Human Head Models from MRI for Head Impact Analysis
Yash Agarwal1, Philippe Young1, Ross Cotton1, Chris Pearce2, Siddiq Qidwai3, Amit Bagchi3, and Nithyanand Kota3

1Simpleware Ltd., Exeter, United Kingdom, 2Atkins, Epsom, United Kingdom, 3U.S. Naval Research Laboratory, Washington, DC, United States

Synopsis

Image-based model generation methods demonstrate the value of creating realistic human head models based on high-resolution MRI data. Head models created by the U.S. Naval Research Laboratory and Simpleware (Exeter, UK) are being used to study head impact and traumatic brain injury; this offers a solution to the problem of limited experimental testing. Results from the modelling methodology and simulation demonstrate a good level of accuracy when compared to experimental benchmarks. The methodology and models have been extended for use in areas such as examining head impact in sports including American football, rugby and cricket.

PURPOSE

Traumatic brain injury (TBI) is a serious condition affecting multiple groups, from soldiers to athletes engaged in activities involving blunt force to the skull. Long-term neurological and other motor disorders are present in between 14% and 20% of veterans of the Iraq and Afghanistan wars,1 while TBIs have been widely reported in relation to sports injuries. Although significant research has been carried out into the mechanisms of TBI, this has been limited by cadaver and other experimental samples. Numerical modelling using computational models from MRI data offers potential for flexible yet robust research into TBI and other forms of head impact.

METHODS

Solutions to the challenge of creating computational models of the human head from scans have typically involved producing computer-aided design (CAD) models, with an idealised geometry used to produce a numerical model, or mesh, suitable for finite element analysis of impact. CAD methods, however, produce issues with complex anatomies and internal structures, leading to a large amount of manual fixing. The majority of models are also designed for specific applications.

For example, a head model has been created from 14 cross-sectional MRI slices and processed using C++ code before being converted into CAD coordinates and spline data,2 while models have been produced by automotive companies to handle particular inputs like crash impact.3 These models have the limitation of being ‘fixed’ for their applications and not easily adapted for new inputs. The NRL-Simpleware model produced from MRI data is designed to solve this flexibility issue by creating finite element meshes from pre-segmented image data, rather than working with pre-defined meshes.

The model was developed by acquiring a whole head in vivo MRI scan of a 25-year old male volunteer at the Exeter MR Centre, UK. The T1-weighted scan produced coronal plane image slices with a resolution of 1.03516 mm x 1.03516 mm, with a slice-to-slice separation of 1.4001 mm. The DICOM image series was imported from the MRI scanner to Simpleware ScanIP software for image processing (Fig 1).

Regions of interest (ROIs) within the greyscale data were segmented and labelled, including brain, skull and muscles. A range of image processing tools were used to fully reconstruct the geometry of the head, skull and neck. Parts not available from the scan, such as facial and neck muscles, were imported into the software for conversion to image masks. This method also allows new objects, such as helmets, to be incorporated within the segmented anatomical data (Fig 2).

Multi-part meshes were generated from the segmented data, with options available for automatically converting ROIs into volumetric FE models using a multi-part Extended Volumetric Marching Cubes (EVOMAC) approach, and for multi-part surface decimation technique that allows control over the size of the model.4 The particular NRL-Simpleware model used for simulation of TBI was produced as an unstructured, all-tetrahedral mesh with 3.72 M volumetric elements.

The NRL-Simpleware model was set up for specific blunt impact (low-to-mid rate) and blast overpressure (mid-to-high rate) simulations in FE solver Abaqus/Explicit. Blunt impact simulations were validated against experimental data in the literature5 and blast overpressure simulations conducted using established thresholds.6, 7

RESULTS

The blunt impact study involved post-processing of the simulation data to quantify injured brain volume as a function of event time based on injury threshold measures in the literature.8 Python scripting was used to calculate evolution of these volumes based on a pressure threshold of 173 kPa9 and a shear strain threshold of 5%,7 with good agreement found with the literature. For blast overpressure loading, a frontal loading with a Friedlander wave profile10 was used with a peak pressure of approximately 430 kPa (Fig 4).

DISCUSSION

Blunt impact loading simulations using the NRL-Simpleware model that were validated against experimental data showed excellent agreement with benchmark results. Blast overpressure results, designed to reproduce conditions associated with military and explosive weaponry, also demonstrated good results for predicting brain injury as a function of both location and time.

CONCLUSION

The NRL-Simpleware model demonstrates promise for both military and clinical research into TBI and other forms of head impact. The novel ability to easily generate new finite element meshes from the segmented image data means that the model can be tailored for specific applications, including electromagnetic simulation. Moreover, the head model is being used in sports research, for example in American football and for studying the effect of cricket ball impact, and is generating promising insights into the mechanics of head injuries and other scenarios that can be modelled from MRI and other image data.

Acknowledgements

This work was supported by the Office of Naval Research (ONR) through the US Naval Research Laboratory's Basic Research Program, and the Department of Defense (DoD) High Performance Computing Modernization Program (HPCMP) using the Air Force Research Laboratory (AFRL) Major Shared Resource Center (MSRC) under project 416, subproject 231.

References

1. U.S. Department of Veterans Affairs. Traumatic brain injury: a guide for patients. http://www.mentalhealth.va.gov/docs.tbi.pdf. Accessed November 10, 2015.

2. Mehta, BV, Rajani, S. Sinha, G. Comparison of image processing techniques (magnetic resonance imaging, computed tomography scan and ultrasound) for 3D modeling and analysis of the human bones. J. Digit. Imaging. 1997; 10: 203-206.

3. Takhounts, EG, Ridella, V, Hasija, RE, Tannous, JQ, et al. Investigation of traumatic brain injuries using the next generation of simulated injury monitor (SIMon) finite element head model. Stapp Car Crash J. 2008; 52:1-31.

4. Young, PG, Beresford-West, TBH, Coward, SRL, et al. An efficient approach to converting three-dimensional image data into highly accurate computational models. Philos Trans R Soc A Math Phys Eng Sci. 2008; 366:3155-3173.

5. Nahum AM, Smith, R, Ward, CC. Intracranial pressure dynamics during head impact. Proc. Of the 21st Stapp Car Crash Conf. 1977: 339-366.

6. Kang, HSR, Willinger, R, Diaw, N et al. Validation of a 3D anatomic head model and replication of head impact in motorcycle accident by finite element modeling. SAE Trans. 1997; 106(6):3849-3858.

7. Margulies, SS, Thibault, LE. A proposed tolerance criterion for diffuse axonal injury in man. J Biomech. 1992; 25(8):917-923.

8. Besenski, N. Traumatic injuries: imaging of head injuries. Eur Radiol. 2002; 12(6):1237-1252.

9. Zhang, L., Yang, KH, King, AL. A proposed injury threshold for mild traumatic brain injury. J Biomech Eng. 2004: 126(2):226-236.

10. Dewey, JM. The shape of the blast wave: studies of the Friedlander equation. 21st international symposium on military aspects of blast and shock. 2010.

Figures

Fig 1. MRI image used to develop the head model

Fig 2. Example of a geometry modification incorporating a helmet design

Fig 3. Model generation procedure following image acquisition

Fig 4. Friedlander wave profile used in frontal blast loading simulation



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
1907