4 FLASH vs. EPI
What’s the difference between the two most common pulse sequences? This first section is a quick summary of the key difference: a FLASH pulse sequence uses a new excitation pulse for each line of k-space, while an EPI pulse sequence covers many lines after a single excitation. [This mini-lecture acts like EPI always acquires all of k-space for a single slice with a single excitation; that’s a single-shot 2D EPI acquisition. It’s possible to cover segments of k-space and add them together later, when you need high resolution and a short echo time; that’s segmented EPI. It’s also possible to cover a 3D volume with an EPI read-out; then your TR is relatively short and you’re doing k-space blips on 2 axes instead of just 1 … but that’s for later!]
Exercises
- In order to acquire a single image or slice with a 64 x 64 matrix (and no undersampling or acceleration), how many excitation pulses are used for a FLASH acquisition? EPI?
- In the same hypothetical, fully-sampled 64 x 64 FLASH and EPI images, if the TE is 20 ms, what is the minimum TR for FLASH? For EPI?
- In the same hypothetical, fully-sampled 64 x 64 FLASH and EPI images with TE=20 ms and TR set to the minimum possible value, that is the total image read-out time for each?
The different k-space trajectories have consequences for image quality. In general, FLASH is cleaner because spins have very little time down on the transverse plane to move away from their intended phase due to chemical shift artifacts (protons on fat actually resonate at a different frequence than protons on water, and the longer you let that go on, the more the signal gets displaced in your image) or unwanted field imperfections in your image (my right ear isn’t actually resonating at the frequency I told it to resonate at, because my head perturbs the magnetic field).
Exercises
Consider the same hypothetical, fully-sampled FLASH and EPI images above with TE = 20 ms. Now, we’ll add in the fact that data were digitized at a rate of 1 data point every 100 microseconds.
4. How long does it take to read out one line (the answer is the same for the FLASH and EPI because resolution and sampling rate are hypothetically the same)?
5. How would you express the sampling rate in terms of Hz?
6. We use the term “pixel bandwidth” to describe the frequency resolution of our image. For a FLASH image, this is computed as the inverse of the time that it takes to read out one line after excitation. What is the “pixel bandwidth” of the hypothetical FLASH image?
7. If your image is perfectly shimmed (you lucky person!) but there’s one air bubble in there that is perturbing the magnetic field in one voxel so that the local magnetic field is 0.00001 T higher than you expected, what is the frequency shift in this voxel?
8. In your hypothetical FLASH image, signal from this pixel won’t show up in the right place. Will it be shifted in the read-out or phase-encode direction? How far will it be shifted?
9. Why does the signal only shift in one direction in a FLASH image?
10. In your hypothetical EPI image, what is the “frequency resolution” of each voxel in the read-out direction? In the phase-encode direction?
11. Where in your image should you look to find the signal from this off-resonance voxel — how var over in the read-out direction, and how far up in the phase-encode direction?
Answer key
You can check your answers against these.