In conventional arterial spin labeling (ASL)^1-3, the inversion of blood water in the neck is followed by a long post-labeling delay (PLD) to allow the blood to reach the tissue before image acquisition, yielding perfusion contrast. ASL can also be used to generate angiographic images^4,5 if the PLD is very short. In cerebrovascular disease, information about the status of the arteries and tissue perfusion are both crucial, but separate angiographic and perfusion acquisitions may be unfeasible in a busy clinical protocol.

In this study a new golden angle radial ASL imaging method is proposed in which the labeled blood is continuously imaged after labeling, allowing both angiographic and perfusion images to be reconstructed from the same raw data set. This makes efficient use of the PLD “dead time” normally required in perfusion imaging, whilst also giving a greater degree of flexibility in the image reconstruction process, allowing the resulting images to be tailored to the hemodynamics of each subject. A 2D multi-slice implementation is presented here, but the same principles could apply to 3D imaging.

A schematic of the Combined Angiography and Perfusion using Radial Imaging and ASL (CAPRIA) sequence is shown in Figure 1. After pre-saturation and pseudo-continuous ASL (PCASL) labeling, a spoiled gradient echo golden angle radial scheme⁶ is used to continuously image the labeled blood passing through the arteries and into the tissue. The azimuthal angle of the i^th line after the n^th PCASL preparation is:$$\phi_{i,n}=(n\frac{t_{max}}{TR}+i)\,\phi_G\qquad\text{where}\qquad\phi_G=111.2^\circ\tag{1}$$where t_max is the maximum temporal resolution desired for reconstruction. This results in relatively even coverage of k-space when combining data within an arbitrary temporal window across multiple PCASL preparations, thereby allowing reconstruction of angiographic images at high temporal resolution and perfusion images at lower temporal resolution from the same raw data set. Acquisition of PCASL control data is interleaved with label data to minimize subtraction artefacts from subject motion or scanner drift. Numerical simulations of the ASL signal⁷, accounting for attenuation due to the imaging pulses^8,9, were used to optimize the TR and flip angle (Figure 2).

Four healthy volunteers were scanned under a technical development protocol agreed by local ethics and institutional committees on a 3T Siemens Verio scanner using a 32-channel head coil. CAPRIA data were acquired in four 10mm slices sequentially, each taking 2.5min. Imaging parameters were: labeling duration 1.4s, imaging time 2s, matrix 192, TR/TE 12/6ms, flip angle 7°, bandwidth 102Hz/Pixel.

For this study, angiographic/perfusion images were reconstructed at 108/252 ms temporal resolution, respectively, using a regridding algorithm¹⁰ and adaptive coil combination¹¹. Perfusion images were post-hoc smoothed to boost SNR and match typical ASL in-plane resolution (3.4mm). To give the appearance of inflowing blood in the angiograms, “inflow subtraction” was also performed^12,13. Simple outflow timing metrics¹⁴ were calculated from angiographic images. Perfusion data were fitted to the ASL kinetic model⁷ including a macrovascular component¹⁵, with T₁ modified to account for the imaging pulses⁸.

The CAPRIA sequence is capable of simultaneously generating dynamic angiograms and perfusion images, as demonstrated in Figures 3 and 4. The expected patterns of vascular filling and tissue perfusion were observed in all subjects, in both the reconstructed images and the derived parameter maps (Figure 5). The golden angle radial scheme is well suited to this application, since dynamic angiograms require high spatial and temporal resolution, but are sparse and relatively high SNR, so undersampling can be tolerated. The converse is true of perfusion imaging, but lower spatial and temporal resolutions are required, so more data can be used to reconstruct each image and the denser sampling near the center of k-space improves the reconstruction. One additional benefit of this readout scheme is the absence of distortion and dropout artefacts that are often observed in conventional ASL perfusion imaging using readouts such as echo-planar imaging.

Acquisition of angiographic and perfusion images after ASL labeling has previously been shown using two separate dedicated readouts combined with a time-encoded preparation¹⁶. This has the advantage of allowing separate optimization of the two readout modules, but unlike CAPRIA, the timing of the resulting images must be decided in advance, so adaptation to the hemodynamics of each subject is not possible retrospectively.

3D CAPRIA should be possible using an extended golden ratio approach¹⁷. The radial trajectory and sparse nature of angiograms would also make this a good target for acceleration methods such as compressed sensing¹⁸. CAPRIA requires further optimization, calibration and validation against established methods, but it is hoped it could provide a comprehensive tool for evaluating blood flow to the brain in diseases such as acute stroke.