Quantifying cerebral blood flow with distortion-corrected pseudo-continuous arterial spin labeling
Michael N Hoff1, Swati R Levendovszky1, and Jalal B Andre1

1Radiology, University of Washington, Seattle, WA, United States


Cerebral blood flow (CBF) may be quantified using pseudo-continuous arterial spin labeling (pCASL), but pCASL suffers from image distortion due to its long echo planar imaging (EPI) readout. Phase labeling for additional coordinate encoding (PLACE) is employed here to remove distortion. EPI images and pCASL subtraction images show improved spatial accuracy when compared to a T1-weighted, anatomical reference image. This results in improved accuracy in CBF quantification, which could potentially improve the assessment of disease-specific patterns indicated by regional CBF abnormalities.


Arterial spin labeling (ASL) offers a non-invasive method of quantifying cerebral blood flow (CBF), an important marker of autoregulatory function1. Traditional ASL acquistions use echo planar imaging (EPI) readouts, which suffer from image distortion due to susceptibility effects2. The ensuing signal mislocalization compromises the utility of ASL to accurately assess disease-specific patterns of regional CBF abnormalities. Phase labeling for additional coordinate encoding (PLACE) removes image distortion3 and Nyquist ghosting in ASL without requiring further data acquisition, and has been successfully employed in animal studies4. Our goal is to extend this approach to human studies, with the hope of increasing the diagnostic and clinical value of ASL by improving the spatial localization of CBF measurements.


PLACE pulse sequence modification: PLACE encoding was achieved by adding gradient areas along the PE direction of odd EPI dynamics (even dynamics were unchanged).

Experiment: Images were acquired of four subjects (one is shown) using a 16-channel head receiver coil and clinical protocols on a 3T Philips Ingenia scanner. A pseudo-continuous ASL (pCASL5) preparation was employed, with a post-labeling delay (PLD) = 2s, and a label duration ($$$\tau$$$) = 1.8s. The acquisition comprised of 30 dynamics - each with an interleaved control and label acquisition - using a single-shot 2D EPI readout with flip angle = 90°, TR/TE = 5000/23.5 ms, and 71/72/20 matrix size and 2.9/2.9/5 mm voxel size along frequency/phase/slice directions. A 4-dynamic reference “M0” EPI image with identical readout and TR but without label preparation was acquired for CBF quantification. T1-weighted spin-echo reference images were also acquired with 256/205/30 matrix size along frequency/phase/slice directions.

PLACE Postprocessing: Figure 1 displays how the PLACE distortion/displacement map is generated. PLACE gradients generate phase ramps along the image PE direction of each odd EPI dynamic (Fig. 1c). Complex image products “C” were formed from each odd dynamic (Fig. 1c) and the conjugate of its neighbouring, even dynamic (Fig. 1b). The phase of C is the phase difference between even and odd dynamics, which is directly proportional to the true signal position3 (Fig. 1d). Figure 1e shows the displacement map C′ with flattened phase that was generated by applying a linear phase ramp in opposition to the phase ramp in C. Sub-pixel interpolation was addressed by expanding (x25) and smoothing C′, and then using the phase of C′ (Fig. 1e) to map expanded and distorted control, label, and M0 magnitude image pixels to their undistorted locations. These corrected images were then rebinned to original size.

CBF Quantificaton: Distorted and distortion-corrected label and control images were corrected for motion artifact in FSL 5.0, and then registered to the corresponding averaged M0 images. Next, subtraction images $$$\triangle M$$$ (control – label) were calculated using a sliding window approach for improved SNR. Quantitative CBF maps were obtained using Equation (1), $$CBF = \frac{6000\cdot\lambda\cdot\triangle M\cdot e^\frac{PLD}{T_{1,blood}}}{2\cdot M0 \cdot \alpha\cdot T_{1,blood} (1-e^\frac{-\tau}{T_{1,blood}})} (1)$$ where $$$T_{1,blood}$$$ = 1.65s, the blood-brain partition coefficient $$$\lambda$$$ = 0.9, and the labeling efficiency $$$\alpha$$$ = 0.856. Finally, coronal and sagittal reformats of the CBF maps were generated.


All four subjects’ pCASL images showed similar distortion correction, although only the subject shown included acquisition of M0 data and quantification of CBF. The green dotted line in Fig. 2 indicates common regions of the subject over several images, with a T1-weighted image used for reference. Red arrows indicate image distortion in the uncorrected single-dynamic EPI image and ASL difference map. Corresponding PLACE-corrected images do not show these distortions, and conform well to the anatomical T1-weighted reference image. Figure 3 depicts the uncorrected and PLACE-corrected quantitative CBF maps of the subject in three projections. Axial images and their coronal and sagittal reformats indicate differences in pixel-by-pixel flow quantification following distortion correction.


The results show eliminated distortion with PLACE in both the acquired EPI scans and in the final quantified CBF maps. This correction permits accurate quantification of CBF, and requires minimal modification to the pulse sequence. Since pCASL already employs multiple dynamics to ensure sufficient SNR, PLACE does not require additional image acquisition, since associated gradients may be added to any of the dynamics. Further studies will investigate the effects of flow on PLACE-correction and subsequent CBF quantification.


The use of PLACE to correct distortion in pCASL provides a useful tool for accurate localization of CBF-weighted signal, and may in general improve the clinical applicability of ASL.


No acknowledgement found.


1. Buxton et al., MRM, 40:383-96, 1998.

2. Wu WC et al., MRM, 58:1020-1027, 2007.

3. Xiang Q-S & Ye FQ, MRM, 57:731-741, 2007.

4. Lu HB et al., MRM, 64:1557-1566, 2010.

5. Garcia DM et al., Proc. ISMRM, 05, 2005.

6. Alsop DC et al., MRM, 73:102-116, 2015.


Fig.1: PLACE distortion correction. a. EPI magnitude image. b. Even-dynamic EPI phase image. c. Odd-dynamic EPI phase image, added PLACE gradient induces a phase ramp along PE direction. d. Phase difference (c-b) map. e. Displacement map computed by applying a linear phase ramp that opposes the ramp in d.

Fig.2: PLACE distortion corrections of pCASL axial images. a. T1-weighted reference image. b. Single dynamic ASL-EPI image, uncorrected. c. Single dynamic ASL-EPI image, PLACE-corrected. d. ΔM ASL difference image, uncorrected. e. ΔM image, PLACE-corrected. Green dotted line provides anatomical reference to T1 image, and red arrows indicate pronounced distortion.

Fig.3: Quantitative cerebral blood flow maps (CBF) with a color key indicating flow rate. Axial maps as well as coronal and sagittal reformats are shown for both a. uncorrected and b. PLACE-corrected images.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)