Dynamic Analysis - Arterial Spin-Labelling

The Arterial Spin-Labelling module performs an analysis of ASL data acquired using alternating spin-labelled/control MR images to produce maps of cerebral blood flow (CBF). The analysis follows that recommended in: Magn Reson Med.73:102-16 (2015) "Recommended implementation of arterial spin-labeled perfusion MRI for clinical applications: A consensus of the ISMRM perfusion study group and the European consortium for ASL in dementia." Alsop D.C. et al. The default values for the settings below come from this paper.

You can analyse images acquired using either pseudo-continuous ASL (PCASL) or QUIPSS II pulsed ASL (PASL) to obtain perfusion values. The units of CBF in the output image are mL/(100 g)/minute.

Note: the ability to obtain meaningful quantitative perfusion values is heavily-dependent on the implementation of the ASL pulse sequence and entering accurate values for the sequence parameters required below. Obtaining an accurate estimate of the labelling efficiency is particularly difficult, and poor estimates will lead to systematic errors in the perfusion values.

You choose the Arterial Spin-Labelling module by clicking the ASL tab in the Dynamic Analysis tool.

dynamic_asl

Set up the input images as detailed in the Introduction, noting that the number of time-points should be set equal to sum the number of labelled plus control images (i.e., twice the number of labelled images).

General analysis settings

This sets the partition coefficient for water in exchange between the blood and the brain parenchyma. 0.9 is the recommended setting.
This sets the longitudinal relaxation time (T₁) of whole blood. The default of 1,650 ms is the value at 3.0 Tesla. At 1.5 Tesla, the T₁ of blood is approximately 1350 ms. Set the value according to the field strength of your MRI scanner.

This sets the labelling efficiency of the pulse sequence used. The default value of 0.85 is that expected for PCASL; for PASL the expected value is 0.98. However, these values are likely to be very implementation-dependent.
Normally, the first image acquired is a labelled image, followed by a control image. Labelled and control images are acquired repeatedly in the same order. If the first image acquired was the control image, then select this checkbox.
In order to produce quantitative CBF maps, the CBF values need to be scaled by dividing by a fully-relaxed (proton-density [PD]) image. To perform this quantitative analysis, select this checkbox and then set the PD image below.
Note: the PD image must be acquired with exactly the same geometry (number of pixels in each dimensions; pixel sizes and slice thickness) and with the same transmitter and receiver gain settings as for the ASL data, otherwise the scaling will be incorrect.
If you don't have such a PD image, then Jim will do a semi-quantitative analysis by using an average of all the "control" images as a substitute for the PD image. As long as the repetition time (TR) is very long compared to the T₁ relaxation times, this this may be a reasonable approximation.
If the scan repetition time is short compared to the tissue T₁, this can result in an underestimation of perfusion values. You can compensate for incomplete longitudinal relaxation by creating a T₁ map using (for example) an inversion-recovery or saturation-recovery pulse sequence. Then you can use the Image Fitter to calculate the T₁ map. If you have a T₁ map, select this checkbox, then set the T₁ map image below.
Note: the T₁ map must have the same geometry (number of pixels in each dimensions; pixel sizes and slice thickness) as the ASL images. The T₁ values in the map must be in milliseconds.
If there is any drift in the signal intensities over time, this can be accounted for using confounds section. The drift can be modelled using a linear term to account for steady drift, and by periodic terms to account for cyclical oscillations. Set the number of cycles over the data collection period using the spinner. Deciding the settings for this correction is best done using the roaming intensities to explore the data.

Pseudo-continuous ASL (PCASL) analysis

If your images were acquired using PCASL, select the option Button to specify that the acquisition was Pseudo-continuous
ASL

. Then set the additional parameters:

The post-labelling delay. This is the time between the end of the labelling period and the start of acquisition.
τ is the duration of the labelling period.

Pulsed ASL (PASL) analysis

If your images were acquired using PASL, select the option Button to specify that the acquisition was Pseudo-continuous
ASL

. Then set the additional parameters:

TI is the time from the Tag/Control pulse to the start of acquisition.
TI₁ is the time from Tag/Control pulse to the QUIPSS II Saturation pulse.

See Magn Reson Med.73:102-16 (2015) for further details of the acquisition parameters and recommendations for processing.

Residuals Image

Selecting the Checkbox to create an image of the residuals

check-box will create output image(s) with the same names as the input image(s), but with the suffix "residuals". The residuals image(s) will be the result of subtracting the modelled ASL signal (including the confounds) from the input data. Residuals images can be useful for investigating whether fitting the ASL still leaves some "structure" - this would be the case if the ASL does not provide a compete description of the input images, and that possibly further correlates should be included in the model. In the case of a "complete" ASL, the residuals will consist entirely of Gaussian-distributed uncorrelated values.

Performing the Analysis

When you click the apply_button

button, the output image created will have a name produced by taking the base name in the image output section and appending the suffix "CBF". The units of CBF in the output image are mL/(100 g)/minute.
Note: physiologically non-meaningful CBF values less than zero are clipped and given the value zero in the output CSF image.

Jim Home