Journal of Clinical Neuroscience 20 (2013) 509–513

Contents lists available at SciVerse ScienceDirect

Journal of Clinical Neuroscience

journal homepage: www.elsevier .com/ locate/ jocn

Clinical Study

Decompressive craniectomy causes a significant strain increase in axonal fiber tracts

Xiaogai Li a,⇑, Hans von Holst a,b, Svein Kleiven a

a Division of Neuronic Engineering, School of Technology and Health, Royal Institute of Technology (KTH), Alfred Nobels Allé 10, SE-141 52 Huddinge, Stockholm, Swedenb Section of Neurosurgery, Division of Clinical Neuroscience, Karolinska Institutet, Stockholm, Sweden

a r t i c l e i n f o

Article history:Received 30 January 2012Accepted 22 April 2012

Keywords:Axonal fiber tractsDecompressive craniectomyDiffusion-weighted imageLagrangian finite strain tensor

0967-5868/$ - see front matter � 2012 Elsevier Ltd. Ahttp://dx.doi.org/10.1016/j.jocn.2012.04.019

⇑ Corresponding author. Tel.: +46 8 79 048 76; fax:E-mail address: xiaogai@kth.se (X. Li).

a b s t r a c t

Decompressive craniectomy (DC) allows for the expansion of a swollen brain outside the skull and hasthe potential to reduce intracranial pressure. However, the stretching of axons may contribute to an unfa-vorable outcome in patients treated with DC. In this study, we present a method for quantifying and visu-alizing axonal fiber deformation during both the pre-craniectomy and post-craniectomy periods toprovide more insight into the mechanical effects of this treatment on axonal fibers. The deformation ofthe brain tissue in the form of a Lagrangian finite strain tensor for the entire brain was obtained by anon-linear image registration method based on the CT scanning data sets of the patient. Axonal fibertracts were extracted from diffusion-weighted images. Based on the calculated brain tissue strain tensorand the observed axonal fiber tracts, the deformation of axonal fiber tracts in the form of a first principalstrain, axonal strain and axonal shear strain were quantified. The greatest axonal fiber displacement waspredominantly located in the treated region of the craniectomy, accompanied by a large axonal deforma-tion close to the skull edge of the craniectomy. The distortion (stretching or shearing) of axonal fibers inthe treated area of the craniectomy may influence the axonal fibers in such a way that neurochemicalevents are disrupted. A quantitative model may clarify some of the potential problems with thistreatment.

� 2012 Elsevier Ltd. All rights reserved.

1. Introduction

The use of decompressive craniectomy (DC) has increased sub-stantially in an effort to reduce intracranial pressure (ICP) follow-ing cerebral injury. However, a consensus on its effectiveness hasnot been achieved among clinicians1–4 due to various complica-tions.5,6 There is debate on the interpretation of DC data reportedby Cooper et al.1 and its application in clinical practice.7,8 DC allowsexpansion of the brain tissue outside the skull, thereby reducingthe ICP.5 However, the treatment also results in the stretching ofaxonal fibers, which has been suggested to contribute to unfavor-able outcomes for patients treated with DC.5 New methods, whichmay clarify the consequences of DC, are necessary to improve thistreatment.

Axons transmit electrical-chemical impulses between neuronsand intact axons, and they are critical for establishing normal neu-rological function. However, when axons are stretched, theircapacity to transmit impulses is attenuated. Stretching even causespermanent loss of functional capability in severe cases.9 Manyin vitro injury models have been developed showing that axonalstretching causes neural injury in different forms, such as neurofil-ament structure alterations,10 mechanical breaking of microtu-

ll rights reserved.

+46 8 21 83 68.

bules in axons11 and axonal swelling.12 Bain and Meaney13

demonstrated, using an optic nerve stretch model, that a strain le-vel of approximately 0.21 will elicit electrophysiological changes,while a strain of approximately 0.34 will cause morphologicalsigns of damage to the white matter. These studies have yieldedconsiderable insight into axonal alterations in response to mechan-ical stretching. In general, however, DC results in complex axonaldeformation, and it is difficult to apply these cellular level thresh-olds to the tissue level because the axons within white matter donot necessarily lie along the same orientation as the direction ofstretch. Therefore, incorporating the axonal fiber tracts into biome-chanical models is necessary to quantify axonal stretching alongthe axons, thus allowing for comparisons with the thresholds ob-tained from previous experiments. Furthermore, information onaxonal fiber deformation, such as axonal shear strain, could alsobe obtained.

The strain level representing the stretching of brain tissue hasbeen quantified in a previous study.14 It has been shown that fol-lowing DC, the strain level and water content in the brain tissuewere substantially increased. This may influence the axonal fibersin such a way that the neurochemical events are disrupted. Axonalfiber tracts extracted from diffusion-weighted images (DWI) havebeen included in a biomechanical model simulating an impactevent in order to study the axonal elongation occurring duringthe primary injury stage.15 However, axonal stretching during the

510 X. Li et al. / Journal of Clinical Neuroscience 20 (2013) 509–513

post-craniectomy period, which may have prognostic value for thecognitive and neurological sequelae of patients treated with DC,has not been studied previously.

Thus, the aim of the present study was to quantify the strain le-vel of axonal fibers following DC to gain better insight into the po-tential damage to axonal fibers caused by DC.

2. Patients and methods

2.1. Patient information

The patient was a 21-year-old male with a severe head injurydue to a fall from a height of 4 m from a ladder at work. The patientwas initially awake but then became unconscious, with a GlasgowComa Scale (GCS) score of 7. A CT scan was performed on admis-sion to the hospital. The scan showed extensive shallow subduralhemorrhage with underlying hemisphere swelling, including mul-tiple contusions at the frontal, parietal and temporal lobes on theright side of the brain (Fig. 1a, b). Due to clinical deterioration withunconsciousness and increased ICP, the patient was treated withDC within 24 hours after the trauma. Following DC, there was evi-dence of edema in the frontal lobe. The CT scans after DC showedthat the brain tissue expanded outside the skull at the treated area(Fig. 1c, d). After 23 days in the Department of Neurosurgery, thepatient was discharged for further neurological rehabilitation dueto a mild, left-sided hemiparesis from which he later recovered.Three months after the trauma, the GCS score was measured as15, and 4 months after the trauma, the patient went back to hisemployment at 75% of his previous working hours.

2.2. Quantification of the brain displacement field in pre-craniectomyand post-craniectomy periods

Displacement fields representing the structural brain changeswere obtained by a nonlinear image registration method based

Fig. 1. Axial CT scans obtained (a,b) before craniectomy showing injuries and (c, d)following decompressive craniectomy showing brain tissue expansion.

on the three-dimensional (3D) CT scanning data sets of the patientboth before and after DC. The diffeomorphic demons (DD) algo-rithm16 implemented in the open-source software Slicer 3D17

was used to account for localized distortions and large deforma-tions while simultaneously preserving the sample topology.16

Quantification of the strain level in the pre-craniectomy period re-quires brain images of the patient obtained before the patient ac-quired the TBI. However, the healthy brain image was notavailable for our patient. Herein, we propose an approach of recov-ering a healthy brain image of the patient before TBI. This was per-formed by morphing the MRI from a similar-age healthy volunteerto the CT scan of the patient according to the cranial shape. Themorphing result shows the image of the recovered healthy brainwith one axial slice and the reconstructed surface (SupplementaryFig. 1, left column, upper row).

The ventricles and the cranial geometry were segmented as bin-ary images for the recovered healthy brain. This was also done forthe pre-craniectomy and post-craniectomy CT scans. A rigid regis-tration step was first used to center the images about the samepoint before applying the DD registration. Supplementary Fig. 1(left column) shows the reconstructed surfaces of the segmentedbinary images. The images in Supplementary Fig. 1 (middle col-umn, left) also show an overlay of the rigidly aligned images, whichprovided the basic initialization for subsequent DD deformableregistration. In Supplementary Fig. 1 (middle column, right), wepresent the corresponding images after DD deformable registra-tion. The images after DD registration show good alignment, whichmeans that the displacement from normal to pre-craniectomyimages (Supplementary Fig. 1, middle column, upper) and frompre-craniectomy to post-craniectomy images has been capturedaccurately (Supplementary Fig. 1, middle column, lower). A 3Dmatching field (Supplementary Fig. 1, right column, upper) repre-senting the spatial transformation needed to displace the healthybrain images to that of the brain after injury (that is, pre-craniec-tomy) was then obtained from the DD registration. The resultingbrain tissue deformation occurring at this stage, described by dif-ferent strain measurements, is referred to as the ‘‘pre-craniectomystrain’’ throughout this manuscript. Similarly, the 3D matchingfield needed to displace the pre-craniectomy brain images to thepost-craniectomy brain images is presented in SupplementaryFig. 1 (right column, lower).

2.3. Extraction of axonal fibers for the pre-craniectomy and post-craniectomy stages

Diffusion-weighted images (DWI) were acquired using a 3-Teslascanner (Siemens Trio-Tim, Erlangen, Germany) on a healthy vol-unteer with the approval of the local Ethics Committee. T1-weighted images were taken simultaneously. The DWI werescanned with 30 gradient directions, and diffusion tensors werethen estimated using a standard least-squares method imple-mented in the Slicer 3D software, which provides a comprehensivetool for DWI and diffusion tensor processing.17 The extracted whitematter tractography, using the streamline method,18,19 containspolylines (Supplementary Fig. 2, left) with a corresponding diffu-sion tensor at each fiber point (Supplementary Fig. 2, middle).The maximum eigenvalue of the diffusion tensor, k1, with its corre-sponding eigenvector N1, is associated with the tangent to the fiberpath. The two other eigenvalues, k2 and k3, together with their cor-responding eigenvectors N2 and N3, are directly perpendicular tothe fiber paths (Supplementary Fig. 2, right).

The extracted axonal fibers were matched to those in the brainof the patient by image registration using a similar procedure, asshown in Supplementary Fig. 1. First, the ventricles and cranialgeometry from the corresponding T1-weighted MRI were seg-mented and converted to a binary image. Then, the resulting image

X. Li et al. / Journal of Clinical Neuroscience 20 (2013) 509–513 511

was matched to the segmented pre-craniectomy image of the pa-tient, resulting in a displacement field representing the inter-sub-ject brain shape difference between the patient and the healthyvolunteer. The obtained displacement field was in turn applied tothe extracted axonal fiber paths, and in this way, the axonal fiberpaths and axonal orientation information of the healthy volunteerwere coupled to those of the brain of the patient at both the pre-craniectomy and post-craniectomy periods.

2.4. Different strain measurements of axonal fiber tracts

From the obtained displacement field, U, a quantitative descrip-tion of the tissue deformation induced by TBI at the pre-craniec-tomy stage and due to DC at the post-craniectomy stage in theentire brain can be derived in the form of a Lagrangian finite straintensor20 according to the following (Equation 1):

E ¼ 12ðFTF� IÞ ð1Þ

Where E is the strain tensor, F is the deformation gradient de-fined as F = I + grad(U), FT is the transposition of F and I is the iden-tity tensor. Once the strain tensor has been determined, theinformation must be summarized efficiently in an understandableformat because there are six unique tensor values for each voxel inthe image. For this purpose, the first principal strain, the axonalstrain and the axonal effective shear strain can be derived.20

The first principal strain, nI, is the maximum eigenvalue of thestrain tensor, E, together with its corresponding eigenvector repre-senting the orientation and the maximum extent to which thebrain tissue is stretched.

The axonal strain is obtained by projecting the strain tensoralong the axonal direction according to the following (Equation 2):

E11 ¼ N1 � EN1 ð2Þ

where N1 is the axonal direction.Similarly, two shear strains that are perpendicular to the axonal

axis are defined as shown in Equation 3:

E12 ¼ N2 � EN1; E13 ¼ N3 � EN1 ð3Þ

where N2 and N3 are the two axes perpendicular to the axonaldirection.

From this, we define the axonal effective shear strain as follows(Equation 4):

Eeff shear ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiE2

12 þ E213

qð4Þ

A sketch illustrates the above different strain measurementsshowing an axon subjected to a uniaxial (stretched in one directiononly) stretch (Fig. 2). The maximum stretching direction is to theright, which corresponds to the first principal strain direction.The axonal strain, E11, is oriented along the axon, and the two shear

Fig. 2. A sketch illustrating different strain measurements showing an axonsubjected to a uniaxial tensile stretch. Before stretching (left). After stretching(right). nI, nII and nIII is the first, second, and third principal strain respectively. E11 isthe axonal strain representing the stretching along the axonal direction, while E12

and E13 are two shear strains perpendicular to the axonal direction. N1 is the axonaldirection, and N2 and N3 are the two axes perpendicular to the axonal direction.

strains, E12 and E13, are perpendicular to the axons. Eeff_shear is acombination of E12 and E13 and can be analyzed as an effectivestrain that has the consequence of ‘‘tearing’’ the axons. If the axoncoincidently lies in the stretching direction, the axonal strain be-comes identical to the first principal strain (Fig. 2).

3. Results

3.1. First principal strain and displacement magnitude of the axonalfiber tracts

The first principal strain and displacement magnitude of the ax-onal fibers for both the pre-craniectomy and post-craniectomystages are presented in Fig. 3. At the pre-craniectomy stage, the fi-bers surrounding the ventricles were most severely distorted dueto the compression of the ventricles, and the strain level furtheraway from the ventricle was less pronounced (Fig. 3, left column,lower). The strain level increased at the post-craniectomy stagein the vicinity of the skull edges where the DC surgery was per-formed (Fig. 3, middle column, lower). The graphs show the dis-placement (Fig. 3, right column, upper) and the first principalstrain (Fig. 3, right column, lower) of the fibers for both the pre-craniectomy and post-craniectomy periods. The greatest axonal fi-ber displacement of up to 12 mm was found in the treated regionof the craniectomy (Fig. 3, right column, upper, red [grey] line). Atthe post-craniectomy stage, the average first principal strain atthese axonal fibers was approximately 0.3, with a maximum valueof approximately 0.49 (Fig. 3, right column, lower red [grey] line),which increased significantly compared with the pre-craniectomystage values (Fig. 3, right column, lower black line).

3.2. Axonal strain and axonal shear strain

Different strain measures are presented for pre-craniectomystage (Supplementary Fig. 3, upper row) and post-craniectomystage (Supplementary Fig. 3, lower row).The axonal strain, E11,and axonal effective shear strain, Eeff_shear, at the axonal fibers aredemonstrated. The first principal strain is also shown with thesame view to provide a general overview of all three strain mea-sures. For the pre-craniectomy stage, the axonal effective shearstrain (Supplementary Fig. 3, right column), Eeff_shear, shows a sim-ilar pattern as the first principal strain (Supplementary Fig. 3, leftcolumn), though with a smaller strain level in general. A negativeaxonal strain value indicates that the axonal fiber is compressed(Supplementary Fig. 3, middle column).

The evaluation of axonal stretching over the treated region ofthe craniectomy is of great interest because axonal stretching isthought to contribute to an unfavorable outcome for patients.1

Thus, the strain levels at two representative regions of the axonalfibers in the treated area are plotted to illustrate more clearlyhow these unique strain measurements differ from each other(Supplementary Fig. 4). Region 1 was selected at the locationwhere the axonal fibers are close to the skull edge, while region2 is located at a greater distance from the skull edge. The first prin-cipal strain is always the largest value among all strain measure-ments, as should be expected. Generally, the axonal effectiveshear strain is smaller than the axonal strain. However, at some fi-ber points, the axonal effective shear strain is indeed larger thanthe axonal strain.

The maximum first principal strain was measured as 0.34 at apoint in region 1, with an axonal strain of 0.28 and an axonal effec-tive shear strain of 0.12 at the same point. For region 2, the maxi-mum corresponding strain levels were 0.2, 0.17 and 0.06. Theaxonal effective shear strain in region 1 was much higher than thatin region 2, representing a larger ‘‘tearing’’ effect to the axonal fi-

Fig. 3. Displacement magnitude and first principal strain at axonal fibers for pre-craniectomy (left column) and post-craniectomy stage (middle column). The axonal fibertracts were extracted from diffusion weighted images using the streamline method in open-source software Slicer 3D. The graphs show the values of displacement magnitude(right column, upper) and first principal strain (right column, lower) at the fiber points over the region of the treatment at both pre- (right column, black line) and post-craniectomy stage (right column, red (grey) line).

512 X. Li et al. / Journal of Clinical Neuroscience 20 (2013) 509–513

bers due to the proximity of the skull edge. In region 2, the axonalstrain was close to the first principal strain, which indicates thatthe axonal orientation at these points is similar to the directionof maximum stretch (i.e., first principal strain direction).

4. Discussion

In this study, we present a new method to quantify and visual-ize axonal fiber tract deformation following DC. The deformation ofaxonal fiber tracts in the form of the first principal strain (nI), axo-nal strain (E11) and axonal effective shear strain (Eeff_shear) werequantified at both the pre-craniectomy and post-craniectomystages following DC. E11 represents the stretch along the axonal ori-entation, and Eeff_shear represents a ‘‘tearing’’ effect to the axons. Thedifferences between the various types of strain measurements atthe same fiber point (Supplementary Fig. 4) clearly demonstratethe importance of incorporating axonal orientation into the model,which provides additional information regarding axonal deforma-tion. Furthermore, the calculated axonal strain (E11) is comparableto the threshold obtained from in vivo axonal stretchingexperiments.

It has been reported that strain levels as low as 0.05 will alterneuronal function, while a strain level higher than 0.20 inducessignificant levels of cell injury in vitro.21 Most of the previous axo-nal injury models use dynamic stretches that are related to rapiddeformation of brain tissue during impact. This is different fromthe type of axonal stretching experienced during the post-craniec-tomy stage, where the axons endure slow dynamic events similarto a quasi-static stretching state. In our model, we found an axonalstretch at the treated hemisphere with an average value of approx-imately 0.30 and a maximum value of 0.49 that was localized closeto the skull edge of the DC (Fig. 3). Using a model of a sciatic nervestretch, Fowler et al.22 reported that even minimal tension, if main-tained for a significant amount of time, may result in a loss of neu-ronal function. Hence, it should also be expected that the centralnervous system will sustain potential damage under a long-dura-tion stretch such as in the post-craniectomy stage, though with a

different threshold level. Additionally, it should be stressed that,as seen in the patient in this study, the displacement and stretch-ing of brain tissue lasts for several days, which may further jeopar-dize the metabolism of the patient.

DC involves the removal of a flap of skull bone with a circular tooval shape (diameter, 11–12 cm), and smaller craniectomies are re-ported to be predisposed to shearing-associated bleeding at theskull edge,23,24 which in turn causes a higher mortality rate.23 Thus,a sufficiently large craniectomy has been suggested to be impor-tant for improved neurological function.24 The axonal effectiveshear strain quantified in this study is greater in regions close tothe craniectomy skull edge than in regions further away from thecraniectomy site (Supplementary Fig. 4). Other strain levels (i.e.,first principal strain, axonal strain) also had their maximum valuesclose to the craniectomy skull edge. Brain tissue herniation hasbeen reported to be predominately located in regions close to thecraniectomy skull edge.3,5,6 Considering this, it seems plausiblethat a high strain level is related to brain tissue herniation. Follow-ing further development of the technology, it should be possible tojudge the outcome of the strain levels before the DC is performed.This may have the possibility of optimizing the size as well as thearea of the craniectomy.

Pre-craniectomy and post-craniectomy CT scans were availableand strain level at the post-craniectomy stage could be readily cal-culated by morphing the pre-craniectomy to the post-craniectomyimages. The strain level at the pre-craniectomy stage is of equalimportance as this will not only provide information regardingbrain tissue deformation at this stage, but also will give compara-ble information with respect to the strain level at the post-craniec-tomy stage. However, quantification of the strain level at the pre-craniectomy stage requires a normal image of the patient beforeTBI, which is not available in most cases. It seems, however, thatthere is no real consensus on the right method to use to recovera healthy brain image. In this study, we propose an approach thatmorphs the MRI from a healthy similar-age volunteer to the CTscan of the patient according to the cranial shape. The assumptionunderlying this approach is that all normal brains, at least at a cer-

X. Li et al. / Journal of Clinical Neuroscience 20 (2013) 509–513 513

tain level of representation, have a similar topological structure,but may differ in shape details. We chose to use an MRI from ahealthy volunteer of a similar age because normal aging can resultin ventricular enlargement.25 The proposed approach has beenshown to give very promising results (Supplementary Fig. 1, leftcolumn, upper row, the recovered healthy brain of the patient).However, the recovery of a healthy brain image of a specific patientis still an open problem and there is no validated registration algo-rithm considered as gold-standard. Further studies across a broad-er range of TBI patients are needed to confirm the validity of thisapproach.

Axonal fiber tracts were included in the patient model by morp-hing the DWI from a healthy brain image into the data for the brainof this specific patient by applying the displacement field contain-ing the inter-subject structural differences. Different methods havebeen proposed in the literature for applying a displacement field toa DWI, we choose to use the ‘‘preservation of principal compo-nents’’ method, which has been suggested for multi-subject non-ri-gid registration of soft tissue.26 DWI for the same patient,whenever available, should be used to represent the fiber tractsof that particular patient; nevertheless, where they are not avail-able, the conclusions drawn from the results of the axonal strainsshould still apply using this method.

There is a lack of studies on sustained axonal stretching andshear deformation under static loading. Therefore, the manner inwhich the strain levels of the axonal fiber tracts found in this studymight affect neural function requires further study. Although it isoften impossible to avoid DC for some seriously injured patients,awareness and a better understanding of the potential dangers ofthe craniectomy should help to prevent further damage to thebrain.

5. Conclusion

This study provides the first quantitative data on the stretchingof axonal fiber tracts caused by DC. The first principal strain, axonalstrain and axonal effective shear strain were quantified in axonalfiber tracts for both the pre-craniectomy and post-craniectomystages. The results show that DC causes a significant strain increasein axonal fiber tracts. Using the present methodology, the deforma-tion (stretching and shearing) of axonal fiber tracts of interest canbe visualized and may therefore serve as a prognostic tool in clin-ical practice on the cognitive and neurological sequelae of patientsundergoing DC.

Conflicts of Interest/Disclosures

The authors declare that they have no financial or other con-flicts of interest in relation to this research and its publication.

Acknowledgements

This study was supported by Swedish Research Council D.nr.621-2008-3400, the Swedish Governmental Agency for Innovationsystems (VINNOVA), and the Chinese Council Scholarship (CSC) forthe first author. We thank Professor Tieqiang Li (Karolinska Insti-tute, Stockholm, Sweden) for providing the diffusion-weightedimages and the MRI in this study.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.jocn.2012.04.019.

References

1. Cooper DJ, Rosenfeld JV, Murray L, et al. Decompressive craniectomy in diffusetraumatic brain injury. N Engl J Med 2011;364:1493–502.

2. Timofeev I, Kirkpatrick PJ, Corteen E, et al. Decompressive craniectomy intraumatic brain injury: outcome following protocol-driven therapy. ActaNeurochir Suppl 2006;96:11–6.

3. Winter C, Adamides A, Rosenfeld J. The role of decompressive craniectomy inthe management of traumatic brain injury: a critical review. J Clin Neurosci2005;12:619–23.

4. Vashu R, Sohail A. Decompressive craniectomy is indispensible in themanagement of severe traumatic brain injury. Acta Neurochir2011;153:2065–6.

5. Honeybul S. Complications of decompressive craniectomy for head injury. J ClinNeurosci 2010;17:430–5.

6. Yang X, Wen L, Shen F, et al. Surgical complications secondary to decompressivecraniectomy in patients with a head injury: a series of 108 consecutive cases.Acta Neurochir 2008;150:1241–8.

7. Servadei F. Clinical value of decompressive craniectomy. N Engl J Med2011;364:1558–9.

8. Timmons SD, Ullman JS, Eisenberg HM, et al. Craniectomy in diffuse traumaticbrain injury. N Engl J Med 2011;365:373–6.

9. Joseph R. Neuropsychiatry, neuropsychology, and clinical neuroscience. emotion,evolution, cognition, language, memory, brain damage, and abnormalbehavior. Philadelphia: Williams & Wilkins; 1996.

10. Chung RS, Staal JA, McCormack GH, et al. Mild axonal stretch injury in vitroinduces a progressive series of neurofilament alterations ultimately leading todelayed axotomy. J Neurotrauma 2005;22:1081–91.

11. Tang-Schomer MD, Patel AR, Baas PW, et al. Mechanical breaking ofmicrotubules in axons during dynamic stretch injury underlies delayedelasticity, microtubule disassembly, and axon degeneration. FASEB J2010;24:1401–10.

12. Smith DH, Wolf JA, Lusardi TA, et al. High tolerance and delayed elasticresponse of cultured axons to dynamic stretch injury. J Neurosci1999;19:4263–9.

13. Bain AC, Meaney DF. Tissue-level thresholds for axonal damage in anexperimental model of central nervous system white matter injury. J BiomechEng 2000;122:615–22.

14. Li X. Finite element and neuroimaging techniques to improve decision-making inclinical neuroscience. Royal Institute of Technology (KTH); 2012.

15. Chatelin S, Deck C, Renard F, et al. Computation of axonal elongation in headtrauma finite element simulation. J Mech Behav Biomed Mater 2011;4:1905–19.

16. Vercauteren T, Pennec X, Perchant A, et al. Diffeomorphic demons: efficientnon-parametric image registration. Neuroimage 2009;45:S61–72.

17. Pieper S, Halle M, Kikinis R, et al. 3D slicer. biomedical imaging: nano to macro,2004. IEEE Int Symp 2004;1:632–5.

18. Basser PJ, Pajevic S, Pierpaoli C, et al. In vivo fiber tractography using DT-MRIdata. Magn Reson Med 2000;44:625–32.

19. Mori S, Crain BJ, Chacko V, et al. Three-dimensional tracking of axonalprojections in the brain by magnetic resonance imaging. Ann Neurol1999;45:265–9.

20. Holzapfel GA. Nonlinear solid mechanics: a continuum approach forengineering. West Sussex, England: John Wiley & Sons, Ltd.; 2000.

21. Morrison 3rd B, Cater HL, Wang C, et al. A tissue level tolerance criterion forliving brain developed with an in vitro model of traumatic mechanical loading.Stapp Car Crash J 2003;47:93–105.

22. Fowler SS, Leonetti JP, Banich JC, et al. Duration of neuronal stretch correlateswith functional loss. Otolaryngol Head Neck Surg 2001;124:641–4.

23. Wagner S, Schnippering H, Aschoff A, et al. Suboptimum hemicraniectomy as acause of additional cerebral lesions in patients with malignant infarction of themiddle cerebral artery. J Neurosurg 2001;94:693–6.

24. Wirtz CR, Steiner T, Aschoff A, et al. Hemicraniectomy with duralaugmentation in medically uncontrollable hemispheric infarction. NeurosurgFocus 1997;2:E3.

25. Hill DLG, Hajnal JV, Rueckert D, et al. A dynamic brain atlasProceedings of the 5thInternational Conference on medical image computing and computer-assistedintervention-Part I. MICCAI ’02. London, UK. UK: Springer-Verlag; 2002. p. 532–9.

26. Alexander DC, Pierpaoli C, Basser PJ, et al. Spatial transformations of diffusiontensor magnetic resonance images. IEEE Trans Med Imaging 2001;20:1131–9.