20 healthy controls, 20 TBI patients, and 20 patients with non-traumatic microbleeds underwent our imaging protocol consisting of Magnetization Prepared Rapid Gradient Echo (MPRAGE), (repetition time (TR) = 2530 ms; echo time (TE) = 3.37 ms; Field of View (FoV) of 256 mm; slice thickness 1 mm); Fluid Attenuation Inversion Recovery (FLAIR) (TR = 5000 ms; TE = 387 ms; FoV = 230 mm; slice thickness 0.9mm); Susceptibility Weighted Imaging (SWI) (TR = 27 ms; TE = 20.0 ms; FoV=220mm, Slice thickness: 1,50 mm) and DTI (TR= 7700ms; TE= 68 ms; FoV= 256mm; Slice thickness: 2mm. There were no major scanning protocol upgrades during the project, furthermore no anesthetics were used during imaging. Imaging was carried out by a Siemens MAGNETOM Prisma 3 Tesla MRI-machine. Age-matched triplets (+/-3 years) of the TBI and microbleeds positive (traumatic-microbleeds-TMB), TBI-negative incidental microbleeds positíve (nontraumatic-microbleeds-nTMB) group and control group were created. Inclusion criteria for the TMB group included age above 18, negative history of neurological, as well as psychiatric pathologies, and any comorbidities that induce MBs in white matter. Another inclusion criterion for the TMB group was a severity of moderate-severe, according to the Mayo classification system. The Mayo Classification System for TBI Severity was developed to classify cases based on available indicators that included death due to TBI, trauma-related neuroimaging abnormalities, GCS, post traumatic amnesia, loss of consciousness and specified post-concussive symptoms. For the nTMB group, where MBs were found incidentally, patients with a history of head trauma were excluded.

In this study modules of the FMRIB's Diffusion Toolbox (FDT) software (FSL v6.0, developed by Oxford University) were used. The raw DICOM images, used in modern medical imaging, were converted into NIFTI format files using a “dcm2niix”-converter. Data was stored in accordance with the General Data Protection Regulation (GDPR) after anonymisation. [8]

Two correction algorithms in the FSL FDT module; eddy correct and topup eliminate various distortion-generating factors on DTI images. After pre-processing FA, MD, AD and RD diffusion maps were calculated using FSL’s DTIFIT, which fits a diffusion tensor model to each voxel. [9-11] Calculation used in probabilistic tractography utilizes the FSL Bayesian Estimation of Diffusion Parameters Obtained using Sampling Techniques (BEDPOSTX) module to determine the proportion of crossing fibers. Subsequently we used PROBTRACKX to perform tractography analysis. [11] These 3D binarized masks were created using the FSL maths program. In addition a comprehensive "Whole-White-Matter mask" was reconstructed by combining these tract-level masks. We established the highest threshold where we observed continuous tracts across all 3D tract masks for each patient we studied.  Currently we can automatically segment 42 tracts using XTRACT.

Fig 1: FIGURE shows the reconstructed forceps minor (blue), major (red), radiatio thalamy anterior (yellow) and fasciculus longitudinalis superior (green) coronal, horizontal planes shown in Fsleyes.

Fig 2: FIGURE shows the reconstructed forceps minor (blue), major (red), radiatio thalamy anterior (yellow) and fasciculus longitudinalis superior (green) in Fsleyes 3d extension.

On raw full head images non-brain tissues were removed. This was performed using the FSL Brain Extraction Tool (BET) module. [12] Linear co-registration of MPRAGE and SWI-images was performed in order to precisely locate MBs within white matter. This was achieved by the FSL - FMRIB's Linear Image Registration Tool (FLIRT)[13]. The central coordinates of MBs located in white matter were recorded and their anatomical localisation was confirmed in coronal, sagittal and axial planes by using MPRAGE images. Identical contralateral coordinates of the MBs in NAWM were also assessed using the FSL Nudge software.

Fig 3: FIGURE show the located MB ont he SWI images and the cooregistered MPRAGE images in the bottom

The coordinates of lesions detected within white matter and coordinates of their identical contralateral loci were then translated to the anatomically corresponding localisation of the control group’s recordings. This was performed by means of nonlinear registration using FMRIB's Non-linear Image Registration Tool (FNIRT) module of the FSL program. [14] We then performed a co-registration between the SWI and FLAIR images using FSL's FLIRT module to exclude MBs co-localized with edema. The functional diffusion maps were also coregistered with the matrix of structural images.

In the following step a total of 9 spherical shells were created surrounding the central, contralateral identical and translated coordinates ranging from 2mm to 20mm according to the 2*2*2mm voxels used in DTI, with a total of 9 spherical shells per 2mm.

To ensure accuracy of the diffusion parameters in perilesional white matter regions, statistical analysis was performed on intersecting voxels of the whole white matter mask and spherical shells created around coordinates of the SWI lesions. Intensity-based *-SD* thresholding was performed on the SWI images to eliminate the formation of biases, due to MBs and other pathologies causing hypointensities, in the measured diffusion parameters.

The evaluation above was performed using a semi-automatic in-house developed software

Fig 4: FIGURE show the reconstructed spherical shells around the MBs from 2 mm up to 2 cm, the intersecting voxels of the whole white matter mask and spherical shells, in 3D extension.

Fig 5: FIGURE show the reconstructed spherical shells around the MBs from 2 mm up to 2 cm, on the top the intersecting voxels of the whole white matter mask and spherical shells.