DTI Acquisition Guide Donald Brien February 2016
Overview The basic sequence: Once or twice refocused SE EPI DTI specific parameters: b-value and directions Standard MRI parameters: TE / TR / Resolution / FOV / Parallel Imaging / Filters / Fat Sat. Orientation and Encoding Direction(s) – Pros and cons Top up and the general problems with EPI imaging (acquisition vs post-processing) Scenarios and Example Images
General MRI Acquisition Considerations There is theory, and then there is practice. Always pilot your protocol and run pilot analysis. Optimizing your study requires: – Thinking ahead. What do you need from the data? – What are the limitations? Time, participant issues – What can be mitigated through parameter setup? – What can be mitigated by post-processing? Are other scans necessary to allow this?
DWI / DTI Measurements ADCMD FA DECFA Tractography
PGSE Sequence The MR signal is attenuated via strong diffusion gradients. This presents many challenges to a successful acquisition.
DTI Challenges We purposefully start with a sequence that is starved of SNR. B-value (s/mm2) Signal
DTI Challenges Bulk motion during gradients creates unpredictable apparent diffusion. Echo-Planar Imaging (EPI) – Acquire entire k-space in one echo – Bulk motion is minimized and more consistent – Many volumes can now be acquired in a feasible time
DTI Challenges EPI Readout We read out the entire k-space in one shot But we pay a price: distortion in the phase- encoding direction
DTI Challenges
Strong diffusion gradients cause large residual eddy currents in the magnetic field These cause ‘apparent’ spatial gradients during readout
TRSE Sequence Eddy-currents can be partially mitigated through sequence
DTI Specific Parameters b-values – 1000 s/mm2 is the standard because most sensitive to changes in diffusion in the brain – Also good choice because most data is acquired at this b- value, which will allow your results to be comparable to normative datasets 9.1:1 DWI:b0 to minimize variance in tensor model # of directions – 20 directions minimum for accurate anisotropy measurements – Benefit to tensors flattens out around 30 directions – diminishing returns
Anisotropy
Tensor Estimates
Common MRI Parameters Fat. Sat. on because lipid signal interferes (off resonance chemical shift) with water diffusion signal. TE short as possible to maximize SNR and same for all measurements TR must be > 3x T1 of tissue (5s or more), but viable (10 mins or less total typically) Resolution tends to be lower to compensate for lower SNR and permit a quick readout
Other Parameters FOV: ~24cm to cover brain and prevent aliasing Slice Gap: 0% - use interleaved ordering Bandwidth: Generally higher for SNR and fast imaging, but testing important Parallel Imaging: GRAPPA for faster k-space sampling – reduce distortion vs lower SNR Partial Fourier: May introduce artifacts, but does reduce TE.
Parallel Imaging
Orientation and Encoding Directions A>>P is preferred for single scan studies – Minimizes asymmetry biases (does not eliminate) Axial or axial oblique? – Axial may reduce ghosting – Measures should be rotationally invariant (if 7+ measurements) either way – Be consistent and nearly axial (ACPC is common)
Encoding Direction
Top Up A >> P P >> A
Multishot Imaging
Scenarios 1 Grey matter diffusion (DWI) – 3 orthogonal DWIs and a b0 would be sufficient as grey matter is mostly isotropic White matter diffusion (DWI) – 6 maximally spread DWIs and a b0 for rotationally invariant measures of white matter MD Whole Brain FA measurements (DTI) – 6 minimum, but 20 would be ideal.
Scenarios 2 Pediatric DTI imaging – Motion susceptibility and time limit with small head 20 directions, 1 AP scan, 1 extra b0, maximize SNR, lower b- value due to developing brain (~700 s/mm2), SAR considerations ALS / Parkinsons imaging – Motion susceptibility and time limit 20 directions, 1 AP scan, 1 extra b0 Healthy Controls – Ideal candidates. Adult, good at staying still. 30 directions, AP and PA, 3 extra b0
Scenarios 3 NHP or other animal or small structure imaging – Very small head. Probably very little motion issues due to anesthesia directions, but can use the increased time to our advantage. More averages, more directions, less acceleration. Multishot scans. Separate k-space into many read outs to drastically decrease susceptibility problems. Many extra b0 AP and PA – susceptibility problems likely exaggerated with small FOV, strong gradients, and anatomical differences (large eyes) Investigate b-value considerations in animal
Scenarios 4 Small structure imaging in awake participants – Imaging a smaller region of interest may require higher resolution and a higher # of directions, but motion and SNR an issue Less slices, more focused scan to reduce TR Train participants on staying still outside of the scanner / Pick participants carefully Cardiac gaiting? Multiband imaging – increase number of directions without increasing time, but preserving SNR. Not available commercially yet. Long reconstruction.
Example – Sequence Choice TRSE – No PI, No pfTRSE – PI - 3Multishot - 5 Multiband – 3 Monopolar, 6/8 pf 2x2x2mm voxels – b=1000, 6 directions, A >> P TE TR TA1:451:278:0100:57
Example – Fat. Sat.
Example – Low SNR
Conclusions 1.What measures will you ultimately need? – MD, ADC, FA, Tractography? 2.What tools do you plan to use? – Resolution / SNR / directions / etc… 3.What trade-offs are you willing to make given your requirements and limitations? – Time constraints, motion issues, resolution requirements 4.What can be mitigated through parameters vs post-processing? – Parallel imaging, partial fourier, multi-shot, multiband 5.What additional scans should you acquire? – Additional encoding, T2w anatomical, b0 field? 6.The interaction of DWI / DTI parameters is complex. – Consult previous research before designing your acquisition. – Even with careful theoretical preparation, careful pilot testing is fundamental to any MRI study.