The purpose of this example is to analyse data that has been acquired with mulitple post-labeling delays. The analysis is on data from the same subject (and the same session) as used in Example 1 and Example 2. This example uses the data first met in Chapter 4 of the primer.
NOTE that we reccomend you use FSL v6.0.1 (or higher) for these exercises.
For this example you should use the Multi-PLD pcASL data, you will also need images from the Single-PLD pcASL data. The multi-PLD pcASL data was acquired using a label duration of 1.4 seconds and PLDs of 0.25, 0.5, 0.75, 1.0, 1.25 and 1.5 seconds.
Get the data
We can proceed to do the full quantification using a single call to
oxford_asl as we did in Example 1 and 2. We will copy
the approach we took in Example 2 and include motion and distortion
correction, and we will use CSF as the reference tissue for
calibration purposes:
oxford_asl -i asltc -o oxasl --casl --iaf=tc \ --tis=1.65,1.9,2.15,2.4,2.65,2.9 --bolus=1.4 --slicedt=0.0452 \ --fixbolus --artoff --spatial --mc \ -c aslcalib --tr 4.8 --cmethod=single --csf=csfmask --cblip=aslcalib_PA --pedir=y --echospacing=0.00095
Note that:
--tis=1.65,1.9,2.15,2.4,2.65,2.9.--bat= option isn't
used), this parameter will be infered from the data. For multi-PLD
data the --bat= option can be used to set the initial (and
prior) value for this parameter, but it is always estimated from the
data. We do not set it here as we are happy with the default value
(1.3 seconds) provided.oxford_asl not to estimate an arterial
(macrovascular) component, --artoff. When we only had
single PLD data this didn't matter, as there isn't information in
single PLD ASL to estimate such a component, but it might be
posisble to infer from multi-PLD data - something we will return to
below.oxford_asl to use the
reference region approach and supplied a CSF mask (in the same space
as the ASL data). If we hadn't it would use the voxelwise method. We
could, as we did in Example 2, supply the structural image, but it
is quicker to supply an existing CSF mask in this case (to save
repeating the registration of ASL to structural image). This CSF
mask is simply the result of a previous run of oxford_asl on this
subject, as we generated in Example 2.
The results directory from this analysis should look similar to
that obtained in Example 2. The main difference is that the
arrival.nii.gz image is now a genuine estimate of
ATT;if you examine this image you should find a pattern of values
consitent with the time it takes for blood to transit between the
labeling to imaging regions. You might notice that the
arrival.nii.gz image was present even in the single-PLD
results, but if you looked at it contained a single value - the one
provided by the --bat= option - which meant that it
appeared blank in that case.
In the analysis above we prevented oxford_asl from
including an extra component in the model that would capture any
arterial (macrovascular) signal present in the data. This is fairly
reasonable for pcASL in general, since we can only start sampling
some time after the first arrival of labeled blood-water in the
imaging region. However, given we are using shorter PLD in our
multi-PLD sampling to improve the SNR there is a much greater
likelihood of arterial signal being present. Thus, we might like to
repeat the analysis with this component included:
oxford_asl -i asltc -o oxasl --casl --iaf=tc \ --tis=1.65,1.9,2.15,2.4,2.65,2.9 --bolus=1.4 --slicedt=0.0452 \ --fixbolus --spatial --mc \ -c aslcalib --tr 4.8 --cmethod=single --csf=csfmask --cblip=aslcalib_PA --pedir=y --echospacing=0.00095
All we did in this run was remove the --artoff
option. The results directory should be almost identical to the
previous run, but now we also gain some new results:
aCBV.nii.gz and
aCBV_calib.nii.gz; follwing the convention for the
perfusion images, these are the relative and absolute arterial
(cerebral) blood volumes respectively. If you examine one of these
and focus on the more inferior slices you should see a pattern of
higher values that map out the strcture of the major arterial
vasculature, including the Circle of Willis. This finding of an
arterial contirbution in some voxels rsults in a correction to the
perfusion image - you may now be able to spot that in the same
slices where there was some evidence for arterial contamination of
the perfusion image before that has now been removed.
In the same way that we could use oxford_asl for
partial volume correction in Example 2 for single PLD pcASL, we can do this
for multi-PLD:
oxford_asl -i asltc -o oxasl --casl --iaf=tc \ --tis=1.65,1.9,2.15,2.4,2.65,2.9 --bolus=1.4 --slicedt=0.0452 \ --fixbolus --spatial --mc \ -c aslcalib --tr 4.8 --cmethod=single --csf=csfmask --cblip=aslcalib_PA --pedir=y --echospacing=0.00095 \ --fslanat=T1.anat --pvcorr
Note that we need to provide partial volume estimates and we have thus used the T1.anat result from Example 2. In practice we do not need to supply the CSF mask in this case.
The results directory now containes, as a further subdirectory, pvcorr,
within the native_space subdirectory, the partial volume
corrected results: gray matter (perfusion_calib.nii.gz
etc) and white matter perfusion
(perfusion_wm_calib.nii.gz etc)
maps. Alongside these there are also gray and white matter ATT maps
(arrival and arrival_wm respectively). The
estimated maps for the arterial component
(aCBV_calib.nii.gz etc) are still present in the
pvcorr directory. Since this is not tissue specific there
are not separate gray and white matter versions of this parameter.
In Example 1 we firstly manipulated the data using
asl_file to examine the perfusion-weighted image and
satisfy ourselevs that there was perfusion information in the
data. We can use asl_file with multi-PLD pcASL (or
almost any other ASL variant). For example, if we wanted to perform
tag-control subtraction we could do as follows:
asl_file --data=asltc --ntis=6 --iaf=tc --ibf=tis --diff --out=asldiffdata --mean=asldiffdata_mean
We have to be careful to do is specify the number of
PLD (TI) in the data: --nti=6. The
asldiffdata_mean.nii.gx contains result of averaging all
the repeats of each invidiual PLD - thus it contains 6 separate
volumes; if you examine these you should see that the first volume is
relatively bright compared to the last couple of volumes. This
reflects that fact that at the first PLD most (but not all) of the
brain has recieved labelled blood-water and thus there is quite a lot
of perfusion signal. By the time of the later PLD the end of the bolus
of labeled blood-water has already reached every part of the brain and
T1 decay is gradually reducing the signal. If you examine the
asldiffdata.nii.gz file you will find it contains the
noisy label-control difference images.
Note that we have also told asl_file the oder in which
the PLD are to be found in the data: --ibf=tis. They were
aquired as a full set of PLDs first, followed by another repeat of all
the PLDs etc. It is possible, but less common, that we could have
acquired all the repeats for a single PLD first, then all at the next
PLD etc.
In this example the early PLDs are a pretty good basis on which to examine if the data contains perfusion information. However, this might be less reliable if we have fewer repeats at each PLD where the influence of noise at each PLD would be more noticable. We can generate a very simple perfusion-weighted image by averaing all the PLD:
fslmaths asldiffdata_mean -Tmean aslpwi
There are various other things we can do with asl_file
to mainpulate the data; for example, splitting the data into separate
files for each PLD using the --split option, or output
the data having changed the oder in which the repeated PLDs appear in
the data according to the pattern using the --out option
with --obf. For more information see the help for asl_file.