Wednesday, October 17, 2012

Slice vs. Slab (3D Time-of-Flight MRA)

"2D" TOF imaging isn't strictly two-dimensional. It's certainly not 1 proton thick, and even if it were,  stacking that many slices on top of one another to build an MR angiogram would take till the end of time   "2D" in 2D TOF is actually a very thin (~1-3 mm) three-dimensional space, represented in two-dimensional k-space, which for all intents and purposes is two-dimensional...

... but what if we were to open up that really thin volume into a bigger volume? Turn that slice into a slab with a gradient echo acquisition of a volume? Well... if you do that, then you've got a 3D Time of Flight MR angiogram...


3D TOF source image of the intracranial arteries on the left.  On the right is a MIP reconstruction of the posterior circulation from that original sequence.  The visibility of the small tortuous vessels in a less-than-perpendicular orientation to the plane of imaging is a strength of the 3D technique.


Why bother with this technique?  It has a couple of advantages over 2D TOF, including the better signal-to-noise and better evaluation of tortuous vessels inherent in a 3D technique.  Speed is variable depending on the amount of resolution you want or need...the speed of the 3D TOF sequence is directly proportional to the number of partitions in the slab (and therefore the number phase-encoding steps)... more partitions = smaller voxels = higher spatial resolution = longer time.

So... in considering how 2D TOF works (see the 10/14/2012 post), how is it that a flowing blood proton bombarded with RF pulses at the bottom of the slab still has signal at the top?  Shouldn't it get saturated and lose signal?  Isn't TOF inherently limited to thin slices?

Blood becomes more saturated as it moves through the slab.
The answer is yes: everything that hold true for 2D TOF holds true for 3D TOF. The flowing protons entering the slab build up transverse magnetization and decrease in signal intensity as they move toward the other side of the slab. The only way 3D TOF can really work is if the flowing protons move at high velocity, so that this effect is minimized.

For this reason, 3D TOF is much less sensitive to slow flow than 2D TOF, and 3D TOF is better for higher velocity arteries, such as the circle of Willis (above) or the aorta. 2D TOF is better for systems with slow flow (such as the peripheral extremities or abdominal vessels)





Theoretical schematic of increasing flip angles through the slab
In addition, other parameters are optimized to try to account for the saturation through the slab. For instance, flip angles tend to be smaller with 3D TOF than with 2D TOF (see the 10/14/2012 post). A smaller flip angle means less build-up of transverse magnetization, so less saturation.  The trade-off is that the background is less saturated as well.

3D TOF can also be combined with gadolinium contrast agents, which would selectively shorten the T1 relaxation time of the blood and allow the larger flip angles..... but the benefit of a non-contrast study is lost.

Some techniques vary the flip angle through the slab, with a smaller flip angle early in the slab to prevent too much saturation, and a larger flip angle toward the end of the slab, where residual magnetization is less of an issue (right).

One technique designed to address the strengths and weaknesses of both 2D and 3D technique is to overlap multiple smaller slabs, thus trying to get the higher signal-to-noise of the 3D technique while trying to keep saturation minimal (MOTSA, multiple overlapping thin-slab angiography).

Magentic transfer techniques can also be used to increase the contrast between relatively homogenous fluid (blood) and heterogeneous fluid with large macromolecules (tissue).

---
1. "MRI Principles" Mitchell DG, Cohen MS. 2nd ed (2004)
2. "Methods in Biomedical Magentic Resoance Imaging and Spectroscopy" Young, ed. (2000)
3. "Magnetic Resonance Imaging: Physical Principles and Applications" Kuperman V. (2000)
4. "The Physics of Clinical MR Taught Through Images" Runge, Nitz, Schmeets, et al. (2005)