INSTALACIÓN DEL MÓDULO WISM2

Most fMRI groups have adopted a ‘box-car’ method of task and rest presentation frequencies. In its simplest form, this consists of simply alternating the two conditions (task and rest) throughout the experiment, ensuring that a similar number of task and rest images are acquired. These studies usually assume that sustained activation occurs during each cycle; however this may not be the case for all areas of the brain and so the paradigm must be carefully considered for each study. One of the major advantages of the method is that it allows non-stimulus correlated motion and signal intensity drifts (such as cardiac, respiratory and other motion artefacts) in the image data to be filtered out, as they are most likely to occur at entirely different frequencies; this increases the statistical power of the subsequent analysis (see Section 3.2.1). One debated issue using this method is the duration (and so the frequency) of the task/rest states. There are a number of points that need to be considered:

• As mentioned above, the haemodynamic response to activation takes between 5 and 8 seconds to reach a steady state. Each 3-D EPI scan (covering the whole head in

Chapter 3: fM R I methodology

one acquisition) takes 5.6 seconds to acquire. Therefore each block must be sufficiently long for the haemodynamic changes to reach a constant level.

• Several studies have suggested that with an extended period of stimulation (from approximately 2-5 minutes after the start of the stimulus) a gradual decrease in the signal change can be seen over the period (Hathout et al. 1994; Frahm et al. 1996; Kruger et al. 1996). This has been explained either by a decrease in neuronal firing with subject habituation or fatigue effects and/or by an increase in oxidative metabolic rate over time which recouples perfusion and oxygenation consumption at a new equilibrium. Contradictory results, however, have shown no evidence of a ‘vanishing’ BOLD signal during sustained stimulation, strengthening the argument for using longer epoch lengths (Bandettini et al. 1995; Kollias et al. 1996; Bandettini et al. 1997; Chen et al. 1998). It has been demonstrated that this discrepancy cannot simply be ascribed to different imaging methods (Howseman et al. 1996, 1998). • It must also be considered that the haemodynamic response function may vary

across brain regions. As mentioned above, paradigms must be carefully chosen to take account of all functional cortex responses as much as possible.

One study has examined the relationship between the activation voxel number and performance time length (duration of task/rest state) during a motor paradigm (Liu et al. 1999). These authors found that performance lengths of 10, 30, and 40 secs produced similar number of activated voxels; however periods as long as 60 secs resulted in subject fatigue, caused more motion artefacts, and also resulted in fewer activated voxels. On the other hand, periods as short as 3-5 secs were easier to perform for the subjects, but produced fewer activated voxels due to less signal change.

The aim of the study described below was to compare two durations of motor task/rest states (5 scans on/off and 10 scans on/off, with a duration of 29 seconds and 58 seconds, respectively). The regions of activation and the suitability of the duration of each state are compared between the two frequencies.

3.3.2.1.1 Methods

Four volunteers participated in the study (3 males, 1 female; mean age 32 years). Two experiments were carried out to compare the fMRI activation maps associated with two task/rest durations (two frequencies), namely 5 scans (29 seconds) and 10 scans (58 seconds). These epoch lengths were chosen as they are long enough to ensure that a prolonged state of detectable activation and so an adequate signal change is reached, but

Chapter 3: fMRI methodology

the subject would not be expected to exhibit large habituation, tiredness or boredom effects during each phase.

A total of 60 scans were acquired per experiment, using the 3-D EPI sequence described in Section 3.3.1. The same task was carried out in both experiments. The task consisted of sequential movement of each finger to the thumb with the right hand at a rate of 2Hz, cued by LED lights set inside goggles fixed on top of the head coil which were left switched on throughout the session. In addition, a T1-weighted axial anatomical scan (TR=31ms, TE=llm s, flip angle=40“, matrix size=256x256, 64 slices, FOV=192mm) was acquired at the beginning of the session, covering the same region as the EPI slab, to permit localisation of significant regions of activation. Datasets acquired in the same session were analysed together using SPM. Images were displayed at an uncorrected p- value threshold of p<().01 and an extent threshold of p<().05.

3.3.2.1.2 Results

All four volunteers demonstrated contralateral sensorimotor cortex flVIRI activation on movement of the hand in all experiments. Examples of the results in two of the subjects are shown in Figure 3.8 and Figure 3.9. There was no visible difference in the extent or location of the sensorimotor cortex activation between the two frequencies. Practically, subjects reported they found a frequency of 5 datasets on/off easier to perform than 10 on/off as they did not feel as fatigued in their hand.

Figure 3.8: Example of contralateral sensorimotor cortex IMRI activation in one axial slice on right hand active movement in one subject acquired using a task/rest frequency of 10 scans (a) and 5 scans (b). SMA activation close to the longitudinal fissure is also seen in the images.

Chapter 3: fMRI methodology

Figure 3.9: Example of contralateral sensorimotor cortex fMRI activation in one axial slice on right hand active movement in one subject acquired using a task/rest frequency of 10 scans (a) and 5 scans (b).

3.3.2.1.3 Conclusions

As described above (Section 3.3.2.1), there are several advantages for selecting as short a task/rest frequency as possible, but also one that is also long enough to ensure that data are collected within a state in which the haemodynamic changes have reached a constant level. This experiment has demonstrated that there was no consistent visible difference between sensorimotor cortex activation using two task/rest frequencies of 5 and 10 scans per epoch. Unlike the study by Liu et al. (1999), we have not shown a significant increase in motion artefacts or a decrease in activated voxels in on/off frequencies of 10 datasets compared to 5. However, the subjects did report fatigue occurring at longer performance periods, in agreement with Liu et al. (1999). For the purpose of this study, therefore it was decided that a task/rest frequency of 5 scans per epoch (29 seconds in each state) would be used for the acquisition of data in adult and child controls and children with brain disease.