DEVELOPMENT OF A FINITE ELEMENT MODEL OF THE HEAD USING THE VISIBLE HUMAN DATA

Gerald Krabbel, Institute of Automotive Engineering,
Berlin University of Technology, Germany

Ralph Müller, Institute for Biomedical Engineering,
Swiss Federal Institute of Technology, Zürich, Switzerland*

*current adress: Orthopedic Biomechanics Laboratory, Harvard Medical School, Boston

ABSTRACT

Head injury is the most frequent severe injury resulting from traffic accidents. Head injury mechanisms are difficult to study experimentally due to the variety of impact conditions involved, as well as ethical issues, such as the use of human cadavers and animals. Finite element modeling is a comprehensive technique through which human head impact tolerance can be studied. A representative finite element human head model would allow the assessment of the injurious effects of different impact conditions and enable the development of enhanced head injury and protection criteria in the automotive environment.

The paper describes the development of a three-dimensional finite element (FE) model based on the digital data set from the head section. The fresh CT scans were used for the skull model and the MRI images for the brain model. The first approach to validate the model and to investigate different boundary conditions by using experimental data are shown.

INTRODUCTION

Head injury is the most frequent type of injury experienced by all seriously injured road users, especially car occupants. This remains true even with the introduction of enhanced restraint technology into all new cars and with the high levels of seat belt use. Most numerical simulation models of vehicle occupants, including finite element models, are based on rigid multi-segment systems resembling current Anthropometric Test Devices (ATD). As the human body's responses to impact are still very different from the responses of current ATD's, models of human body parts or segments are needed to obtain more realistic information about each body part's responses in crash events and to analyse various impact situations.

Mathematical models are valuable tools in the study of trauma. They can be used to predict body response to injury-producing conditions that cannot be simulated experimentally, and they can predict responses that cannot be measured in surrogate and animal experiments. Most important, mathematical modeling is the only means by which valid experimental animal and cadaveric data can be extrapolated to living man. Finite element analysis appears to be the most appropriate technique for analysing the complex geometrical and mechanical properties of human structures. Limitations to finite element modelling in the past were due to the assumptions and approximations incorporated in those models. These simplifications were necessary as finite element methods and CPU were restricted.

The accurate representation of the anatomically specific geometry in the finite element model enhances its function as predictive tool for skull fracture and brain injuries. With the aim of the head model we will study and gain insights to the biomechanics of head injury by simulating the real head injuries experienced in the automotive crash environment. The analysis was performed with the with the FE code PAM-CRASH, an explicit, large deformation, Lagrangian dynamic finite element program that is used widely for dynamic crash simulation and for nonlinear structure analysis.

SEGMENTATION

The fresh CT scans were scaled at an identical pixel size (0.9 mm) and transformed into tiff image format. The original measurement protocol included high-resolution images (512 x 512 pixel matrix) with a slice increment of 1 mm. In a first step, in order to construct the three-dimensional bone geometry, inner and outer bone contours were extracted slice per slice from the stack of 223 tomograms with the help of a semi-automatic tracking algorithm (Rüegsegger et al., 1993). Thereby outer contours were generated for all slices, inner contours were generated only if the thickness of the skull justified a volumetric representation for the elements. Results from the segmentation scheme are depicted in Figure 1 for four horizontal slices aligned along the superior-inferior direction of the head.

Figure 1 - Four horizontal slices aligned along the superior-inferior direction of the skull.

In a second step, a volume of interest (VOI) was defined to include the segmented bone in full detail representing a volume of 156 x 224 x 223 voxels. Figure 2 shows a view of the skull model in full resolution using an extended Marching Cubes triangulation algorithm (Lorensen and Cline, 1987; Müller et al., 1994). In Figure 3 a sagital side cut of the skull is depicted demonstrating the excellent segmentation capabilities of the implemented tracking and surface reconstruction algorithm. Because of the complex shape of the skull as seen in Figure 2 and 3 common mesh generators would need a lot of user interaction. For this reason, the FE models were constructed using an fully automatic voxel-based hexahedron mesher (VHM) creating eight-noded cube-shaped elements (Keyak et al., 1990; Müller and Rüegsegger, 1995).

Figure 2 - Three-dimensional visualization of the skull in full resolution (1 mm voxel size) including a total of 762,912 triangles and 381'418 nodes. The triangles were displayed using Gouraud shading on a pesonal workstation resulting in a pseudo-realistic representation of the skull.

Figure 3 - Three-dimensional visualization of a
sagital side cut of the skull.

VHM incorporates a simple, straight-forward approach: If a cube belongs to the bone, a hexahedron is generated at that place; if the cube is outside of the object no element is created. The grid size may either be adapted to the actual scanning resolution or may be reduced by scaling the digital model. In order to reduce the number of elements in our three-dimensional FE model, we had to scale down the binary volume data by a factor of five resulting in an element size of roughly 4.5 x 4.5 x 5.0 mm³. Since a traditional VHM approach leads to models with edged surfaces an interpolation scheme was incorporated using nodal averaging, which resulted in a FE model of the skull with smooth surfaces as depicted in Figure 4. Figure 5 shows the interior of the skull model. The same steps have been taken to generate the FE model of the brain using the MRI data (Figure 6).

Figure 4 - Finite Element Skull Model.

The skull model consists of 13,962 solids with 20,876 nodes, the brain model consists of 9,273 solids with 11,152 nodes. The material properties used were linear-elastic for the skull and visco-elastic for the brain.

Figure 5 - Mid-sagittal view of the of the human skull
model showing the cranial cavity.

Figure 6 - Interior view of the of the human skull model showing
the base of the skull with the foramen magnum.

Figure 7 - Finite Element Brain Model.

VALIDATION

The first approach to validate the model and to investigate different boundary conditions by using experimental data is shown in Figure 8. A rigid cylinder impacted the head in the mid-sagittal plane in an anterior-posterior direction (Kallieris et al., 1994). The results of the finite element analysis were compared with the test data. Good correspondence was found for the contact forces, the accelerations at the center of gravity and the dynamic motion. Furthermore the calculated distribution of the VonMises stresses showed good agreement with the observed skull fractures. The motion sequence from an 10 ms impact simulation is shown in a QuickTime movie (1,7 Mb).

Figure 8 - Cadaver head impact test set-up.

CONCLUSION

In order to develop a representative finite element human head model, we used the digital data set from the head section (CT and MRI images) within the framework of the Visible Human Project. The anatomical image data were transformed into a three-dimensional finite element model and implemented into the finite element code PAM-CRASH. After comparisons and necessary adjustments are made to the model, it is expected that the model will be able to predict the risk of head injuries in a crash event, otherwise difficult to achieve experimentally. Once validated, a mathematical model can help understanding injury mechanisms and quantify mechanical parameters related to a specific impact event so that injury tolerances can be formulated.

ACKNOWLEDGEMENTS

This work was supported in part by the Federal Highway Research Institute Germany.

REFERENCES

P. Rüegsegger, B. Münch, M. Felder, Early detection of osteoarthritis by 3D computed tomography, Technology and Health Care, Vol. 1, pp. 53-66, 1993.

R. Müller, T. Hildebrand, P. Rüegsegger. Non-invasive bone biopsy: A new method to analyse and display the three-dimensional structure of trabecular bone. Phys. Med. Biol., Vol. 39, pp. 145-164, 1994.

W.E. Lorensen, H.E. Cline. Marching Cubes: a high resolution 3D surface reconstruction algorithm. Comput. Graphics, Vol. 21, pp. 163-169, 1987.

J.H. Keyak, J.M. Meagher, H.B. Skinner, C.D. Mote, Automated three-dimensional finite element modelling of bone: A new method. J. Biomed. Eng., Vol. 12, pp. 389-397, 1990.

R. Müller, P. Rüegsegger. Three-dimensional finite element modelling of non-invasively assessed trabecular bone structures. Med. Eng. Phys., Vol. 17, pp. 126-133, 1995.

D. Kallieris, A. Rizzetti, R. Mattern. Verhalten der Kopf-Hals-Einheit bei der dynamischen Belastung-Vergleich zwischen Dummy und Leiche. Zentralblatt Rechtsmedizin, 42. Band, Heft 6, 1994.

自己当年也做过类似的一些工作，只是当时找不到从MRI文件和CT灰度文件建模的合适软件，只是自己连了粗略的外轮廓。感觉生物CAE是个交叉学科，成熟的解决方案没有看到有什么突破。这个是哈佛医学院发的论文，思路还是相当清楚的，希望对同领域的同行有所借鉴。