Computer-Based Estimation of Spine Loading during Self-Contained Breathing Apparatus Carriage

WANG Shitan(王诗潭), WANG Yunyi(王云仪)

College of Fashion and Design, Donghua University, Shanghai 200051, China

Abstract: Firefighters’ low back disorders (LBDs) are closely related to excessive spine loading when using the self-contained breathing apparatus (SCBA) continuously. The purpose of this study was to quantify firefighters’ spine loading and evaluate the effects of strap lengths of SCBA on altering spine loading. Computer-based musculoskeletal models of three varying-strapped SCBA conditions and a control condition (CC) with no SCBA equipped were developed. The model was driven using three-dimensional (3D) inertial motion capture data from twelve male subjects performing a walking task and the predicted ground reaction force (GRF). Electromyography (EMG) activities were also recorded to validate the results from the model. The 4th-5th lumbar vertebra (L4/L5) joint reaction forces, as well as erector spinae and rectus abdominis forces, were finally obtained. Results demonstrated that carrying SCBA significantly increased the compressive force and anteroposterior shear force at the spine. The risk of potential LBDs increased by about 17.77%. Dynamic balance of erector spinae and rectus abdominis contraction was also disturbed when carrying SCBA, indicating a higher risk of spine muscle strain. Adjustment of SCBA strap length was an efficient method to influence spine loading. The medium-fitting strap (MS) with a length of around 101 cm generated minimum joint reaction forces and achieved the optimum dynamic balance of spine muscle contraction, which was recommended for firefighters.

Key words: firefighter; self-contained breathing apparatus (SCBA); low back disorder (LBD); musculoskeletal model

Introduction

Low back disorders (LBDs) have been reported as a common and costly problem in the fire service[1]. Continual carrying self-contained breathing apparatus (SCBA) is considered an important risk factor for the occurrence of LBDs[2].

As the heaviest item of firefighters’ personal protective equipment (PPE), SCBA (between 12 kg and 18 kg) directly generates intensive mechanical load on the spine. In addition, carrying SCBA during training and firefighting may alter firefighters’ trunk kinematics, such as more trunk rotation or forward inclination, which further imposes high compression and shear forces on the spine[3-4]. LBDs may occur when the mechanical load-bearing tolerance of the spinal column is exceeded. For instance, American National Institute for Occupational Safety and Health (NIOSH) determined a threshold of 3 400 N for peak compressive force in the 4th-5th lumbar vertebra (L4/L5) joint[5].

Although the risk of LBDs associated with SCBA carriage is generally acknowledged, the assessment of spine loading is rarely performed due to the technical challenges and ethical restrictions, and thus the risk of LBDs has not been fully examined. Mcgiletal.[6]developed a polynomial method to estimate L4/L5 joint compressive force by inputting the flexion/extension, lateral bend, and axial twist components of the net L4/L5 joint moment. However, this mathematical model only obtained joint compressive force in a specific posture. Shear forces and muscle forces could not be obtained. Several researchers developed the electromyography (EMG)-assisted musculoskeletal model to calculate compressive force, anteroposterior (A-P) and mediolateral (M-L) shear forces at the L4/L5 joint, as well as the trunk muscle activation, during simulated fireground operations[7-8]. However, these models can only analyze the human biomechanics involving body movement, which cannot investigate how joints and muscles respond when humans interact with the environment, such as carrying a bag and lifting a box.

Computer-based musculoskeletal modeling has been developed in recent years by technologies, for example, artificial intelligence[9-10]. The computational model could simulate a virtual human-environment interaction. The model could also be driven using real-time motion capture data and export the internal joint and muscle kinetic responses. In other words, the computer-based musculoskeletal model is a virtual replica of the real human bone, joint, muscle, ligament and integrated environment, namely a digital twin model. In this paper, a musculoskeletal simulator, namely AnyBody modeling system, is employed to quantify the spine loading by examining spine reaction force and muscle activity when the wearer interacts with the SCBA.

In addition, previous studies have commonly focused on reducing the weight of SCBA to alleviate its biomechanical strain, such as a lighter SCBA cylinder or redesigning SCBA systems[11-13]. However, a lighter SCBA cylinder provides limited breathing air capacity, while a low-profile SCBA is high-cost and impractical to each firefighter. Most firefighters currently carry a traditional SCBA system designed with a single-cylinder composed of carbon fiber and aluminium and carried on a frame with shoulder strap, a hip belt and a chest belt. When the SCBA weight is inevitable, variation of SCBA strap lengths to alter the weight distribution on the body may be selected as an alternative solution. Moreover, strap lengths could be conveniently adjusted by firefighters themselves, and meanwhile, it is low-cost compared to the redesign of the SCBA system.

The objective of this study is, therefore, to assess firefighters’ spine loading when carrying SCBA and the risk of potential LBDs. The effectiveness of strap length adjustment on altering spine loading is also evaluated. To quantify the spine loading, the subject-specific musculoskeletal models with no SCBAand varying-strapped SCBA conditions are developed to obtain the reaction force generated in the spine and the muscle activity around the spine region.

1 Methodology

This study had two phases. In the first phase, a three-dimensional (3D) motion capture experiment was conducted to collect body kinematics that were then used to drive the musculoskeletal model. EMG measurements were also synchronously conducted for validating the results of the musculoskeletal simulation. The second stage involved creating a musculoskeletal model to predict spine reaction forces and muscle forces based on the kinematic measurements.

1.1 Motion capture experiment

1.1.1Participants

Twelve young male students with ages of (24.4±2.0) years, heights of (174.6±2.4) cm, masses of (67.0±3.5) kg, body mass indices (BMIs) of (22.0±1.0) kg/m2, and body fat percentages of (16.5±3.4)% have participated in the experiment. These recruited subjects were the candidate of the initial firefighters and meet or exceed the China Army Basic Fitness Assessment recruitment standards[13]: (i) 17.5 ≤ BMI < 30.0; (ii) a minimum of 35 push-ups in 2 min; (iii) a minimum of 5 single-leg squat ups. All subjects were healthy, non-smokers and with no history of musculoskeletal or cardiopulmonary conditions. Before participation in this study, the recruited subjects were informed of the detailed data collection procedures and provided their written informed consent to the protocol.

1.1.2Testensembles

Four sets of test ensembles were examined (shown in Fig. 1), including a control condition (CC) without SCBA and three kinds of SCBA carrying with different strap lengths: loose-fitting strap (LS) with a strap length of 109 cm, medium-fitting strap (MS) with a strap length of 101 cm and tight-fitting strap (TS) with a strap length of 93 cm. The turnout gear, protective gloves, helmet, inner clothing, cotton socks and running shoes were also provided for each subject. The order of participations was counter-balanced to avoid any possible order effect.

Fig. 1 Four types of test ensembles: (a) CC; (b) LS; (c) MS; (d) TS

1.1.3Experimentalprotocol

The inputs of the musculoskeletal model including joint angles were collected by using the inertial motion capture systems-Xsens MVN Link (Xsens Technologies B.V., Enschede, the Netherlands), sampling at 120 Hz.

Before the dynamic trials, the subjects were instructed to wear an Xsens Lycra test suit. The test suit had 17 wireless sensors which were secured on specific anatomical locations, including the head, sternum, pelvis, upper legs, lower legs, feet, shoulders, upper arms, forearms and hands. During dynamic trials, the subjects were asked to perform a 50 m walking task at their fastest pace. To ensure consistency, all subjects were asked to begin walking with the left foot. In a randomized order, three repetitions were performed for each test sample, totaling 12 trials for each subject. The EMG data were recorded simultaneously during the kinematic data acquisition by using a Noraxon system (Noraxon, Scottsdale, Arizona, USA), sampling at 100 Hz, to ensure that the kinematic data and EMG data correspond to the exact same motion.

1.2 Model development and simulation

An AnyBody modeling system (version 7.12.2, AnyBodyTM Technology A/S, Aalborg, Denmark) was used to perform musculoskeletal analyses based on the motion capture data. The AnyBody modeling system is a musculoskeletal simulation software to investigate the mechanical functions of the living body.

The modeling process was described in detail in Ref. [14]. The motion trajectories from the motion capture experiment were filtered with a low-pass fourth-order Butterworth filter (cutoff frequency, 8 Hz). The subject-specific musculoskeletal model was then developed by modifying the generic GaitFullModel programming code. As shown in Fig. 2, the lumbar spine model in GaitFullModel contains five vertebrae with three degree of freedom (DoF) spherical joints, 188 muscle fascicles and a model of intra-abdominal pressure. Once the model is constructed, the model runs for kinematic analysis and ground reaction force (GRF) prediction by using the method in Ref. [15].

Fig. 2 Diagram of the lumbar spine model: (a) front view; (b) side view; (c) back view

Fig. 3 Process of musculoskeletal modeling: (a) motion capture experiment; (b) subject-specific model development; (c) kinematic analysis; (d) inverse dynamic analysis; (e) kinetic output

Fig. 4 Force curves at the L4/L5 joint in four test samples: (a) compressive force; (b) A-P shear force; (c) M-L shear force

Inverse dynamics approach was used to compute muscle forces and joint reaction forces over one walking cycle. During the inverse dynamics analysis, the muscle redundant problem was solved according to a third-order polynomial optimization.

Minimize function：

(1)

subject to

Cf=d,

(2)

(3)

Figure 3 depicts the process of musculoskeletal development and analysis. The musculoskeletal analysis was performed for walking with four test samples, and twelve subjects’ trials were all involved for modeling.

1.3 Data processing and analysis

Kinetic data obtained from the musculoskeletal model were processed by using Origin Pro software (version 9.0, Microcal Software, Inc., Northampton, Massachusetts, USA) to extract the peak reaction forces and peak muscle forces. The peak compressive force, A-P and M-L shear forces at the L4/L5 joint were extracted as well as the muscle force around the spine [(erector spinae (ES) and rectus abdominis (RA)]. In addition, muscle forces of trapezius (TR), rectus femoris (RF), biceps femoris (BF) and tibialis anterior (TA) were also extracted. Joint forces and muscle forces were both normalized to each subject’s body weight (BW) and reported as dimensionless parameters.

The EMG activity was also calculated for six muscles: TR, ES, RA, RF, BF and TA, for comparison with the predicted muscle forces from the musculoskeletal model. EMG recordings were selected over three cycles for every trial, and data from three cycles were averaged together to avoid any signal noise. The raw EMG signals were full-wave filtered by using an eighth-order Butterworth low pass filter with a frequency of 20-500 Hz, and then full-wave rectified and smoothed with the help of root mean square (RMS) calculation.

Statistical tests were performed by using the SPSS software (version 22, SPSS Inc., IBM, Armonk, NY, USA) with significance set atp<0.05. Quantitative kinetic data from twelve subjects were averaged, and standard deviation (SD) of the mean values was calculated for each test sample. A three-way repeated-measures analysis of variance (ANOVA) test was carried out to compare the kinetics among four test samples, followed by the Bonferroni post hoc comparisons.

2 Results

2.1 Model validation

The model validation was conducted for each subject-specific musculoskeletal model by comparing the predicted muscle forces with the measured EMG.

Pearson correlation coefficientrwas used to compare the agreement between the predicted and measured forces, categorized similarly to Taylor[16], as “weak” (r≤ 0.35), “moderate” (0.35 0.90). The root-mean-squared errors (RMSE) for predicted muscle forces and SD for measured EMG were also calculated. Table 1 presents the comparison results for the predicted muscle forces using the musculoskeletal modelversusthe measured EMG data.

Table 1 RMSE, SD and r between predicted and measured muscle forces

Predicted muscle forces at all six test muscles were significantly correlated to measured EMG (p<0.05). Strong correlation was found in the TR, RA and RF (r=0.68, 0.71 and 0.79, respectively), with RMSE lower than the experimental SD (1.44versus2.88, 1.25versus2.37, and 1.32versus2.31, respectively). ES, BF and TA showed a moderate correlation (r=0.56, 0.53 and 0.58, respectively) with a lower RMSD of 3.12, 2.60 and 1.85 than the experimental SD of 3.22, 2.72 and 1.88, respectively. In general, predicted muscle forces from the model quantitatively agreed with the measured EMG. Several deviations may be due to EMG signal delay or the incomplete synchronization of EMG collection and motion capture data collection during testing.

2.2 L4/L5 joint reaction forces

Figure 4 depicts the force-time curves at the L4/L5 joint during walking in four test samples.

The compressive force and the A-P shear force presented three peaks during a walking gait, and the maximum force occurred at about 15%, 55% and 90% phase, respectively. These three phases correspond to the initial contact, toe-off and end swing, and during these stages the body presents the maximum flexion or extension.

The peak reaction forces in four test samples are summarized in Fig. 5. Compared to the CC, walking with SCBA significantly increased L4/L5 joint compressive forces and A-P shear forces by about 58.26% and 20.59%, respectively (p<0.05).

Fig. 5 Peak reaction forces at the L4/L5 joint in four test samples

Significant main effect due to SCBA strap lengths was also found for the compressive force and the A-P shear force (p<0.05). The maximum forces were all reported for the LS among three SCBA strap lengths, while the MS showed the minimum forces. The compressive force and A-P shear force in MS were 24.08% and 14.40% lower than those in LS, respectively (p<0.05).

2.3 Spine muscle force

The force-time curves at ES and RA are shown in Fig. 6. Similar to the compressive force and the A-P shear force, the force curve of RA fluctuated three times during a walking gait, and the peak force occurred at the initial contact, toe-off and end swing, respectively. In contrast, ES only showed one peak, and the maximum peak force occurred near the end swing phase.

Fig. 6 Force curves of spine muscles in four test samples: (a) ES force; (b) RA force

Fig. 7 Peak muscle forces in four test samples

Figure 7 depicts the peak muscle force during walking in four test samples. Carrying SCBA significantly increased the muscle force for ES and RA by 20.11% and 26.70%, respectively (p<0.05). ES showed a significant difference among three SCBA strap lengths, during which the LS had a significantly higher force than MS (p<0.05). It was also observed that the force gap in BW between ES and RA increased from 0.89% in the CC to 1.00%-1.11% (p<0.05) when carrying SCBA. Similarly, the MS had the minimal force gap among three SCBA strap lengths.

3 Discussion

A computer-based modeling methodology was presented to estimate the internal loading of the spine during walking with SCBA carriage. The predicted muscle forces showed good agreement with the measured EMG data at all test muscles, indicating the good capability of the model.

Carrying SCBA significantly increased the L4/L5 joint compressive force in BW from 10.09% (706.79 N) in CC to 14.96%-16.57% (1 047.10-1 159.71 N). The compressive force increased from 20.19% to 33.00% of the maximum admissible compressive lumbar force of 3 400 N, indicating that the injury risk increased by 17.77%. Moreover, it was noted that the A-P shear force increased significantly with the addition of SCBA. The increase of the shear force components in the lumbosacral joint could be caused by the center of mass (COM) deviation of the combined trunk mass and external load. The COM deviation was associated with instabilities during the gait, resulting in an increased risk of fall accidents[17]. This finding could be further supported by the significant increase in the ES force with the adding of SCBA. The ES belongs to trunk flexor muscles with the primary function to control trunk flexion. The higher thigh flexor forces implied greater trunk forward flexion to maintain postural balance. However, the flexor muscle strength of the initial firefighter is generally weak; overstressed muscle contraction is prone to rapid muscle fatigue[18]. The contraction of flexor and extensor muscles was more unbalanced when carrying SCBA compared to the no SCBA condition, which may be associated with an increased risk of spine muscle strain. M-L shear forces in this study had no significant difference among the four test samples, indicating that it was less crucial for the ergonomic evaluation of spine loading.

Significant main effects due to strap lengths were observed for L4/L5 joint compressive forces and A-P shear forces, as well as ES forces. From these results, variation of SCBA strap lengths was an efficient strategy to alter spine loading in load-carrying activities. It could be seen that when the SCBA was carried at the LS, the peak L4/L5 compressive force was the highest (1 160 N), which was 33.13% of the threshold of 3 400 N. Similarly, the LS showed the highest exertion imbalance between the ES and RA among three SCBA strap lengths. These findings indicated the highest risk of LBDs and potential muscle strain when carrying SCBA in an LS condition. This may be due to the forward flexion of the trunk when the shoulder strap was loosened, resulting in higher mass inertia moment on the spine, thus increasing the stress on the axial and posterior aspects of the spine[19].

It was interesting to note that the variation of spine loading was not the same proportion as the increase or decrease with the strap length of SCBA. The lowest spine loading was observed when the SCBA was carried at the MS. It generated minimum L4/L5 joint reaction force in all three planes, and meanwhile, spine flexor and extensor muscles achieved the maximum balance. This may be explained by the even weight distribution of SCBA on the human back and thus might reduce mechanical irritation of the intervertebral disc. These results indicated that carrying SCBA in the MS length could be beneficial to alleviate spine loading and potential LBDs risk compared to the LS and TS lengths.

4 Conclusions

This study quantified the effects of carrying SCBA on initial firefighters’ spine loading and the risk of potential LBDs. Experimental results found that carrying SCBA generated higher compressive forces and A-P shear forces at the spine. The risk of potential LBDs increased by 17.77%. In addition, the dynamic balance of spine flexor and extensor muscles was disturbed when carrying SCBA, indicating a higher risk of the spine muscle strain.

The variation of the SCBA strap length was a feasible and convenient strategy to adjust spine loading since it significantly altered the compressive force and the A-P shear force, as well as spine flexor muscle activity. A medium-fitting strap length of around 101 cm was observed to generate the lowest spine loading, which was recommended for initial firefighters with 172-178 cm height.

Several limitations should be considered in this study. Walking was the only motion measured in this study, which might restrict the results’ application. Future research is necessary to test more complex firefighting-simulated tasks, such as climbing, rescuing, and filling hoses. Furthermore, to better predict the risk, it will be necessary to complement this study by analyzing the cumulative effect due to the time in the discs.