,6 / 2,7 / 1,1 / 2,8 / 1,7
BB(Fn) / 21,0 / 15,0 / 13,2 / 16,4 / 4,1
S(Fn) / 585,8 / 590,0 / 585,1 / 587,0 / 2,7
S(Fs) / 11,7 / 10,1 / 9,0 / 10,3 / 1,4
FR(Fn) / 88,1 / 86,0 / 91,7 / 88,6 / 2,9
FR(Fs) / 20,7 / 16,0 / 17,3 / 18,0 / 2,4
Table 2 shows an experiment that was conducted three times. The forces for each of the three measuments and the mean and standard deviation. This illustrates a small variation. BT: backrest top, BB: backrest bottom, S: seat, FR: footrest.
The friction coefficients µ, between the body and the seat and footrest was estimated to µ=0.5 because the surface material was rubber. For the backrest µ=0 was used because the foam could rotate around the backrest bars.
The model is illustrated in Figure 3 where the model sits in the chair in the same position as in Figure 1. The armrest, legrest, and head rest were removed from the original model since they were not present in the wheelchair used in the experiment. The backrest was changed to be two independent backrest bars as it was in the experiment. The size and position of the seat, backrest bars and footrest were set and driven by the input file generated by the Matlab scripts.
The human model was driven by the markers recorded during the experiment such that the posture of the human body in the model corresponds to the experiment.
The connection between the chair and the human model was modeled as contact elements perpendicular to the contact surfaces and capable of taking Coulomb friction into account.
Comparison
The comparison between the experimentally obtained forces and the estimated forces from the AnyBody model was done by comparing absolute force values and trends when changing an input parameter. The force values were plotted as a function of the parameter that was changed.
The parameters that were used for the comparison could be split into input and output parameters. The input and output parameters are listed in Table 1.
3.RESULTS
The results from forces measured during the experiment and estimated forces from the AnyBody model were compared as absolute values and trends while changing one parameter at the time.
Figure 4. Weight of the subject in 14 experiments measured as the sum of vertical forces on the dynamometers.
The first result calculated was the absolute weight of model and subject for each of the 14 experiments conducted. The result is depicted around the weight of the AnyBody model which is illustrated as a straight line due to the constant weight of 681 N. The subject weight was calculated from the reaction forces on the chair, and the mean was also 681 N with a SD of 8 N.