DTI learning notes

BOOK reading

Two reasons why we need to address and correct motion artifacts in DTI studies.
1) DTI-specific issue
2) Common to all quantitative MRI that requires multiple MR images such as T2 map, magnetization transfer ratio, and perfusion map.

DTI-specific issue:  degradation of diffusion-weighted images by ghosting effect.  DTI is a technique hypersensitive to motion due to its use of strong diffusion gradients.  Using a diffusion constant of normal brain tissue (D=0.8*10-3 mm2/s) and typical gradient separation (delta=30 ms), we can calculate that water molecules move about 7um on average during the measurement (7um=square_root[2Ddelta], based on Einstein's equation).

Two most notable shortcomings of single-shot EPI is the limitation in resolution and image distortion due to B0 susceptibility problems.  The single-shot EPI acquires all phase-encoded lines after one excitation.  As the echo train length gets longer, the extent of image distortion increases; i.e., the higher the resolution, the more distortion we get.  This is an inherent limitation of EPI.  The distortion is caused by magnetic field (B0) inhomogeneity, which is especially severe at the boundary of air and tissue; most notably around the sinus, such as inferior regions of the frontal lobe, the anterior pole of the temporal lobe, and the pons.  The amount of distortion increases as the field strength increases.

First, we can use a lower resolution as much as possible.  Second, there are techniques that measure the B0 field distortion, calculate a map of image distortion, and undo the distortion.  This approach requires and additional scan for the field mapping.  Therefore, it cannot be applied to existing data unless such a scan has been acquired for each study.  This approach should be able to correct low-frequency distortion (global distortion) fairly accurately, but it is a challenging task to correct high-frequency distortion (severe local distortion).  One of the most effective approaches in recent technologies is probably parallel imaging.

B0 images can be used as non distorted reference images to unwrap the distorted DWIs.  12-mode linear transformation (affine transformation) is highly effective.