MR protocols

From SBIwiki

Jump to: navigation, search

This page is intended to inform KI-KS MRC users about a few most frequently used MRI methods (structural, functional, spectroscopy and diffusion MRI).


fMRI @3T: pulse sequences & MRI coils

fmri with Gradient Echo EPI for 3T

One may find many papers with this or similar tittle while the truth is - there is no such protocol that can be regarded as optimal for all areas of the brain, that match all possible paradigms, etc. Though, there are certain guidelines that one may use in order to optimise the time-space resolution to a particular conditions used just in your study.

fmri Temporal & Spatial resolution

Many efforts have been devoted to improve TR parameter from hardware (fast switching gradients, muli-element coil arrays) and pulse-sequence development. All this allow to choose shorter TR. The ultimate goal is to have TR- temporal resolution below 500 ms that allow adequate sampling not only BOLD signal but also physiological noise. If the latter is properly sampled it can be properly filtered away from the dataset. In the context of functional connectivity consider shortest possible TR (below 2000 ms) that would imply fare number of slices.

The product pulse sequence was modified that allows to reduce TR to 2000 ms and use echo-spacing in EPI, esp= 460 µs. A new multiband EPI allows TR=400 ms, full brain coverage.

Most often used the in-plane resolution is 3x3 mm, thickness 3 mm, that gives rise the voxel volume of 27 mL. At 3T the optimum resolution minimizing physiological noise is on the order of 8 mL (2x2x2 mm). This would provide ca. 3 times less signal, more slices for full-brain coverage and substantially increased TR. The phase-read matrix would increase to 128x128.

Resting State fmri at TR=400 ms full brain, multi-band acceleration= 5, in-plane acceleration= 2 ( tot x 10) shows very clean time courses and power spectrum allocated only in rest. state band. The current limitation is image reconstruction time that is for 240 volumes takes 6 h. computer time. The raw image size is ca 16 G., that may consume a lot of storage space.

Modified product EPI

"Pulse sequence time dig" - from the side of scale bar

features: flexible resolution in acquisition parameters, image resolution can be set equal to acquisition matrix or to 128x128, number of dummy scans = 5, echo spacing esp= 460 µs Echo train length depends on resolution. With standard setting 72x72 and R=2, ETL= 16.5 ms Allows to acquire images with "negative blips" - in opposite direction which can be used for correction of image distortion instead of conventional field map in FSL.

Field Map for FMRI

The protocol for field mapping use non-epi based pulse sequence. This is 2D GRE (gradient echo) with two echos. The field map rely on the echo time difference (not the actual eco time) that is fixed to 2 ms ( to map magnetic field variations over the range of 200 Hz).

The SPM field map toolbox ( then Realign & Unwrap function) of FSL Fugue should be used.

The difference between original fmri and unwrapped fmri :

Structural MRI

With 32 ch ( actually with 8 ch array as well) use the same pad thickness under the head all the time to keep the same distance between the occipital lobe and the coil! See more under FreeSurfer, FSL.

All structural protocols use:

T1w IR-SPGR BRAVO™ pulse sequence with GE 3D geometry correction.

Standard-1: Sagittal plane, 1 mm isotropic resolution ca 6 min scan time. ( no acceleration) , requires 8 ch array or 32 ch array Standard-2: Axial plane, 1 mm isotropic resolution ca 6 min scan time. ( no acceleration) , requires 8 ch array or 32 ch array High resolution T1w: Sagittal plane, 0.7 mm isotropic, ca. 7 min, 32 ch array Reduced time T1w ( for children, and less cooperative subjects): Sagittal T1w: 5 min

T2w Cube™ variable flip angle with GE 3D geometry correction. High res: 0.7 mm resolution ca. 7 min, 32 ch array Standard: 1 mm resolution 6 min, 8ch or 32 ch array

New: a buit in 3D intensity correction "Pure" has become available for 3T (see left: 0.8 mm resolution iso, + PURE)

3D T1w- 32 ch. array bias field removal may affect structure segmentation
See more under FreeSurfer, FSL.


Diffusion Tensor Imaging MRI for quantitative voxelwise tract-based spatial statistics (TBSS)

The recommended pulse sequence is a modified version of GE product dual spin-echo DWI. The pulse-sequence uses an improved fat suppression (SSGR slice selective gradient reversal technique) and keeps the raw image size, cubic voxels avoiding interpolation of the final image to nearest FFT size. It also allows for acquisition with negative phase encoding gradients (blips) – the short 30’ experiment used to correct image distortion (FSL tools "TOPUP")
The diffusion data is free from eddy current distortions due to the self-compensating diffusion gradient arrangement.
The max. number of diffusion directions is 150. The b-vectors are generated according Derek Jones proposal – b-vec uniformly distributed on a unit sphere using Columb repulsion forces algorithm.
Several diffusion protocols are recommended that can be found in Protocol Manager/Template/A_KI_diff_DTI_template:

1. DTI_2.3mm_b1000_60 8’03’’ - standard DTI, 60 directions
2. DTI_2.3mm_b1500_60 8’31’’ - allows DTI + tractography
3. DTI_2.3mm_b2200_60 8’11’’ - optimal for HARDI and 8 elements coil array
4. DTI_2.0mm_b1500_45 7’14’’ - hi-res DTI 45 directions
5. DTI_2.0mm_b1500_60 9’30’’ - hi-res DTI 60 directions

The FA maps acquired at standard b=1000 s/mm2 can be still contaminated by partial volume effect from CSF in the areas close to ventricles. Hence, the larger b-value (1500 s/mm2) can be used still producing similar FA maps with reasonable SNR > 17.

The difference between FA maps (fig. below) are attributed to the better resolved directionality for the major fiber bundle at b= 1500.

Diffusion Tractography with HARDI

Due to low coil sensitivity at high b-values required to resolve crossed fibers the 32 ch. NOVA coils is a preferable choice. Another 32 ch. coil from MR Instruments does not offer any advantage over 8 ch. array when high b-values are used. The goal is to maximize DW contrast for crossing fibers by using higher b-values and number of gradient directions (> 60) but SNR should not go below a certain limit which augmented by coil construction, b-value/ min_TE and voxel size. On the fig. below the MR Instrum. 32 ch. array was used with vox.size 2.3 mm -iso and two b-values : 1500 and 3000. The b-balue 3000 results in plenty of spurious short-length tracks < 4 cm length. One may se, that there is no advantage for using b= 3000 over a more trustful b= 1500 with low sensitivity coils. When more sensitive NOVA32 coil is used the situation will change.


This is "Neurite orientation dispersion and density imaging" a multi-shell DWI acquisition schema .

Quantitative Susceptibility Weighted Imaging, QSM

The protocol is based on Axial 3D-mGRE (multi-echo gradient echo pulse sequence), where 6 to 8 echo-images are acquired. The images are flow-compencated in slice direction. The Real/Imaginary pairs of each echo is saved in dicom format. Matlab recon is recommended.

MR Spectroscopy

There are several protocols for single voxel spectroscopy utilizing PRESS, mega-PRESS (for GABA).


Be aware the goggles that come with Eay-Tracker may cause artifacts

(A) Often the folding artifacts are observed in 3D T1w images. The origin is from the cables to the goggles. Apparently, the cable material contains 1H - protons. The signals are generated outside field-of-view and folds onto the image from two phase encoding directions:

What's to do? 1) Tape the cable bundle to the coil so that they approach the coil from the back side (Superior) /ask for help. 2) Use Sagittal 3D T1w

(B) B0- magnetic susceptibility The NNL LCD goggles will affect magnetic field homogeneity in frontal part of the brain, that was depicted by field maps: with and without goggles:

Goggles attached "Anterior to phantom" - from the side of scale bar
no Goggles

The gray scale does reflect the change of phase for the spins.
The phase spread range is ca. within pi-radians or less than 180°, otherwise the phase wraps will be observed.
This is not a dramatic effect since at the interface of temporal lobe the phase may change > 1000°.

Personal tools