Measuring arterial and tissue responses to functional challenges using arterial spin labeling Yi-Ching Lynn Ho a,b, ⁎, Esben Thade Petersen a,b , Xavier Golay a,c,1 a Department of Neuroradiology, National Neuroscience Institute, Singapore b Centre for Functionally Integrative Neuroscience, University of Aarhus, Aarhus, Denmark c Laboratory of Molecular Imaging, Singapore Bioimaging Consortium, Agency for Science, Technology and Research (A⁎STAR), Singapore abstract article info Article history: Received 19 December 2008 Revised 17 July 2009 Accepted 17 July 2009 Available online 25 July 2009 The measurement of cerebral blood flow (CBF) in functional MRI studies that aim for non-invasive, quantitative and reliable measurements is a challenge. Here, we tested the feasibility of a recently developed, model-free CBF technique to study vascular dynamics upon functional challenges. Multiple inversion time- point signals were measured from arterial and tissue compartments, allowing for the calculation of CBF through a process of deconvolution. Using graded visual stimulation known to produce increasing hemodynamic responses, we recorded significant and graded ΔCBF and Δτ m (microvascular arrival time change) that were highly comparable to those estimated by a standard 3-parameter fit based on the general kinetic model, though the absolute values had weaker agreement. Estimated arterial blood volumes (excluding substantial arteriolar contribution) did not show significant change with visual stimulation. Bolus arrival times in the microvascular compartment shortened more as compared to the arrival times from the arterial compartment during visual stimulation, suggesting larger involvement of the microvasculature in local neuronal response. While there are limitations, the model-free analysis method has the potential to offer useful vascular information in fMRI studies. © 2009 Elsevier Inc. All rights reserved. Introduction The measurement of cerebral blood flow (CBF) in functional MRI studies is gaining momentum, due to advances and accessibility of non-invasive, arterial spin labeling (ASL) techniques (for a review, see Petersen et al., 2006b). There is furthermore, recognition of the need for quantitative fMRI, since it is known that the amplitude of the common BOLD-fMRI signal depends on underlying vascular contribu- tions like CBF and CBV (cerebral blood volume) (Ogawa et al., 1993). Nonetheless, it remains a challenge to get robust estimates of CBF using ASL during brain activation. CBF is typically quantified based on tracer kinetics theory with a single compartment, originally proposed by Kety and Schmidt (1948) and later extended by Buxton et al. (1998) as a general kinetic model for ASL data. The general kinetic model takes into account the magnetization differences between the control and labeled experi- ments in ASL, for which the travelling time of the magnetized blood is an important factor. Among some of the complexities with ASL techniques, the time taken by the leading and trailing edges of the labeled arterial blood to arrive at the capillary exchange site has been shown to reduce with positive functional activity, as pointed out by several studies using motor and visual paradigms (Buxton et al., 1998; Gonzalez-At et al., 2000; Hendrikse et al., 2003; Yang et al., 2000). These studies pointed out the error of assuming unchanging arrival times if a single inversion time-point acquisition was used, as had been commonly done, until methods such as Q2-TIPS (Luh et al., 1999) and QUIPSS II (Wong et al., 1998) were developed to improve this issue. Inter- and intra-subject variations in arterial, arteriolar and capillary (further defined in the text as “pre-venous”) vascular responses may further compound the problem. The issue of dynamic arrival times can be circumvented through the use of multiple post-labeling delay times (TI) in a single ASL scan with repeated, small flip-angle acquisitions (Gunther et al., 2001), a method which Hendrikse et al. (2003) modified to study perfusion in human visual stimulation. In that latter study, the authors separated arterial hemodynamics from the tissue perfusion by maximum cross correlation of the visual activation with post- labeling (TI) delay. Maximal activation at short TIs (200–400 ms) characterized the voxels with large arterial input, while maximal activation at longer TIs (600–1000 ms) characterized voxels with relatively more tissue contribution. While compelling, this separa- tion is arbitrary and depends on several factors, such as the distance of each imaged voxel from the labeling slab and more importantly, the fraction and composition of the vascular system as covered by individual voxels. A simple solution proposed by Ye et al. (1997) is to use bipolar crusher gradients to exclude the inflowing arterial signal, particularly NeuroImage 49 (2010) 478–487 ⁎ Corresponding author. Department of Neuroradiology, National Neuroscience Institute, 11 Jalan Tan Tock Seng, 308433, Singapore. Fax: +65 6358 1259. E-mail address: yiching.lynn.ho@gmail.com (Y.-C.L. Ho). 1 UCL Institute of Neurology, Queen Square, London, UK. 1053-8119/$ – see front matter © 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.neuroimage.2009.07.040 Contents lists available at ScienceDirect NeuroImage journal homepage: www.elsevier.com/locate/ynimg