Detecting functional activation separated by periods up to 4 weeks by arterial spin labeling

Introduction: The ability of arterial spin labeling (ASL) to measure cerebral blood flow (CBF) independent of task frequency makes it attractive for longitudinal studies to follow disease progression or track treatment response1. However, ASL's poor spatial resolution and low signal-tonoise ratio makes image alignment from separate sessions difficult. Registration errors will reduce sensitivity by increasing variance and Type I errors. Variations in basal CBF can further reduce reproducibility over longer periods of time2. The aim of this study was to quantify between-session variance and demonstrate that with the appropriate steps, ASL can detect activation-induced changes in regional CBF over periods extending up to a month. Methods: Seven right-handed volunteers (22.7 _ 1.3 years, 2 male) were scanned during three sessions separated by a week and a month. Registration errors between sessions were minimized by creating an immobilizing head mold for each subject during the first session, which was reused in the following sessions to replicate head position. Each session consisted of two sets of rest and sequential finger tapping task epochs (6 min, 48 label/control pairs). pCASL GRASE3 images (1.5s label duration, 1.2s post-label delay, 24 axial slices, matrix = 64 x 64, FOV = 24cm) were acquired on a Siemens 3.0T Biograph system. Data were processed using SPM8 (UCL, London, UK). Variability and reproducibility of CBF were assessed using within-session coefficient of variance (wsCV) and intra-class correlation coefficient (ICC), respectively. In addition, statistical parametric maps were created from rest and task data acquired in the same session and from data acquired in different sessions (i.e. task and rest periods separated by a week and a month). To remove the variability in basal CBF, data were scaled by their respective resting greymatter CBF. Areas of activation were identified after correction for multiple comparisons using the family-wise error rate (p < 0.05). Results: ICC for within and between-session was 0.86 and 0.62, respectively, and wsCV was 9.1% and 10.0% respectively (Fig 1). Removing variability in basal CBF increased the within and between-session ICC to 0.873 and 0.781, and decreased the wsCV to 4.71% and 5.74% respectively. Absolute and normalized gray matter CBF activation maps from within and between-session analyses are shown in Figure 2.Discussion: Variance and ICC values were similar to other studies4. wsCV images showed some regional heterogeneity, but only a marginal increase when compared to between-session images. This is reflected in the comparison of activation maps generated from rest and activation images from the same session and from data acquired on separate days. The remarkable similarity in these maps after removing the variability in resting CBF indicates that registration errors between sessions were minimal. These results demonstrate the feasibility of conducting voxel-wise analysis of CBF images acquired on different days (in this case, a month apart) and highlight the potential of this technique for longitudinal studies.