Introduction

Arterial spin labeling (ASL) imaging can provide insight into stroke. Here we evaluate the 2D and 3D alternatives provided by the University of Oxford.

Methods

The included PDF provides sequence details for the images acquired. The following scans were exported (several were saved with and without intensity normalization and also created derived images):

1. T1 anatomical scan
2. 2D pCASL (97 volumes)
3. 2D pCASL PA (2 volumes)
4. 3D pCASL M0 (2 volumes)
5. 3D pCASL 6PLDs 8 averages (96 volumes)
6. 3D pCASL PA (4 volumes)

Note that recent versions of dcm2niix have been updated to extract important details direct from your DICOM images (so you do not need to look at the PDF files to determine these parameters. The Oxford team suggested the extraction of the following parameters that can help in BASIL analyses (use the '-b y' option in dcm2niix to save the BIDS sidecar). Here data is provided for the 2D sequence:

"RepetitionTime": 4.1,
"InversionTime": 1.8,
"BolusDuration": 0.7,
"TagRFFlipAngle": 20,
"TagRFDuration": 0.0006,
"TagRFSeparation": 0.001,
"MeanTagGradient": 0.8,
"TagGradientAmplitude": 6,
"TagDuration": 1.4,
"MaximumT1Opt": 0.5,
"InitialPostLabelDelay": [
	0.25,
	0.5,
	0.75,
	1,
	1.25,
	1.5	],

Methods

The 3D sequence clearly has better signal to noise. However, it tends to be very blurry in the slice (head-foot) direction. Further, it is very sensitive to motion artifacts. Therefore, the use of the 2D sequence for clinical patients seems justified.

Analysis

Here we will describe the analysis of the 6 Post-Label-Delay 2D sequence. This dataset is extremely similar to the sample provided in Example 3 ASL Primer, and in general the analysis will be similar. The data provided here has 97 volumes (the first volume is our M0 calibration image, and we need to separate this from the others for analysis). When analyzing multiple PLDs, make sure the ordering of your "PLDs" and "Repeats" is correctly specified. As described in the ASL Primer, you can validate this by looking at the signal intensity across volumes using a tool like MRIcroGL or FSLeyes. The figure below shows this - on the left is our data and on the right is data with a different ordering. The signal decays with post-label delay - the eight staircases for our data reveals they are stored one after another (PLD1, PLD2...PLD6, PLD1, PLD2...), whereas with the data on the right the images with similar signal are right next to each other (PLD1, PLD1,...,PLD1,PLD2, PLD2...).

Links

• Dr Wang provides an alternative pCASL sequence for Siemens scanners.
• ASLtbx is a popular tool for ASL analyses.
• Overview of Arterial Spin Labeling
• BASIL web page provides rich information on this tool.
• The FSL training course provides sample ASL data and a tutorial for analyzing this data.