Loading...
 

BOLD fMRI - Acquisition Protocols and Methods


If you are about to propose to carry out a new project at LNiF involving MR, the project proposal template asks you to specify the MR measurement methods you plan to use (a slide on "Data acquisition protocols"). Unless you require non-standard methods, you should be able to find an optimised and well-tested EPI protocol from those we have set up, or suggest one based on these, with sensible changes. If you need more detailed information about scanner functionality you might find it here).

This excel file summarizes the effects of number of slices on space coverage, minimum TR, optimal flip angle and SNR. This information is useful to keep in mind possibilities and consequences of faster acquisitions with short TR.

If you have non-standard requirements (e.g. real-time fMRI, perfusion fMRI, very fast EPI with limited brain coverage, etc) or don't find the information you need to enable you to decide what would be right for your experiment, ask the MR group.

Back to: MRI Information

Selecting an EPI protocol


A number of whole-brain EPI protocols have been optimised and tested, covering a range of spatial and temporal resolutions. A quite complete list of the parameters of these is given in this file. The most commonly used protocol is whole brain EPI with 3x3x3 mm3 voxels and a TR of 2.2 seconds - protocol lnif_ep2d_3x3x3_TR2200. You may wish to change some parameters to give you the TR you want, match the parameters of a study you wish to reproduce or to improve SNR in a particular area - see Which EPI parameters can be changed and Which EPI parameters shouldn't be changed.
--------------

How have EPI protocols been optimised?


----------------

Which EPI parameters can be changed?

Number of dummy/preparation scans

Dummy scans are executions of the MR sequence (RF excitation of the sample, gradients) with no acquisition of data. That allows longitudinal magnetisation
equilibrium to be (quasi) attained (see figure below)

Image
The first scans have higher contrast than scans acquired after longitudinal equilibrium has been achieved, however (cf Vol 1 and Vol 4 below).
Image

Particularly images with TR<2 s have low contrast between grey matter, white matter and CSF, and image registration can be poor.

There is a choice to be made, therefore, between using dummy and having stable global signal over the first scans, and acquiring the first excitations to have early high contrast volumes which can be used to aid image registration, in which case the first few scans have to be skipped in the analysis.

The default behaviour of Siemens EPI sequences is to require that a number of dummy scans be used that lasts at least 3 s (it also must be an integer number of scans, so if your TR is 2200 ms you will have 2 dummy scans). If there is no option for the number of dummy scans in the specials card then you are using either a Siemens sequence or an arlier sequence by Maxim Zeitsev (mz...). With these sequences you will not be able to set the number of dummy scans and it accord to this 3 s rule. In most cases this is insufficient. For instance, the third data point would be acquired in the graph of "Global Signal" above, although equilibrium longitudinal magnetisation clearly still hasn't been reached. 10 s of dummy scans should be used.
---------

Number of volumes to skip before issuing first trigger

Triggers can be issued by the sequence at the first volume acquired (after dummy/preparation scans), or after a specied number of volumes (the "Number of Volumes to Skip" - NVS). If no dummy scans are used it can make sense to set the NVS to a value which allows equilibrium magnetisation to have reached equilibrium. See Experiment Timing - Suggested Schemes for Triggering the presentation of stimuli (particularly option 3).

Repetition time TR

The parameter that you are most likely to want to change from default value is the repetition time (time taken to acquire one volume), TR. If you need to reduce the TR from that in one of our EPI protocols, you can do so by first reducing the number of slices. You should also change the flip angle.

Flip Angle

The flip angle of our EPI protocols is set to the Ernst angle, to give maximum SNR. If you have changed the TR from the default value in one of the "Common Tested" protocols, you should change the Flip angle to be equal to the Ernst angle at the new TR. You can calculate this using this Excel calculator or this table, both of which are based on a T1 of grey matter at 4 T of 1600 ms. The table is also on the wall near the console.

Number of Slices

The number of slices acquired can be changed to reduce TR (see above), or to allow breaks in the acquisition every TR (sparse sampling) in which auditory stimuli can be presented and verbal responses given.

Distance Factor

Expressed as a percentage of slice thickness, this specifies the gap between slices. To increase temporal resolution and because the profile of slices isn't sharp, this is generally set to 10-20%. Larger distance factors can be used but make the acquisition prone to motion artefacts due to spin history effects.

Phase-encode direction

Our EPI protocols are set up with the phase-encode (PE) direction anterior-posterior, with PE blips ...???. This polarity can be changed to the positive scheme. Reasons why that might be advantageous are listed here???, along with data indicating where this gives higher signal, where lower. We strongly recommend against changing the PE direction to left-right, because it introduces distortions which are not symmetric across the midline.

Inclusion of Fat Saturation pulses

Because fat has a different resonance frequency to water fat signal is chemically shifted and overlaps with other areas of the image. Fat saturation pulses can be used to suppress this signal by selectively exciting then crushing it. These pulses are not included in protocols by default but can be added at the cost of increased TR. Although the shifted fat signal is quite small in slim subjects, fat sat pulses should be considered if a group of subjects are to be studied who may often be overweight (e.g. diabetic patients).
----------

Which EPI parameters shouldn't be changed lightly?

Echo Time

To optimise BOLD sensitivity and temporal resolution the echo times of our EPI protocols have been set to approximately 80% of the T2* in magnetically homogeneous regions at the respective resolutions. Unless your study is focussed on a magnetically highly inhomogeneous area you should not change the echo time.

Receiver Bandwidth

The intensity of aliased images which may overlap with the main images, or Nyquist ghosts, is influence by the receiver bandwidth. This has been carefully selected to avoid acoustic resonances which worsen the effect.
----------

Experiment Timing - Suggested Schemes for Triggering the presentation of stimuli

Trigger signals from the scanner can be used to initiate the presentation of stimuli. These can be set to be issued after an arbitrary number of scans, which may be preceeded by preparation scans (during which time the MR signal approaches longitudinal equilibrium). To understand the correspondence of MR scans and behavioural events (stimuli, responses) it is necessary know when triggers were issued by the scanner and whether dummy/preparation scans were acquired (and if, how many). Here we explain the role of dummy/preparation scans, suggest three schemes for triggering the presentation of stimuli, and show how to determine in retrospect the number of dummy scans that were used and when the first trigger was issued.

The following illustrations are based on experiment requiring 100 image volumes, where 5 scans is sufficient to achieve equilibrium magnetisation
-------------

1 - Use enough dummy scans to get equilibrium longitudinal magnetisation (no high contrast first scans)

timing_recommendation_1.PNG
-------------

2 - Acquire first (high contrast) scans, extend first baseline period in paradigm

timing_recommendation_2
-------------

3 - Acquire first (high contrast) scans, keep baseline period to duration to be used in analysis, skip volumes before issuing trigger

timing_recommendation_3

AutoAligning Scans


Auto aligning is setting up a protocol with a specified geometry in the space of the head rather than magnet, and realising this by acquiring a circa 2 minute dual echo "AAScout" scan which the system registers to a brain template.

Advantages:It

  • enables an exact and reproducible slice prescription
  • is useful for resumption of scanning after break in a session
  • enables reproduction of exact geometry in longitudinal studies (pre-post TMS, neuro-rehabilitation post-therapy etc)
  • allows structures of interest to be reliably covered with limited slab protocols
  • improves the overlapping coverage between a group of subjects


Disadvantages:

  • It adds 2-3 minutes to session
  • The system can (rarely) fail to register the AAScout to the template
  • The adopted geometries not always identical, so geometry still needs to be copied between PSF and EPI, for instance
  • Takes a little undoing if you decide not to use the adopted geometry


AutoAligning is a 2-step process

  1. – Plan scans on the AAPatient template (phantom pilot session) or pilot subject (subject pilot session)
  2. – Acquire an AAScout (measurement session)


To plan scans so that they AutoAlign , to use AutoAlign and to undo the geometry determined by AutoAlign see this description for operators this description for operators.
--------

Distortion correction

Image


Echo-planar images suffer from distortion in regions of non-uniform magnetic field. This is a particular problem in the frontal and ventral brain regions, and areas close to the auditory canals. Distortion scales approximately with field strength and can be as large as 1-2 cm in frontal regions at 4 T. Distortion may be corrected for via the field mapping (FM) and point-spread function (PSF) methods.

The Point-Spread Function  Method


References:
- Zeitsev et al, Magn Reson Med. 2004 Nov;52(5):1156-66
- Zeng et al., Magn Reson Med. 2002 Jul;48(1):137-46

Performing the PSF distortion correction online


This is the default setting for most acquisitions (protocols). The EPI scan needs to be preceded by a PSF scan with the same parameters. The option to do this online correction must be set by editing the EPI protocol in the Exam Explorer (DiCo card, check the 'Apply DiCo' box) as the DiCo card doesn't work in the main scan planning window. Settings for the distortion correction can also be set. We recommend unchecking the 'Amplitude Correction' box and checking the 'Discard Fake Ref Image' box

Performing the PSF distortion correction offline


There are some (probably rare) circumstances in which you might want to perform the PSF correction offline. One is that you might be using an EPI-based sequence which doesn't allow the online correction (such as DTI or PASL). Alternatively you may wish to apply a PSF other than the immediately preceding one to an EPI run. E.g. in the sequence of scans

PSF Scan 1
EPI Run 1
PSF Scan 2
EPI Run 2

you may want to apply PSF Scan 1 to EPI Run 2. In either of those cases the offline PSF distortion correction is for you. Read how to do it here


Acquiring EPI data without a PSF correction

The Field Mapping Method

  • Background: Field maps are maps of the variation of B0 over the object, and can be acquired before or after the functional run (or both before and after, to see if there has been movement). Phase images from two echos (spin-echo or gradient echo) are required to calculate field maps. A third scan or further echos can be acquired to check field maps. Considerable post-processing is required in the field mapping method, including sorting of the echos, separate channel data and phase and magnitude images, calculation and refinement of the field maps and correction of the images. Contact one of the MR group if you wish to use field mapping to discuss the programs and processing you'll need to do.

Image


Although the use of multi-channel coils is advantageous at high field, field maps needed for the distortion correction are based phase images, and phase images reconstructed by the scanner (using the sum-of-squares of the separate channels) contain non-physical discontinuities result due to arbitrary phase offsets (are just wrong). High quality field maps may be generated from separate channel data, though, and the consistency between channels used to identify unreliable voxels. Using high quality field maps generated in this way, the correction with the field mapping method is excellent, and images show little residual distortion. Because it requires postprocessing, the field mapping method is only recommended for specialist applications where PSF may not work - e.g. high resolution protocols using GRAPPA, where the data rate prohibits an online correction. An ISMRM abstract comparing the Field Mapping and Point Spread Function methods of distortion correction can be viewed here

  • How to acquire the Field Map data:
    • The dual-echo field map sequence is in the protocol folder LNIF-Tested\EPI at the console. We have field map protocols for a range of spatial resolutions. The sequences are already prepared to save magnitude and phase images for each separate channel.
    • This basic sequence needs to be adapted each time so that it matches the following parameters of the target EPI that will be corrected afterwords:
      • matrix size
      • FoV
      • in-plane resolution
      • slice thickness
      • slice gap
      • slice orientation
      • number of slices
      • fat suppression on or off
    • A common problem in setting the above parameters is inflexibility due to TR. So, first set an exaggerated high TR (+2000ms of what it is), then set the above listed parameters, then adjust TR to the allowed minimum value. The flip angles should be adjusted to math the Ernst Angle of the TR in gray matter. You can calculate this using this Excel calculator or this table, both of which are based on a T1 of grey matter at 4 T of 1600 ms. The table is also on the wall near the console.

  • How to do the field map correction:
    • Summary: you will request an account in a unix computer called 'crunch', you will copy the nifti data that you will distortion correct to crunch, and in crunch you will run a script that uses matlab and FSL programs to first calculate the field map from your multi-echo structural data and then calculate the distortion correction in the EPI from the field map.
    • Get an account in Crunch, a 64-bit unix computer resource from the MRI Methods Group
    • Get familiarized with unix and how to mount lnif-storage to exchange data with crunch. Your default shell will be bash, all instructions below work for bash shells.
    • You should have a home account with a matlab folder and a startup.m file to set you paths: /home/UNITN/name.lastname/matlab. Your startup.m file should like like the example below. *ENSURE YOU ADD THE LAST LINES IN ORDER TO HAVE A MATLAb PATH TO THE FIELD MAPPING PROGRAMS*

%******* Freesurfer *************

fshome = getenv('FREESURFER_HOME');

fsmatlab = sprintf('%s/matlab',fshome);

if (exist(fsmatlab) == 7)

path(path,fsmatlab);

end

clear fshome fsmatlab;

%********************************

%********* FreeSurfer FAST ********

fsfasthome = getenv('FSFAST_HOME');

fsfasttoolbox = sprintf('%s/toolbox',fsfasthome);

if (exist(fsfasttoolbox) == 7)

path(path,fsfasttoolbox);

end

clear fsfasthome fsfasttoolbox;

%********************************

%****** Field Mapping Distortion Correction *******************

path(path, '/home/UNITN/jorge.jovicich/matlab/FieldMapping_Matlab_Scripts');

%********************************

    • In your folder /home/UNITN/name.lastname/matlab edit the following text file (past and copy from below) and call it for example fm_dist_cor.sh. This is a bash shell script, please read the comments at the top to get an idea of what it does. After saving it make it executable (type the following unix command from the directory where you have the script: chmod a+x fm_dist_cor.sh):

#!/bin/bash

##########################################################################

# bash script to submit batch jobs for field map estimation and EPI distortion correction

# The main function takes three arguments and does all the calculations.

#

# fm_calc_main('arg1', 'arg2', 'arg3')

# arg1: This argument is for the nifti data path containing the multi-echo structural and EPI series to correct

# arg2: Series number for the multi-echo structural data acquired with the same spatial resolution and slice

# prespcription as the EPIs to correct. In the scan_list.txt file this corresponds to the series like:

# MGE fieldmap_2x2x2. 128x128x22, axial(11°), 2 echos at 6 10 ms, MCMP

# arg3: Series number for the EPI series to correct

#

# Sample command line:

# fm_calc_main('/data/users_data/jorge.jovicich/FM_data/19670329GEMI_200908130830/nifti','10','13')

#

# Program outputs:

# - Under the path of arg1 there will be a new folder called FM_results_jj with the field map data

# - Under the path of arg3 there will be a new compressed nifti file with the distortion

# corrected EPI: Image_fmcor.nii.gz.

# To look at or further process the distortion corrected data unzip doing 'gunzip Image_fmcor.nii.gz'

#

# Assumptions

# - add the following matlab path to find the needed functions:

# /home/UNITN/jorge.jovicich/matlab/FieldMapping_Matlab_Scripts

#########################################################################

date

/usr/local/bin/matlab -nodesktop << EOF #starts Matlab

fm_calc_main('/data/users_data/jorge.jovicich/FM_data/19670329GEMI_200908130830/nifti','10','14')

exit

EOF #end of Matlab commands

    • The script above allows you to run the whole field map and distortion correction processing without having to open or modify matlab files. For each EPI series you need to correct add a line with the function fm_calc_main_jj('arg1', 'arg2', 'arg3') pointing to the right subject, structural multi-echo data and EPI series to correct. When you have multiple EPI series to correct in a session, the code will figure out that the field map needs to be calculated only once, and applies it to each EPI series you put in arg3. When the code finishes each EPI correction the results are saved with the name "Image_fmcor_nii" in the same folder where the original uncorrected data was ("Image.nii"). The size of the field map corrected file is double the original file because FSL saves data as 32-bit real (floats) and the input data is 16-bit integer data.
    • Try the fm_dist_cor.sh script out pointing at your own data and let us know how it works!



--------

Physiological Data


If you have saved physiological data using the EPI sequence, as described here, then you will have respiration belt and photo plethysmograph data sampled at 50 Hz, which will be called PmuSignals_00x.resp and PmuSignals_00x.puls respectively. The acquisition of that data starts before and end after the EPI acquisition, but can be matched to the acquisition of scans using a marker in the file indicating the number of seconds before the scan begins (e.g. "1SectoStart") using this MATLAB program, which also deletes junk prior to the marker and parses the data from a not very user-friendly one-row format to a single-column format (contact Simon for most recent version).

A little work still needs to be done to match this to the MRI data. The physiological data in the parsed file still over-runs the end of scanning, so values after the end of scanning need to be removed (after 50*NR*TR data points). Likewise, if no dummy scans were acquired and initial scans are being skipped in analysis, these points need to be skipped in the physiological data (50*TR*NumberOfTimePointsSkipped data points).

After this, the physiological signals can be downsampled to the TR to form useful regressors.

--------

Slice acquisition order


Slice order can be set in the protocol to either "Ascending", "Descending" or "Interleaved".

"Ascending" means consecutive, in the order inferior->superior (foot->head) for axial
"Descending" means consecutive, in the order superior->inferior (head->foot) for axial
"Interleaved" means...
ascending interleaved with even slices first, then odd slices if the number of slices acquired is even (2,4,6,8,...end; 1,3,5,7,..end-1)
ascending interleaved with odd slices first, then even slices if the number of slices acquired is odd (1,3,5,7,...end; 2,4,6,8,..end-1),
where the numbering is of the slices above is in position order (1 = closest to feet, 2 next closest, end is at the head)

If you want to know with which acquisition scheme your images were acquired look in the text section of the DICOM header. This is in text_header.txt if you used dicom_sort_convert_main.m to sort your images to NIfTI or Analyze, otherwise in every DICOM file in your EPI time-series. Open one of these with an editor (e.g. Wordpad) and search for "### ASCCONV BEGIN ###". Not far below you will find a field called

"sSliceArray.ucMode"

which will have one of the following values:

"0x1" which means "Ascending"
"0x2" which means "Descending"
"0x4" which means "Interleaved"

The Proof

Scan 1) Even slices: In this acquisition the subject had their head straight for slices 1-3 (acquisition order), and rotated it for slices 4-6. You can see that the acquisition order is (2,4,6,1,3,5 - by position)

Image

Scan 2) Odd slices: In this acquisition the subject had their head straight for slices 1-4 (acquisition order), and rotated it for slices 5-7. You can see that the acquisition order is (1,3,5,7,2,4,6)

Image

Scan 3) Acending or descending. In this acquisition the subject had their head rotated for slices 1-3 (acquisition order), and straight for slices 5-7. You can see that slices closest to the feet were acquired first.

Image


--------

AutoAligning Scans


Auto aligning is setting up a protocol with a specified geometry in the space of the head rather than magnet, and realising this by acquiring a circa 1 minute dual echo "AAScout" scan which the system registers to a brain template.

Advantages:

  • It enables an exact and reproducible slice prescription
  • Useful for resumption of scanning after break in a session
  • Enables reproduction of exact geometry in longitudinal studies (pre-post TMS, neuro-rehabilitation post-therapy etc)
  • Allows structures of interest to be reliably covered with limited slab protocols
  • Improves overlap of coverage in a group of subjects


Disadvantages:

  • Adds 2-3 minutes to session
  • The system can (rarely) fail to register the AAScout to the template
  • Adopted geometry not always identical, so geometry still needs to be copied between PSF and EPI, for instance


AutoAligning is a 2-step process

  1. – Plan scans on the AAPatient template (phantom pilot session) or pilot subject (subject pilot session)
  2. – Acquire an AAScout (measurement session)


To plan scans so that they AutoAlign , to use AutoAlign and to undo the geometry determined by AutoAlign see this description for operators (???need this link???).
--------

How to convert *.IMA files from the scanner into anonymised sorted DICOM or convert to Analyze/NIfTI format?


Images exported from the scanner have the extension "IMA". These are DICOM images.

We have developed a MATLAB GUI-based program for converting IMA images to Analyze format or sorting them into directories and renaming them with the DICOM (dcm) extension, and anonymising them in the process. This is available for download on the MATLAB exchange server. Most of this was developed at LNiF, but it also making use of lots of functions from other sites (credited).

Image

Main Features

1) The program produces a list of the scans you acquired in your session
2) It sorts data into directories
3) It 'lightly' anonymises DICOM files, removing the Patient Name field only from the DICOM headers, meaning you can always trace scans back to subject and session via PatientID and Accession numbers*.
4) It allows you to convert your data to NIfTI format (.nii), or two-file Analyze (.hdr and .img - useful for users of SPM before SPM5)

Installation Instructions

You will need MATLAB, and have the "image processing" toolbox installed.
If you want to use the anonymized DICOM files you create in Brain Voyager, you will need to use MATLAB version 2006b or later.

1) Download the zip file below and extract it to your computer.
2) Add the directory to which you extracted it to your MATLAB path (File>Set Path>Add with Subfolders (select directory) > Save>Close)
3) Run MATLAB
4) Run the GUI by typing "dicom_sort_convert_main" at the command line prompt
5) Select the directory in which you *.IMA files reside with the "Select Unsorted DICOM directory" button (see screeshot below).
6) Unless you want the images to be sorted to a different directory you don't need to use the "Select Destination directory" button. Subdirectories for the sorted and converted data will be created ('dicom' and/or 'analyze', depending on your choice of conversion).
7) Check the boxes to convert (either to NIfTI or Analyze), or to sort your .IMA data into directories (or both). Alternatively you can just create a list of scans.
8) Hit the "Sort and Convert" button.

Bugs/strange behaviour

1) At the moment, the directory for analyse or nifti images is called 'analyse', regardless of whether they are really analyze or nifti
2) There are some warnings about missing DICOM fields: ignore.

Problems?

1) Read these instructions through carefully
2) Type "help dicom_sort_convert_main" at the MATLAB command line
3) If you can run the program but it crashes with an error, copy this to a file and sontact Simon Robinson for help.

* - The Patient Name field is removed from dicom headers. Final responsibility that no sensitive information is released from the lab lies with each researcher. Please check that your data is anonymised by using a freeware DICOM viewer and looking at the header information (i.e. IrfanView , from http://www.irfanview.com/ . Load your image then go to Image>Information).

Using the scan_list

Modifying the parameters listed in the scan list


--------

Right-Left image orientation issues

  • We acquired a reference dataset with a volunteer turning his head towards his right during different acquisitions (3D MPRAGE, 2D EPI, 2D FSE, 2D DTI). This anonymized dataset can be used as a reference dataset for any arbitrary processing workflow that you have, so that you follow what you are doing with a clear and known input.
  • The anonymized dataset is available to all users at the following address: \\Lnif-storage\jorjov_n\MR-DATA-FOR-USERS\Left-Right-Orientation-Reference.
  • Here you will find the original dicom files exported from the scanner, as well as various examples of nifty generated data from the dicoms. Since there are as many possible workflows as MR users (at least!) we cannot check for all possibilities, so the key thing is that users check for left-right issues using your own processing environment.


--------

Back to: MRI Information

Created by simon.robinson. Last Modification: Monday 25 of October, 2010 22:55:31 CEST by jorge.jovicich.