Introduction

Fatal and injurious automotive collisions occur at a rate of 1.7 million per year in the United States, and accounted for nearly 33,000 deaths in 2014.17 Among such collisions, frontal crashes are the most prevalent, accounting for over 50%.17 Brumbelow and Zuby2 reported that chest injuries are the most common injuries observed among those seriously injured in frontal crashes of vehicles with good safety ratings. During motor vehicle collisions, chest deflection can produce rib fractures in vehicle occupants, thus chest deflection can be used to predict thoracic injury.3,13,14,22

Several different values for chest compression ([deflection/chest depth] × 100%) injury threshold in frontal and side impacts have been reported, including 30% ,20,25 34%,13 35% (half-chest compression),24 37%,15 and 38%.27 In a series of studies, Kemper et al. 10,11 used strain gages to identify fracture timing, and found that fractures had occurred at chest compressions as low as 16%, 12%, and even 6%. The current injury assessment reference value (IARV) in use for chest deflection in the Hybrid III 50th percentile male anthropomorphic test device (ATD) is 63 mm, or 29% compression.7 Because of the utility of chest deflection data, it is desirable in automotive testing to obtain such data in order to predict injuries and evaluate vehicle safety.

Historically, chest deflection has been observed on PMHS through various methods including surface accelerometers, video capture of surface targets, transthoracic rods.3,18 Concerns about accuracy in these methods, and particularly about the transthoracic rod’s effects on thoracic response and injuries, motivated the development of the chestband, which is now commonly used to acquire chest deflection.4–6,20,21,23

A chestband, also called an External Peripheral Instrument for Deformation Measurement (EPIDM), is a thin strip of steel with a series of strain gages attached to it at known locations, all encased in polyurethane.6 The gage readings provide a measure of the curvature of the chestband at each gage location, and can be used to digitally reconstruct the chestband contour. Chestbands are commonly used in biomechanics as an instrument which is wrapped around the thorax to provide thoracic contour data throughout impact, from which chest deflection may be calculated.4,5,20,23

Several studies have validated the accuracy of chestband contours reconstructed from strain gage measurements.1,4,6,19 Taking validation a step further, measurement errors were specifically examined and quantified in the study by Bass et al.1 Although it has been verified that the measurements are accurate, there has not yet been a study which has examined whether, and to what extent, the use of chestbands may alter the actual response of the thorax to impact. A few possible sources of chestband effects are added stiffness from the chestband itself, initial compression of the thorax from wrapped chestbands increasing the observed stiffness, and chestband interactions with the individual ribs inducing stress concentrations and altering strain patterns. Should such effects exist, the validity and value of the collected data could be compromised. Thus, there exists a concern that by wrapping a chestband, or especially multiple chestbands, around the thorax of a post-mortem human surrogate (PMHS), the thorax characteristics and impact response may be altered. Of particular concern are the effects on small, frail subjects, as their small size and low bone strength could amplify the amount of relative compression induced by chestband placement, thus also amplifying potential chestband effects. The purpose of the present work is to determine the effects of chestbands on both the global and local responses of the human thorax to frontal impact.

Materials and Methods

Chestband effects on global thoracic response were investigated through a series of 22 low-energy, non-injurious, frontal ram impacts to two PMHS. The number of chestbands present on the subject was varied throughout testing in order to provide a comparison of response with and without chestbands. Speed was also varied in order to have multiple points for comparison. Chest deflection and thoracic stiffness were used to represent the global thoracic response, while rib strain was used to characterize the local response.

Subject Selection

Two PMHS were used in the study, the first being a small, frail (osteopenic) female, and the second being a mid-size male. The PMHS were made available through The Ohio State University’s Body Donor Program and all applicable Body Donor Program and University guidelines were reviewed and followed. The small female was selected in order to evaluate chestband effects under conditions in which such effects should be most pronounced. The mid-size male was selected to represent the size of subjects most commonly used historically in biomechanics studies.8 The small female, subject 1, was 83 years old, 162 cm in height and 59 kg in mass. Areal bone mineral density (aBMD) was obtained using dual-energy x-ray absorptiometry and the dual femoral neck T-score was calculated as −1.6, which indicates osteopenia.9 The male, subject 2, was 64 years old, 178 cm tall, with a mass of 79 kg and a lumbar spine aBMD T-score of −0.6.

After the BMD was checked, each of the two subjects had a pre-testing CT scan performed to screen for pre-existing injuries and thoracic abnormalities, and no problems were found. Each subject’s anthropometry measurements were collected prior to instrumentation, and an anthropometry table is included in the supplementary material, Table S1. Additionally, the breasts were removed from the female in order to eliminate their influence on thoracic response and chestband effects.

Test Setup

Frontal impacts were conducted on each PMHS using a 23 kg pneumatic impactor, with a flat plate impactor surface (6″H × 12″W) centered at mid-sternum (Fig. 1). The PMHS was seated against a 90°, flat back fixture, thus impeding spinal motion. The rigid-back setup enabled the ram displacement, from initial contact to peak displacement, to be used as a measure of chest deflection. The PMHS was supported, and the seated posture maintained by use of a head harness which was placed on a linear track, allowing for anterior–posterior motion of the head during impact. The PMHS’ arms and forearms were flexed 90°, which prevented impactor interaction without raising the thoracic region.

Figure 1
figure 1

(a) Test setup including a 23 kg impactor ram, 90° back fixture, and PMHS with arms and forearms flexed at 90°. (b) Position of the 2 chestbands on the PMHS.

Full size image

Instrumentation

The instrumentation used was nearly identical for both subjects, the only difference being that subject 2 had additional strain gages on the ribs. For both subjects, two chestbands were wrapped around the PMHS’ thorax; the superior chestband at the level of the axilla (40 gage, Humanetics Innovative Solutions, Plymouth MI), corresponding to the sternal end of the third level ribs, and the inferior chestband at the level of the xiphoid process (59 gage, Humanetics Innovative Solutions, Plymouth MI), corresponding to the sternal end of the sixth level ribs. Chestbands were wrapped at 45 N (10 lbs) of tension as measured by a handheld hook scale, secured using Velcro at the two ends, and two sutures were placed in the back near the spine to maintain chestband positioning throughout all tests. A 6 df motion block containing 3 accelerometers and 3 angular rate sensors (6DX Pro Sensor, Model 2000 g 18 K°/s, DTS, Seal Beach CA) was placed on the 4th thoracic vertebra (T4) to verify there was no spinal motion, and to provide a correction in the event of spinal motion occurrence. Strain gages (Model CEA-13-062UW-350/P2, Micro Measurements, Wendell NC) were placed anterolaterally (approximately 60% of the curve length, the costotransverse junction [CTJ] being 0% and the costocondral junction [CCJ] being 100%) on ribs 3–8 bilaterally, in order to quantify lateral rib strain during each impact and to identify fracture timing. In the case of subject 2, up to three strain gages were placed on each rib 2–8. Ribs 3–8 had gages placed at 30% and 60% of the curve length. At the 90% (of curve length) location, gages were placed on ribs 2–7 in order to be centered under the chestbands. Instrumentation placement on the thorax is shown in Fig. 2.

Figure 2
figure 2

Instrumentation placed on each PMHS thorax prior to impact testing.

Full size image

The test setup used a linear potentiometer (Model CLWG-600-MC4, Celesco, Chatsworth CA) on the impactor, which served as the consistent measure of chest deflection throughout testing. Other instrumentation included an accelerometer (Model 7264C-2K, Endevco, Irvine CA) on the impactor to acquire impact velocity, as well as a load cell (Model 2944JFL, Humanetics Innovative Solutions, Plymouth MI) between the ram shaft and the impactor face.

Test Matrix

The combined set of 22 impacts included variations both in the impact speed and the number of chestbands used. A baseline case was defined as 0.8 m/s impact velocity with 0 chestbands present. The baseline scenario was tested first, repeated twice to check repeatability, then also repeated intermittently during the testing process. Deviations from the baseline scenario were systematically applied, including adjustments to impact velocity and to the number of chestbands. Impact velocities were approximately 0.8, 1.0, 1.5, and 2.0 m/s. The number of chestbands was 0, 1, or 2. As chestbands were placed, removed, and replaced throughout testing, marks were made on the subject to identify proper placement and both the chestband placement and subject posture were verified visually for consistency using a lateral-view camera on a tripod.

The general approach to testing with each subject was to begin with a low-energy impact, and gradually increase the impact energy until damage to any of the ribs occurred. Strain gage data were reviewed after each impact to identify any fractures. Once rib damage was observed, testing was ended because the system being tested (the thorax) had changed. The order in which the impacts were conducted is portrayed in Fig. 3.

Figure 3
figure 3

Flow of testing for the two test subjects. Each box represents an impact. Each * indicates a baseline impact within the test flow to verify that the system had not changed.

Full size image

Data Collection and Analysis

All data were collected using a SlicePro data acquisition system (DTS, Seal Beach CA) at a sampling rate of 20,000 Hz. Time zero in each test was defined as the time of initial contact between the impactor face and the thorax, and was determined electronically by using conductive tape on the contacting surfaces, with wires attached to the tape such that contact completed an electrical circuit. In all cases, initial contact occurred near the bottom edge of the impactor, as the chest depth was greater at the inferior thorax than at the superior thorax. All data were filtered during post-processing in Matlab using a two-direction (phaseless), second order Butterworth filter with a cutoff frequency of 300 Hz, which is comparable to the SAE Channel Frequency Class (CFC) 180 (SAE J211, 2007). Although chestbands were placed on the subject, chest deflection data was obtained from the ram displacement—not from the chestbands. This provided a consistent measure for deflection for both impacts with and without any chestbands. Normalized chest deflection and thoracic stiffness were examined as the basis for evaluating the effects on global response, while rib strains were used to evaluate effects on localized loading. For a diagram showing the structure of the data collection and analysis process, see Figure S1 in the supplementary material. The analytical methods for the impact velocity, deflection, and stiffness are explained in greater detail in the following subsections.

Impact Velocity

Multiple redundant sensors were used to calculate impact velocity. The filtered ram acceleration (from accelerometer) was integrated to get the ram velocity. Additionally, the filtered ram position (from potentiometer) was differentiated to get the velocity. The differentiated ram position was selected as the best method for calculating velocity because of vibrational noise experienced by the accelerometer. Impact velocities for each test are included in the supplemental material.

Chest Deflection

The values for absolute chest deflection (c) were obtained for each test as the ram displacement after initial contact, as obtained from the impactor potentiometer and not from the chestbands. In order to remove variation in actual impact velocity (v a) as a potential factor for variation in peak chest deflection, the deflections were normalized (d) about the target impact velocity (v t), as shown in Eq. (1). A linear relationship between deflection (stored energy) and velocity (kinetic energy) was deemed sufficient for normalization scaling because only small deviations in actual impact velocity were being accounted for. Target velocities, actual velocities, and peak absolute deflections are included in the supplemental material.

$$d = c\frac{{v_{\text{t}} }}{{v_{\text{a}} }}$$
(1)

With the normalized chest deflections calculated, deflection differences were then calculated as the basis for comparing the deflection for each number of chestbands used. Uncertainty was then calculated for individual deflections and extended to deflection differences. The individual uncertainty was calculated as the 3-degree-of-freedom 95% confidence interval, which was 3.2 times the standard deviation of the first 4 baseline impacts. The uncertainty in each chest deflection difference was calculated as the root sum of squares of two impacts, each impact having an uncertainty equal to the individual deflection uncertainty.16

In order to determine whether observed differences were significant, a paired two-tailed t-test with six degrees of freedom was conducted on the combined set of eight deflection differences comparing response with chestbands to response without chestbands (shown in the results, Fig. 6b). The null hypothesis of the test was that there is no difference between response with and without chestbands (d 1 or 2 − d 0 = 0). The alternative hypothesis was that there is a difference between response with and without chestbands (d 1 or 2 − d 0 ≠ 0). A pooled standard deviation was used for the test, which was calculated as the average standard deviation from the two subjects because an equal number of data points were used from each subject to calculate the individual standard deviations.

Thoracic Stiffness

The A/P force from the load cell was inertially compensated to account for the mass of the impactor plate, thus representing the complete force exerted on the thorax. The compensated force (F) was then used to produce force–deflection curves as well as to calculate the thoracic stiffness. Stiffness (k) was calculated for the linear spring model containing equal potential energy (PE) as the thorax at peak deflection (c) (Eq. 2). Stiffness values are included in the supplemental material.

$$PE = \frac{1}{2}kc^{2} \to k = \frac{2PE}{{c^{2} }},\;{\text{where}}\;PE = \mathop \smallint \nolimits_{0}^{{{\text{Peak}}\;{\text{Deflection}}}} F \cdot dc$$
(2)

The methods for calculating stiffness differences, stiffness difference uncertainty, and for testing the significance of stiffness differences were identical to the methods evaluating the differences in the chest deflection.

Rib Strain

Analysis of rib strain sought to determine whether chestbands concentrate the input load, thus altering the observed rib strain in the contacted ribs. Thus, analyses were focused specifically on those ribs which were in direct contact with a chestband (ribs 3–4 and ribs 6–7). As with chest deflection, rib strain was first normalized by impact velocity in order to directly compare impacts. Changes in peak strain were then calculated for bilateral ribs 3–4 with the first chestband added to the system (s 1–s 0). Similarly, changes in peak strain were calculated for ribs 6–7 when adding the second chestband after the first was already in place (s 2–s 1).

The uncertainty for each strain difference was calculated for each rib by the same methodology as was used in the uncertainty of deflection differences. Three paired two-tailed t-tests were conducted to evaluate significance: (1) the dataset of ribs 3–4, (2) the dataset of ribs 6–7, and (3) the combined dataset of ribs 3–4 and 6–7. In each instance, a pooled standard deviation was calculated as the average standard deviation for each rib within the dataset. The three t-tests had 15, 15, and 30 degrees of freedom, respectively.

Results

Injuries

Rib fractures were identified during testing by sudden, sharp drops in the strain gage output during loading.10 For subject 1, fractures occurred in the 3rd and 4th right ribs during the 17th impact, which was the second impact to occur at the 2.0 m/s impact velocity (see Fig. 4). One chestband, at the level of the axilla, was present on the subject during the injurious impact. Both fractures were transverse and were located slightly posterior of the strain gages, but still anterior of lateral (approximately 60% of rib curve length). Subject 2 had one fracture occur on the right 8th rib during the 7th impact, which was the first impact to occur at 1.5 m/s. The fracture was located at 95% of the rib curve length (anterior, near the CCJ), and both chestbands were in use when the injury occurred.

Figure 4
figure 4

(a) Fractures in right ribs 3 and 4 from subject 1 and right rib 8 fracture from subject 2. (b) Each fracture is shown across the top, and each associated strain gage plot is shown across the bottom.

Full size image

Global Response Data

Figure 5 contains a series of plots for PMHS 1 from which calculations for chest deflection and thoracic stiffness were conducted. Figure 5a shows the normalized deflection time-histories for baseline impacts, and provides a visual idea of the variability present among similar impacts. The peak deflection values from these baseline impacts were used to calculate the standard deviation for the chest deflection statistical analysis, described in the methods. Figure 5b shows the normalized chest deflection time-histories for impacts conducted at 1.0 m/s and 1.5 m/s, providing a comparison between impacts with varying number of chestbands. Figures 5c and 5d show force–deflection curves for the same impacts presented in Figs. 5a and 5b. Stiffness values were calculated from these curves, the values from Fig. 5c being used to calculate standard deviation and the values from Fig. 5d being used to calculate response differences.

Figure 5
figure 5

Global response data from PMHS 1. (a) Normalized chest deflection for baseline impacts. (b) Normalized chest deflection for impacts at 1.0 m/s (solid) and 1.5 m/s (dashed). (c) Force vs. deflection for baseline impacts. (d) Force vs. deflection for impacts at 1.0 m/s (solid) and 1.5 m/s (dashed).

Full size image

Chest Deflection

One example of the peak deflection values (PMHS 1 at 0.8 m/s) is shown in Fig. 6a. Additional deflection values for the various impacts are included in the supplementary material, Table S2. In order to visualize any effect of chestbands on chest deflection, differences in normalized deflection based on the number of chestbands are shown in Fig. 6b. Because Fig. 6b is showing differences in deflection, a value of zero indicates that the deflection did not change when the number of chestbands changed (i.e., chestbands had no effect). Looking at the individual impact comparisons, a difference of 0 is contained within the 95% confidence interval for the deflection difference in every case except for one. Viewing the combined set of 8 deflection differences which compare impacts with at least one chestband to impacts without chestbands, the differences in deflection appear to vary about zero, having a 0.8 mm average increase (approximately a 2% change) with chestband use. Using the same combined set of differences, a two-sided t-test identifying the significance of the change in chest deflection from using chestbands produced a p value of 0.35. When comparing impacts with one chestband vs. two chestbands, differences in chest deflection were even smaller, showing an average 0.3 mm decrease in deflection for adding the second chestband.

Figure 6
figure 6

Effects of chestbands on thoracic deflection. (a) Peak normalized deflection for impacts on PMHS 1 at 0.8 m/s. (b) Differences between peak deflections for impacts at the same speed, comparing 0 chestbands with 1 chestband (d 1 − d 0), 0 chestbands with 2 chestbands (d 2 − d 0), and 1 chestband with 2 chestbands (d 2 − d 1). Error bars indicate the 95% confidence interval.

Full size image

Thoracic Stiffness

Results for thoracic stiffness appeared much the same as results for chest deflection, and are shown in Fig. 7. As an example of the observed thoracic stiffness for the various impacts, Fig. 7a shows stiffness values for impacts on PMHS 1 at 0.8 m/s. Stiffness values for all impacts are included in the supplementary material, Table S2. Calculated stiffness differences for comparison are shown in Fig. 7b. Every individual impact comparison has a 95% confidence interval which crosses the line for zero difference. The combined set of differences comparing impacts with at least one chestband vs. impacts without chestbands showed a 0.4 N/mm decrease in stiffness (approximately a 3% change) with chestband use. The associated two-sided t-test for statistical significance yielded a p value of 0.42.

Figure 7
figure 7

Effects of chestbands on thoracic stiffness. (a) Stiffness values for impacts on PMHS 1 at 0.8 m/s. (b) Differences between thoracic stiffness for impacts at the same speed, comparing 0 chestbands with 1 chestband (k 1 − k 0), 0 chestbands with 2 chestbands (k 2 − k 0), and 1 chestband with 2 chestbands (k 2 − k 1). Error bars indicate the 95% confidence interval.

Full size image

Rib Strain

Peak strain values for impacts on the right third rib are shown in Fig. 8a as an example of the strain values from which differences were calculated. The peak strain values for all ribs are included in the supplementary material, Table S3. Average differences in peak strain for each rib level contacted by a chestband are shown in Fig. 8b. Rib levels 3–4 were in contact with the superior chestband and the rib levels 6–7 were in contact with the inferior chestband. Average changes in rib strain were 42 µS (2% change) and 235 µS (4% change), from use of the superior chestband and the inferior chestband, respectively. T-tests were conducted to evaluate whether the observed strain differences were significant. The statistical test was performed on both datasets individually as well as for the combined dataset in the figure. None of the t-tests produced statistically significant p values: rib levels 3–4 had a p value of 0.7, rib levels 6–7 had a p value of 0.09, and the combined set of levels had a p value of 0.42.

Figure 8
figure 8

(a) Peak strain values in the right 3rd rib during impacts with zero chestbands and one chestband. (b) Average change in peak strain for ribs in contact with a chestband. Ribs level 3–4 are shown on the left, comparing impacts with one chestband vs. zero chestbands. Ribs level 6–7 are shown on the right, comparing impacts with two chestbands vs. one chestband. Error bars represent a 95% confidence interval for the average strain difference.

Full size image

Discussion

The present study was reliant upon data from repeated impacts to the same thorax, and some of the test setup details described in the methods were selected to maximize both repeatability and the number of impacts which could be conducted without thoracic damage. The first was the impactor face. Previous studies have used a linear impactor to interact with the thorax, and have commonly used a 23 kg impactor with a 6″ circular impactor face.12,13,26 The current study applied the same impactor mass, but used a larger rectangular impact face in order to minimize the chance of fracture due to edge effects. Additionally, the testing began at the lowest impact velocity which could be repeatedly produced by the impactor, and was gradually increased.

Global thoracic response was characterized in the present study by the chest deflection and the thoracic stiffness. It was observed in tests with chestbands that discrepancies between chestband-based deflection measurements and impactor displacement-based deflection measurements were within 2 mm (5%). In order to maintain consistency across all tests in the analysis, including impacts without chestbands, the impactor displacement-based measure was used for all tests. Looking at chest deflections for individual impact comparisons in Fig. 6b, the 95% confidence interval, which portrays the variability among repeated baseline impacts, crosses over zero in every case except one. This indicates that the observed differences in chest deflection could simply be due to random experimental variation rather than to the use of chestbands. Furthermore, when the deflection data were combined and tested for statistical significance, it was found that the data was far from significant with a p value of 0.35 for a mean deflection difference of less than 1 mm from chestband use. Additionally, the slight average increase in measured deflection from chestband use could be attributed to the thickness of the chestbands and ribbon cables sticking out from the thorax and initiating contact sooner.

Calculated stiffness values in the present work were observed to increase as impact velocity was increased. Impacts at 1.5 m/s showed stiffness values very comparable to the hub loading case in Kent et al.12 For example, CADVE146 (a small older female) in Kent et al.12 had a reported effective stiffness of 3687 N/100% deflection, whereas the 19.1 N/mm stiffness reported for impact 14 on PMHS 1 (a small older female) in the current study (Table S2) corresponds to an effective stiffness of 3629 N/100% deflection. Stiffness differences from chestband use were even smaller than differences in chest deflection, relative to their individual variability (Fig. 7b). The 95% confidence intervals for the observed differences in stiffness included zero in every case, again suggesting that random experimental variation may be the cause of the observed stiffness differences. The attribution of observed differences to random variation was supported by the statistical treatment of the dataset, having a p value of 0.42. Both aspects of global response failed to show any effect from the use of chestbands.

Local response was evaluated based on the strain observed in the individual ribs. As observed in Fig. 8b, the average difference in strain for each rib level from chestband contact was much smaller than the variability in strain from test-to-test. Viewing the observed differences statistically, changes in rib strain from chestband use were not significant, whether looking at the first chestband, the second chestband, or both. As with global response, the local response also failed to show any effect from chestband use and interaction with the ribs.

The development of the chestband was motivated by concerns regarding accuracy in existing methods, and particularly about the transthoracic rod’s effects on thoracic response and injuries.6 The chestband improved upon available techniques through an ability to observe complete chest contours rather than just point comparisons, and did not have to disturb the thoracic cavity. Previous use of the thoracic rod inherently assumed it did not affect thoracic response, and chestband use has also relied upon an inherent assumption that they do not alter thoracic response.4,5,20,23 The results of the present work provide evidence to support the assumption of negligible effects on the thoracic response, both at the global and local response levels. Consequently, chestbands may be used with some confidence that the thoracic system under study has not been significantly changed by the instrumentation.

Limitations

While the present work provides evidence to support the assumption of negligible thoracic effects from chestbands, there are also important limitations to consider in the application of the results. First, the sample size of two PMHS makes it impossible to provide definitive conclusions. Additionally, the need to avoid fracture for the repeated loading necessitated low-energy impacts and thus did not replicate impacts of the same energy typically utilized in vehicle occupant crash studies. It was expected, though, that any observed chestband effects should be most greatly manifested in low-energy impacts and should become more negligible as the impact energy increases. Thus, effects should be negligible at higher energy impacts because the effects appear to be negligible at lower energy impacts. Another limitation in the application of this study is that it specifically examined frontal impact, assuming that the results will be similar for oblique and lateral impacts. Finally, while the results failed to show a significant effect in thoracic response as a result of chestband use, the statistical methods employed do not necessarily indicate the chestbands have no effect on response. The data serves as evidence to support the common assumption of negligible chestband effects, but does not constitute conclusive proof.