Seminar Abstract
“Diffusion MRI Brain Atlas at 3.0T”
Quantitative measurements of brain diffusion parameters (apparent diffusion coefficient [ADC] and fractional anisotropy [FA]) provide new insight compared to conventional MRI. Underlying structural changes at the cellular level can give rise to changes in diffusion parameters that is observable at image spatial resolutions. A standard set of diffusion parametric values (and their spatial distribution) provides a basis for identifying and monitoring abnormality in a given patient relative to the reference. However, the inherent normal variations present in normal human brains will complicate distinguishing abnormalities from normal variants. Clearly, a single brain cannot accurately represent the parameter variance in the normal population. Probabilistic atlas methods, addresses this challenge by providing methods to capture the population variability. On the other hand, intensity-based models are focusing on reconstructing an average representation of structural anatomy by averaging over multiple MRI scans.
In this work we utilize affine and free form transformations to successfully create a minimally distorted average morphometric atlas of diffusion parameters (FA, ADC) acquired at 3.0 T. This atlas converges towards the centroid of population data set and reflects the variability in diffusion parameters for the ten subjects on a voxel-by-voxel basis. The atlas can be utilized as a tool that characterizes morphological or functional information in order to distinguish abnormal anatomical and functional variations. A pre-requisite for atlas creation is the acquisition of diffusion weighted MR images that are free from geometric distortions. We have integrated parallel imaging methods to collect high-resolution, near-isotropic voxel images with whole brain coverage at 3T with relatively high SNR. Parallel imaging methods reduces geometric distortions, but at 3T still require post-processing to further reduce these artifacts. As a part of this work, we have also developed and implemented an algorithm that successfully corrects this type of distortions in MR images at high magnetic fields. Average shape and parametric atlases constructed from these distortion free diffusion weighted images will be presented as well as some preliminary work on compact representations of shape and parameter (ADC and FA) variability within the population based on Active Appearance Models.
