Microinfarct Disruption of White Matter Structure:

A Longitudinal Diffusion Tensor Analysis

E Auriel, MD, MSc; BL Edlow, MD; YD Reijmer, PhD; P Fotiadis, BSc; S Martinez-Ramirez, MD; J Ni, MD; AK Reed, BA: A Vashkevich, BA; K Schwab, BA;

J Rosand, MD, MSc; A Viswanathan MD, PhD; O Wu, PhD; ME Gurol MD, MSc;

SM Greenberg MD, PhD

Supplementary Methods

Coregistration

The following steps were performed toidentify the DWIlesion ROI and corresponding control ROI at each time point (see also Fig. e-1):1) placement of a ROIcontaining each DWI lesion using FSLView from the FMRIB Software Library (FSL)1; 2) skull stripping of the non-diffusion weighted, b=0 sec/mm2 (b0) and T1-weighted MEMPRAGE root-mean-square (T1) datasets using FSL’s Brain Extraction Tool (BET); 3) linear registration of the b0 to the T1 dataset using FMRIB’s Linear Imager Registration Tool (FLIRT)version 5.51 with affine transformation and mutual information cost function (matrix 1); 4) linear registration of the T1 datasetto the Montreal Neurological Institute (MNI) 152 T1 2mm voxel template (MNI152, distributed with FSL) using FLIRT with affine transformation and correlation ratio cost function (matrix 2); 5) creation of a non-linear warp (warp 1) from T1 space to MNI152 space using FMRIB’s Non-linear Imager Registration Tool (FNIRT) with matrix 2 used as the preliminary linear matrix for the FNIRT command; 6) non-linear coregistration of b0 to MNI152 using FNIRT, using warp 1 as the nonlinear warp andmatrix 1 as the preliminary linear matrix; 7) application of non-linear transformation (using matrix 1 and warp 1) to coregister the lesion ROI from b0 space into MNI152 space using FNIRT; 8) generation of a control ROI in the contralateral hemisphere in MNI152 space using the fslswapdimfunction with “-x y z” parameters, thereby placing the control ROI in the same neuroanatomic location as the lesion within the contralateral hemisphere; 9) transformation of the control ROI from MNI152 space to T1 space by inverting the non-linear warp (inverted warp 1) and using FNIRT with inverted warp 1; and 10) transformation of the control ROI from T1 space to b0 space by inverting linear matrix 1 (inverted matrix 1) and using FLIRT with inverted matrix 1 and a mutual information cost function.Precise localization of the control ROIs in native b0 space was confirmed by comparing the location of the lesion ROI and the control ROI on the color-coded fiber orientation maps using a standard color DTI atlas2 and by comparing the trajectories of the associated WM bundles using a standard tractography atlas. 3Manual corrections of the control ROI were performed when necessary by consensus of three investigators (EA, BE, PF) without reference to quantitative FA or MD data to minimize bias.

The following procedure was used for coregistration of lesion and control ROIs from the lesion-detection scan to the pre- and post-lesional scans from the same subject: 1) the pre-lesional, lesional, and post-lesional b0 datasets were skullstripped using BET;2) the skulled-stripped lesion-detectionb0 dataset was coregistered tothe skull-stripped pre-lesional and post-lesional b0 dataset using FLIRT with affine transformation and the output registration matrix then used to linearlytransform the lesion and control ROIs to the skulled strippedpre-lesional and post-lesional b0 datasets;3) after linear coregistration was complete, FNIRT was used to non-linearlycoregister the lesion-detectionb0 dataset to the pre-lesional and post-lesional b0 datasets(non-skull-stripped scans were used in this non-linear step);4) the command applywarp wasused to apply the warps generated by FNIRT in step 3 to transform the lesion and control ROIs from the lesional b0 dataset to the pre- and post-lesional b0 datasets.

Diffusion tractography analysis

Diffusion data were corrected for subject motion and eddy current distortions with ExploreDTI ( Fiber tracking was performed based on constrained spherical deconvolution{Tournier, 2007 #14644}using an ROI approach to delineate the fiber tracts. 7 The applied termination criteria for fiber tracts were 1) a fiber orientation distribution (FOD) value of 0.1 (the harmonic degree of the estimated FOD coefficients was limited to 4); and 2) an intervoxel angle of curvature > 45 degrees.
References

1. Smith SM, Jenkinson M, Woolrich MW, et al. Advances in functional and structural MR image analysis and implementation as FSL. Neuroimage. 2004;23 Suppl 1:S208–219.

2. Oishi K. MRI Atlas of Human White Matter. Second edition. Amsterdam: Elsevier; 2011.

3. Catani M, Thiebaut de Schotten M. A diffusion tensor imaging tractography atlas for virtual in vivo dissections. Cortex. 2008 Sep;44(8):1105–32.

4. Mac Donald CL, Johnson AM, Cooper D, et al. Detection of blast-related traumatic brain injury in U.S. military personnel. N Engl J Med. 2011 Jun 2;364(22):2091–100.

5. Leemans A, Jones DK. The B-matrix must be rotated when correcting for subject motion in DTI data. Magn Reson Med. 2009 Jun;61(6):1336–49.

6. Jeurissen B, Leemans A, Jones DK, Tournier J-D, Sijbers J. Probabilistic fiber tracking using the residual bootstrap with constrained spherical deconvolution. Hum Brain Mapp. 2011 Mar;32(3):461–79.

7. Reijmer YD, Leemans A, Heringa SM, et al. Improved sensitivity to cerebral white matter abnormalities in Alzheimer’s disease with spherical deconvolution based tractography. PLoS ONE. 2012;7(8):e44074. 1. Smith SM. Fast robust automated brain extraction. Hum Brain Mapp. 2002;17(3):143–55.

Figure e-1. Coregistration of DWI hyperintense lesion to corresponding contralateral ROI and to longitudinal images.

The coregistration steps outlined in the figure are briefly summarized as follows (see Supplementary Methods for details): 1) coregistration of native b0 to native T1; 2) coregistration of native T1 to MNI space; 3) non-linear transformation of lesion (solid circle) from native b0 to MNI; 4) creation of contralateral control ROI (dotted circle) in MNI space; 5) transformation of control ROI from MNI to native T1; 6) transformation of control ROI from native T1 to native b0; 7a and 7b) transformation of lesion and control ROIs from lesional b0 dataset to pre-lesional b0 dataset; and 8a and 8b)transformation of lesion and control ROIs from lesional b0 dataset to post-lesional b0 dataset.

Figure e-1