UTILIZING COMPUTED TOMOGRAPHY SCANS FOR ANALYSIS OF MOTORCYCLE HELMETS IN REAL-WORLD CRASHES Kathryn L. Loftis 1, Daniel P. Moreno 1, Joshua Tan 2, Hampton C. Gabler 1, Joel D. Stitzel 1 1 Virginia Tech- Wake Forest University Center for Injury Biomechanics 2 Wake Forest University Center for Biomolecular Imaging ABSTRACT In 2008, there were more than 5,000 motorcycle crash fatalities in the United States. Many states have motorcycle helmet laws that are meant to protect riders during a crash. After recruiting motorcycle occupants injured in crashes, a protocol was established to scan three different types of motorcycle helmets commonly worn (cap, ¾ shield, and full face shield) using a computed tomography (CT) scanner. The protocol developed was for a GE 64 slice PET/CT Discovery VCT scanner with axial images from anterior to posterior helmet acquired in helical mode. It had 512x512 resolution and the full face and ¾ face shield helmets were scanned with greater voxels in the axial plane compared to the skull cap helmets. New helmets were scanned as exemplary images for comparison with helmets involved in motorcycle crashes. After CT scans were gathered, three-dimensional reconstructions were made to visualize scratches and impacts to the exterior of the helmets. Initial work was also conducted in analyzing interior components, and a trend was seen in decreased thickness between the interior foam and shell with sides of the exterior helmet thought to have contacted roadside barriers or the ground during motorcycle crashes. These helmet analysis methods have been established, and will be used to investigate multiple motorcycle crashes in conjunction with occupant injuries and direct head impacts to improve helmet design and the understanding of head injuries. This work also establishes the basis for development of finite element models of three of the most common helmet types. Keywords: motorcycle, injury prevention, computed tomography, head injury, finite element modeling, 3D reconstructions INTRODUCTION According to NHTSA, there were more than 5,000 motorcycle fatalities in the United States (US) in 2008, with an increase in deaths from 2007. NHTSA reported that in 2008, 63% of motorcyclists were wearing Department of Transportation (DOT)-compliant helmets. [1] s approved by the Federal Motor Vehicle Safety Standard (FMVSS) 218 have been shown to be significant countermeasures for reducing severe head and neck injury and a full facial coverage helmet provided increased protection for reducing facial injuries [2]. With many state-sponsored training classes advocating the greater use of helmets as a means of decreasing rider fatalities, it is important to evaluate helmet technology and provide the most comfortable fit while also providing protection [3]. One important method for determining the effectiveness of wearing a helmet during a crash is to develop finite element models for testing. This would allow different helmet types to be tested in the same conditions to evaluate new helmet designs and injury metrics for the skull and brain for association with injury severity measures [4]. To develop these FE models, it is important to have accurate representations of helmets with appropriate material properties for the shell and interior foam [5, 6]. This study has developed computed tomography (CT) scanning protocol for three different types of motorcycle helmets. New, unused helmets were scanned as the control group, and then helmets from riders who were involved in motorcycle crashes were scanned. The hypothesis was that the space between the interior helmet foam and the helmet shell would be increased on the side of the helmet that
had sustained an impact due the compression of the foam leaving a larger gap. To investigate this hypothesis, a measurement technique was developed for taking measurements off the axial slices and three-dimensional renderings were created of each helmet. This work provides the basis for developing the geometry for incorporation into future FE models, and provides insight into the characteristics of a helmet after impact loading. METHODS After gathering new motorcycle helmets, and helmets from recruited motorcycle crash patients, CT protocols were developed for scanning. Study protocol of motorcycle patients was approved by the WFUBMC Institutional Review Board (IRB). A GE 64 slice PET/CT Discovery VCT scanner (GE Healthcare, Milwaukee, WI) was used, with 512x512 resolution and scanning specifications of 120 kv and 300mA. Axial images were acquired from anterior to posterior helmet in helical mode. For the skull cap helmet, there was 0.625 mm slice thickness and scan field of view (SFOV) of 32 cm. Given the different helmet types, different protocol was developed for the larger helmets. The ¾ and full face shield helmets had the same 0.625 mm slice thickness, but had an SFOV of 50 cm. All CT scans were then imported into Mimics (Materialise Interactive Medical Image Control System, Leuven, Belgium) for viewing and measuring. Measurement techniques were developed for recording the free space between the interior helmet foam and the exterior shell of the helmet. For each scan, Hounsfield value ranges of -1024, -826 were chosen for taking measurements, as this provided the best contrast for identifying each different part of the helmet. The most anterior and posterior slices on the helmet were recorded as locations, and this was used to calculate three evenly-spaced measurement slices for each helmet anterior, middle, and posterior (Figure 1a). The median of each of these slices was then found by scrolling through the sagittal slices. Using this median measurement for each of the three chosen slices, 4 measurement locations were calculated two left and two right evenly-spaced on each side. The distance between the inside of the shell and the outside of the interior foam was then measured and recorded (Figure 1b). By recording these thicknesses at evenly spaced points on either side of the center, the left side could be compared to the right side to investigate trends in free space between the struck and unstruck sides of the helmets. Where no pixels with darker Hounsfield units could be seen between the shell and foam, the measurement was recorded as zero. Posterior Anterior Median Posterior Measurement Middle Measurement Anterior Measurement Right, Right, Left, Superior Superior Left, Median a) b) Figure 1. Measurement techniques a) locating three evenly spaced slices for measuring b) taking measurements between interior foam and helmet shell at evenly spaced locations from the median helmet on this slice After taking the measurements, the helmet shell was segmented by starting with a Hounsfield unit range for bone and then manually adjusting until the helmet shell was fully segmented. Using the Shell Foam
segmentation, a three-dimensional (3D) rendering of the helmet shell was created to locate contact points on the exterior where impact had occurred. These points were also correlated with exterior helmet photos for the helmets that had been involved in crashes. RESULTS While this was a small dataset, results identified trends that should be investigated further. The first three helmets listed were the unused helmets (MC000-1, MC000-2, MC000-3). These represented three main different types of helmet styles: skull cap, ¾ face shield, and full face, respectively. The middle slice of each CT scan is shown in Figure 2, with the 3D reconstruction of the helmet exteriors below each to visually compare the differences between the helmets. In these images, the artifact created from the metal parts of the helmets can also be seen. MC000-1 MC000-2 MC000-3 Figure 2. New helmets displaying the three different types by their middle CT scan axial slice and exterior shell 3D reconstructions from segmentation MC000-1 had the most consistent measurements throughout the helmet. MC000-2 was difficult to measure due to the inconsistent interface between the interior foam and the shell. MC000-3 had no measurable space between the foam and the shell at any of the superior measurement points. The next five helmets were all involved in crashes and were skull cap-type helmets, but only three of these had evidence of direct impact. Measurement of the free space was highly dependent upon the different helmet types. MC004-D had the least amount of interior foam and support of all these helmets. Each of the measurements can be seen in Table 1, with all units in mm.
MC001-P MC001-D MC003-D A C Figure 3. 3D segmentation of helmet shells, showing exterior contact points B Table 1. Motorcycle helmet free space measurements (units = mm) Number MC000-1 MC000-2 MC000-3 MC001-D MC001-P MC002-D MC003-D MC004-D Location Right, Right, Superior Left, Superior Left, Anterior 1.55 1.41 1.41 1.55 Middle 3.51 3.09 3.37 2.81 Posterior 4.36 2.11 2.11 2.25 Anterior 0.77 NA NA 0.62 Middle 0.94 12.99 12.99 0.71 Posterior 0.62 2.48 2.48 NA Anterior 73.4 0 0 69.55 Middle 1 0 0 0.67 Posterior 1.07 0 0 0.71 Anterior 0 0 0 0.84 Middle 0 1.27 2.39 0 Posterior 0.56 2.67 2.81 1.41 Anterior 0 2.71 2.2 0 Middle 2.2 2.54 2.2 1.02 Posterior 3.22 2.88 3.05 3.05 Anterior 1.13 0.71 0.85 1.28 Middle 2.13 0.99 0.99 1.13 Posterior 1.56 1.42 1.42 1.13 Anterior 0 2.11 2.11 1.05 Middle 0.75 3.16 3.01 2.41 Posterior 2.11 2.71 2.56 2.86 Anterior 18.25 3.09 3.93 12.92 Middle 1.12 12.07 11.79 0.56 Posterior 2.11 7.58 7.3 1.4 After subtracting the left side measurements from the right side measurements at each slice (anterior, middle, posterior), the average difference was calculated. According to this average, the side with the larger space was recorded, also noting which side of the helmet had loading evidence, if applicable. The results of these measurements can be seen in Table 2, with helmets having an exterior contact point identified in bold. Because the left side measurements were subtracted from the right side, values that are negative indicate that the left side had a larger space than the right. Table 2. Motorcycle helmet free space measurement results Side with more free space at superior measurement, Average difference Side with more free space at inferior measurement, Average difference impacted side MC000-1 Left, -0.09 mm Right, 0.94 mm None MC000-2 None, 0 mm Right, 0.19 mm None MC000-3 None, 0 mm Right, 1.51 mm None MC001-D Left, -0.42 mm Left, -0.56 mm Right MC001-P Right, 0.23 mm Right, 0.45 mm Left MC002-D Left, -0.05 mm Right, 0.43 mm None MC003-D Right, 0.10 mm Left, -1.15 mm Right MC004-D Left, -0.09 mm Right, 2.2 mm None The results of the segmentation showed it was possible to view exterior scratches and contact points on the helmets to identify where impacts had occurred. The three helmets with exterior contacts
are shown in Figure 3. Exterior contact points were also correlated with helmet photos for agreement. The head, face, and neck injuries associated with each of the crashes is shown in Table 3. MC0004-D was a special crash, with the rider becoming hung on a cable barrier under his chin, creating significant neck, jaw, and cervical spine injuries. Table 3. Motorcycle cases and injuries to the head, face, and neck as a result of the crash Case Head, Face, Neck Injuries MC001-D None MC001-P Subarachnoid hemorrhage, right occipital condyle fracture, C7 lamina fracture, facial abrasions and contusions MC002-D C-spine transverse process fractures, neck laceration, nasal & right maxilla fracture, large neck hematoma MC003-D Right maxilla fracture MC004-D Serious neck and cervical spine injuries, including spinal cord and trachea, and left mandible fracture DISCUSSION Results of the helmet scans showed large differences in construction and interior foam for each type of helmet. While most of the helmets had a hard outer shell, a thick interior foam, and then an adjustable interior thinner foam for fit to the head, the MC004-D helmet did not have the thick layer of foam covering the interior surface of the helmet shell. This resulted in the large values seen in the anterior slice for the right and left inferior measurements. Even within helmets of the same style category, variations in construction were identified. After analyzing the measurement results, a trend was seen in the helmets that sustained an impact. For these helmets, the struck side had less free space between the interior foam and outer helmet shell. While the hypothesis was that the struck side would have larger free space due to the compression of the foam, the data showed otherwise. It may be that the interior helmet becomes shifted towards the struck side after an impact, reducing free space on the struck side and increasing on the opposite side. The struck helmets consistently had more space on the opposite side for both superior and inferior measurements, except for MC003-D. After visually locating the exterior contact on the helmet shell for MC003-D, the superior free space measurements may not have been greatly affected by the impact to the low right side of the helmet, resulting in the right side free space measuring slightly more than the left superiorly. For all helmets without strikes, the inferior measurements were greater on the right side while the top measurements were either greater on the left side or showing no difference. These similarities could be due to slightly off-center rotations of the helmets in the scanner or partial volume effects, where the interface between the helmet shell Hounsfield unit values (white) and the air Hounsfield unit values (black) occur within the same voxel, producing a combination of the two values (grey). The injuries associated with each helmet showed that these motorcyclists likely had reduced head injury severity due to the helmets. This was especially seen for MC001-D, where there was a visible contact to the helmet, but no head injury was sustained by the rider. MC001-P was the only case occupant that sustained brain injury as a result of the crash, and the helmet showed a significant contact on the left side. There were also no serious scalp lacerations as a result of direct contact of the head with an exterior object. MC004-D sustained serious injury to his neck due to contact with a cable barrier. These helmet scans had a large amount of artifact due to metal components which could not be removed because helmets had to be returned to the owners. This artifact made it difficult to perform segmentations of the interior foam. Another limitation was the resolution of the images, which discounts some of the small thickness measurements calculated. Values listed as NA were unavailable for measurement due to the helmet construction. Future work will focus on enhancing the measurement and segmentation techniques. This may involve automation, so that more measurements could be
collected, and extended to measuring the foam thickness. With more sophisticated segmentation techniques, it will also be possible to segment the interior foam and compare volumes. These techniques may also be useful for work with sports helmets, such as football helmets and bicycle helmets. CT scan segmentations, as seen in Figure 3, provide important geometry for further development of finite element models of different helmet styles and impact types. By taking CT scans of helmets before and after controlled laboratory impact tests and correlating foam and helmet measurements with known insults produced in the lab, it will be possible to create finite element models that are able to investigate head injury reductions between different helmet designs and foam types. This methodology could also be used to correlate foam compression measurements with impact forces and accelerations, so that estimates of impact energy could be made for real-world helmet impacts that produced head injury. This could be extended to other helmet types as well, including football helmets and bicycle helmets. CONCLUSIONS In conclusion, this pilot study developed the protocol for scanning different types of motorcycle helmets. After scanning, methods were established for recording measurements from the CT slices and creating 3D segmentations of the helmet shells for viewing contact points on the exterior. Small differences were measured between the interior foam and the helmet shells between left and right sides, and there were large variations in measurements between helmets of different styles. A trend was seen in helmets with exterior contact points, having reduced space between the interior foam and shell on the struck side. This information provides the basis for a standard methodology of measurements for continued research on helmets and their protective effects. ACKNOWLEDGMENTS The authors would like to acknowledge the Center for Injury Biomechanics students, faculty, and staff for their contributions to this work and also the Center for Biomolecular Imaging for use of the CT scanner and their help developing the protocol. REFERENCES [1] NHTSA, "Traffic Safety Facts 2008." DOT HS 811 159, Dept of Transportation, Ed.: National Center for Statistics and Analysis, Washington, D.C. 20590, 2010. [2] H. H. Hurt, J. V. Ouellet, and D. R. Thom, "Motorcycle Accident Cause Factors and Identification of Countermeasures." vol. 1: Technical Report, L. A. Traffic Safety Center, CA, Ed.: National Technical Information Service, Springfield, VA, 1981. [3] A. Daniello, H. C. Gabler, and Y. Mehta, "The effectiveness of motorcycle training and licensing," in Transportation Research Record: Journal of the Transportation Research Board, No. 2140, Transportation Research Board of the National Academies, Washington, DC, 2009, pp. 206-213. [4] C. B. Meeks, S. E. Day, and A. D. Foreman, "The influence of non-visible damage on the performance of aircrew helmets," in Survival and Flight Equipment Association Edinburgh, UK: SAFE Europe, 2007. [5] P. Mahahan and P. K. Pinnoji, "Finite Element Modeling of ed Head Impact Under Frontal Loading," Transportation Research and Indian Institute of Technology Delhi Research Review, vol. 1, p. 4, October 1, 2009. [6] N. J. Mills, S. Wilkes, S. Derler, and A. Flisch, "FEA of oblique impact tests on a motorcycle helmet," International Journal of Impact Engineering, vol. 36, pp. 913-925, 2009.