ABSTRACT: The purpose of this study is to utilize a Monte-Carlo Simulation (MCS) to study the behaviour of water diffusion within the white matter to have a better understanding of the ultrahigh-b DWI (UHb-DWI) signal-b of spinal cord white matter. Recently, UHb-DWI method presented a novel biomarker with very high specificity to demyelination in the cervical spinal cord. It is an advanced version of diffusion-tensor MRI (DTI) and diffusion-kurtosis MRI (DKI) with greatly simplified data acquisition. While conventional DTI relies on the water diffusion within the extra-axonal (EA) space, UHb-DWI provides information about water exchange across the myelin sheath as well as intra-axonal (IA) and EA spaces. The MCS of the water diffusion within the one-dimensional white-matter, such as spinal cord and optic nerve, has been very helpful to understand the deep microscopical environment with different pathology in active lesion of the cervical spinal cord (CSC) in patients with multiple sclerosis (MS). MCS experiment was performed with various degrees of demyelination and axonal loss. The measured UHb-DWI signal from the active lesion and normal-appearing white-matter region MS CSC were matched to the signal-b curves, which were numerically synthesized for various fractions of demyelination and axonal loss, to search for the best matching curve. Although we present the date from the cervical spinal cord of MS CSC, the method can be applied to other white-matter diseases, such as amyotrophic lateral sclerosis (ALS) and degenerative cervical myelopathy (DCM) as well. DATA DESCRIPTION: There are two raw data sets: (1) diffusion-weighted MRI data and (2) Numerical Monte-Carlo Simulation (MCS) data. Raw diffusion MRI data were collected with 8 signal averages, of which individual average image is stored in its own DICOM file, therefore one needs to average over the average count before further analysis. Each data set contains two folders: (a) UHbDWI_b0 for estimating T1 and (b) UHbDWI_bMax_7384, UHb-DWI data measured with the maximum b of 7384 s/mm2 along direction perpendicular to the nerve direction. MCS data can be processed using the included processing software, written in Python 3.x.Signal-b curve can be generated using these MCS data and software for diffusion gradient applied along angle from the fiber direction, for instance 0o for axial DWI and 90o for radial DWI.
Steps to reproduce
Data set 1: MRIData The data was collected using 2D single shot diffusion-weighted stimulated-echo-planar-imaging with reduced-FOV (2D ss-DWSTEPI-rFOV). Because we target on about 1 cm^2 cross-section of spinal cord, it is important to use a high-sensitivity RF coil. For these data, we used a home-developed 8 channel cervical-spinal-cord RF coil, which is dedicated only for CSC. The diffusion gradient was applied perpendicular to the spinal cord, with maximum b of 7834 s/mm^2. To achieve such high b-value at clinical MRI system, of which the maximum gradient strength was only 40 mT/m, we used 2D ss-DWSTEPI-rFOV with varying diffusion time, i.e., mixing time TM, up to 450 ms. However, because TM was varied, measured DW images experience different T1 decay during the mixing time, therefore, it is crucial to remove T1 decay factor from all DW images. Thus, we measured another set of 2D ss-DWSTEPI-rFOV data with b=0, from which T1 map was constructed by fitting signal-time curve to a single-exponential function in pixel-by-pixel, and used to remove T1 decay factor from each DWI. Then the resultant DW images are ready for further UHb-DWI analysis. Data set 2: MCSimulData These numerical data were generated for 10000 water molecules on an 80x80 μm2 pixel with 3199 axons, which were randomly distributed. Myelin layers of a fraction of axons, which were randomly selected, were completely peeled off to mimic demyelination. The water molecules made random hopping every 1 μs and their position vectors were saved every 100 μs, for 250 ms total diffusion duration.