7A Introduction
The aim of this experiment was to measure cerebrovascular reactivity (CR) and baseline CBF, inflow delay and BOLD signal in normals over the course of a day. Secondly, this experiment also served as a pilot study to evaluate the feasibility of longitudinal CBF studies using the FAIR technique.
Longitudinal CBF studies are only possible if the changes in the parameter of interest that occur from session to session are predominantly physiological and not determined by volunteer repositioning, machine drifts and such (the so-called ‘session effects’). For a single-slice FAIR experiment, it is predominantly inter-session movement (repositioning errors) that results in session effects. A specially designed repositioning device was therefore used to minimise repositioning errors and volunteer movement in general.
The following questions are the basis for this study:
- What is the variation of CR and baseline BOLD, CBF and inflow delay over the course
of a day?
- Are there any (obvious) patterns in the diurnal BOLD/CBF/inflow delay/CR
fluctuations?
To my knowledge, there are no reports of baseline measurements of BOLD, CBF or inflow delay over the course of a day. PET, the main quantitative CBF technique so far, is limited in its utility for longitudinal studies due to the repeated radioactive label injections that would be required. This experiment is therefore of interest simply to establish the diurnal variability of these parameters.
The CR over the course o f a day, however, has been the topic of several studies (Ameriso et al., 1994; Placidi et al., 1998; Qureshi et al, 1999). This is because the CR is seen as an indicator of vascular reserve, and a lowered CR as a warning signal for stroke.
Ameriso and colleagues (Ameriso et al, 1994) have reported a significant (25%) decrease in relative CR in the morning compared to the afternoon and evening. This measurement was done using transcranial Doppler. Their findings fit in nicely with the clinical observation that stroke occurs more often in the early hours of the morning.
Placidi and colleagues and Qureshi and colleagues have replicated this morning CR reduction for patient groups with sleeping disorders (snoring) that are thought to be more at risk for stroke. Placidi also studied a control group of normals, for which he did not find a significant reduction in relative CR. Placidi studied quite a small group (8 volunteers), while Ameriso studied 20 subjects. A tentative hypothesis for this study was therefore that there is a reduction in CR in the early morning session compared to the afternoon and evening. All three aforementioned studies were measuring relative CR with Doppler; for this diurnal study in which absolute CR was measured using FAIR, the sensitivity to changes in CR had yet to be established.
7.2 Intersession movement effects
7.2.1 Minimising inter-session movement
A special re-positioning device was built in-house by Peter Aston. It was designed to minimise inter-session movement, i.e. repositioning errors and other volunteer motion. In particular, translations along the magnet bore and nodding were to be avoided, as these are out-of-plane movements for the axial slice o f interest. This device is pictured in Figure 7-1. The device is clamped to the rf coil at the back and fastened with screws. The top of the volunteer’s head touches an arc so that repositioning errors and volunteer movement along the bore are minimised. A blunt pin touches the volunteer’s forehead: a circle is drawn around this pin at the beginning of the first session. A black cross is then also drawn on the volunteer’s forehead; this cross is defined by the scanner’s slice positioning laser beams. At the beginning of each new session the head is repositioned until the pin falls within the circle again and the black cross and laser beams overlap. Furthermore, standard rf coil side-clamps are used to minimise movements sideways and the head rests in head shaped black cushioning. The volunteer’s chin is supported by a neck brace, to prevent nodding movements further.
pin to mark position forehead (to avoid nodding)
arc that rests on top head (to prevent movement along bore magnet)
7.2.2 The effects out-of-plane motion; choice of masks
The chances of significant out-of-plane movements are much larger for multi-session studies, mainly due to repositioning errors. As the FAIR technique implemented here only measures a single slice, a control experiment was performed to establish the effects o f an out-of-plane shift. To that end, two volunteers were scanned following the imaging and analysis procedures of section 5.4.7. A rest state CBF measurement was made, the imaging slice was then shifted downwards by 5mm (= slice thickness) and the CBF was measured again for the new slice. Thus the effect of a repositioning error was mimicked, and a ‘multi session’ data set generated with an exactly known out-of-plane shift between the two ‘sessions’.
Assuming CBF to vary minimally over the two back-to-back CBF measurements (~ 45 minutes), the CBF values should be very similar for the first and second slice (the first and second ‘session’), if the grey matter/white matter/CSF composition of the voxels in the masks of interest does not change dramatically over the 5 mm shift. Table 7-1 gives the absolute mean CBF and inflow delay differences (A) between the two sessions, for the global and grey matter masks. These global and grey matter masks were constructed for each ‘session’ following the criteria in section 5.4.7.
The global masks give smaller differences between sessions. This is to be expected, as any changes in partial voluming are more likely to average out over the large ensemble. The disadvantage of a global mask is that it will contain a mix of white and grey matter.
The grey matter masks are defined on the basis o f their Tlapp characteristics, CSF fraction and fit convergence as defined in section 5.4.7. These masks therefore use no prior knowledge of brain CSF and grey and white matter distributions. There is a way of
incorporating this knowledge into the mask definition: by using information from
segmented structural data, separately acquired. Using the sequence parameters listed in the ‘Materials and methods’ section below, whole brain structural and BOLD data were therefore also acquired for each session. If one coregisters the segmented structural data separately to each session’s BOLD volume, a grey matter ‘structural’ mask can be defined per session consisting of voxels with a certain percentage grey matter. This method
assumes there is no significant motion between the FAIR and the BOLD acquisitions of each session, as these can not be coregistered to each other.
The results for these grey matter - structural (‘grey-struc’) masks are given in Table 7-2. The percentage of grey matter in each mask has been chosen in such a way as to give (close to) equal numbers of voxels to the standard grey matter masks used before, to facilitate comparisons. Voxels in the grey-struc masks that have not given converging fit results or have a fitted CBF > 200 ml/lOOg/min (vessels) were nulled, as is done by default for the standard grey matter masks.
volunteer 1 volunteer 2 grey matter mask (85/106) global mask (522/524) grey matter mask (95/111) global mask (580/601) |A CBF| in ml/lOOg/min 8.46(10%) 3.80 (5%) 1.22 (2%) 0.31 (1%) |A inflow delay] in s 0.08 (19%) 0.06(13%) 0.01 (2%) 0.03 (6%)
Table 7-1 Mimicking the effects of repositioning errors between sessions: changes in CBF and inflow delay for two volunteers after a 5 mm out-of-plane shift o f the imaging slice. The number of voxels in each sessions’s mask is given in brackets below the mask name. The difference in CBF or delay as percentage of its start value is given in brackets behind the difference.
volunteer 1 volunteer 2 grey-struc mask, 85% (100/96) grey-struc mask, 87% (100/106) |A CBF| in ml/lOOg/min 0.76(1%) 4.00 (7%) |A inflow delay] in s 0.01 (2%) 0.06(13%)
Table 7-2 Using masks derived from structural scans, looking at the effects o f repositioning errors
between sessions: changes in CBF and inflow delay for two volunteers after a 5 mm out-of-plane shift of the imaging slice. The number of voxels in each session’s mask is given in brackets below the mask name. The grey matter percentage in the mask name is derived from the structural segmentation results. The difference in CBF or delay as percentage of its start value is given in brackets behind the difference.
For these different masks, the changes in CBF for a 5 mm out-of-plane shift range from 1- 10% and for inflow delay from 2-19%. The grey-struc masks are doing slightly better on average, with maximum changes in CBF of 7% and in inflow delay of 13%. These
measured CBF and inflow delay differences can be taken as an estimate of the minimum change in these parameters that is required before it can be considered to be physiologically significant. This estimate is valid only if out-of-plane shifts from session to session do not exceed 5 mm.
It is not immediately clear which is the optimal grey matter mask to use to minimise the impact of shifts, while still sampling a reasonably pure grey matter segment. Apart from minimising differences in CBF and delay for these two sessions (which is not a 100% reliable method, as that makes the assumption that the minimal CBF and delay difference is the correct one), it is also informative to look at GL inversion data over sessions.
GL data are independent of CBF and should therefore vary little over sessions. The mean sum of square differences between GL data over sessions for a mask is an indication of how well that mask keeps the mean voxel intensity constant. Considering this analysis includes GL intensity data over the whole TI range, this means this favours masks that select voxels with similar voxel Tiapp values over sessions.
An overview of the sums of squared differences between GL arrays of different sessions (for all TI values) is given in Table 7-3 for the standard grey matter and the grey- struc masks. Apart from data of the two volunteers above, this Table also includes results for two volunteers of the 4 session study described below in section 7.3. As these are GL inversion data, this analysis is independent from the CBF results described later on.
volunteer 1 volunteer 2 volunteer 3 volunteer 4 total
grey matter 61.52 15.40 2851 271.0 3.199-10^ (100%) grey-struc mask, 86% 5.704 21.66 2250 130.3 2.408 10^ (75%)
Table 7-3 Sum of squared differences (/lO in a.u.) between GL inversion data (for all TI values) of different sessions for standard grey matter and 86% grey-struc masks.
Overall, the grey-struc masks give a 25% lower sum o f squared differences between GL arrays o f different sessions. O f course motion will affect all voxels in the imaging slice and the choice of voxel mask has nothing to do with that. However, it might be that the added