3.1 Introduction to Neuroimaging

The idea of localisation of function within the brain has only been accepted for the last century and a half. In the early 19th century Gall and Spurzheim, were ostracised by the scientific community for their so-called science of phrenology[1]. They suggested that there were twenty-seven separate organs in the brain, governing various moral, sexual and intellectual traits. The importance of each to the individual was determined by feeling the bumps on their skull. The science behind this may have been flawed, but it first introduced the idea of functional localisation within the brain which was developed from the mid 1800's onwards by clinicians such as Jackson [2] and Broca[3]. Most of the information available on the human brain came from subjects who had sustained major head wounds, or who suffered from various mental disorders[4]. By determining the extent of brain damage, and the nature of the loss of function, it was possible to infer which regions of the brain were responsible for which function.

Patients with severe neurological disorders were sometimes treated by removing regions of their brain. For example, an effective treatment for a severe form of epilepsy involved severing the corpus callosum, the bundle of nerve fibres which connect left and right cerebral hemispheres. Following the surgery patients were tested, using stimuli presented only to the left hemisphere or to the right hemisphere[5]. If the object was in the right visual field, therefore stimulating the left hemisphere, then the subject was able to say what they saw. However if the object was in the left visual field, stimulating the right hemisphere, then the subject could not say what they saw but they could select an appropriate object to associate with that image. This suggested that only the left hemisphere was capable of speech.

With the development of the imaging techniques of computerised tomography (CT) and magnetic resonance imaging it was possible to be more specific as to the location of damage in brain injured patients. The measurement of the electrical signals on the scalp, arising from the synchronous firing of the neurons in response to a stimulus, known as electroencephalography (EEG), opened up new possibilities in studying brain function in normal subjects. However it was the advent of the functional imaging modalities of positron emission tomography (PET), single photon emission computed tomography (SPECT), functional magnetic resonance imaging (fMRI), and magnetoencephalography (MEG) that led to a new era in the study of brain function.

In this chapter the mechanisms of the techniques mentioned above are outlined, together with an assessment of their strengths and weaknesses. Then an introduction to the physical and structural anatomy of the brain is given, and some of the terminology used in neurology is introduced. This is followed by a more detailed explanation of functional MRI and how such experiments are performed.

3.1.1 SPECT and PET

The imaging modalities of single photon emission computed tomography (SPECT) and positron emission tomography (PET) both involve the use of radioactive nuclides either from natural or synthetic sources. Their strength is in the fact that, since the radioactivity is introduced, they can be used in tracer studies where a radiopharmaceutical is selectively absorbed in a region of the brain.

The main aim of SPECT as used in brain imaging, is to measure the regional cerebral blood flow (rCBF). The earliest experiments to measure cerebral blood flow were performed in 1948 by Kety and Schmidt[6]. They used nitrous oxide as an indicator in the blood, measuring the differences between the arterial input and venous outflow, from which the cellular uptake could be determined. This could only be used to measure the global cerebral blood flow, and so in 1963 Glass and Harper[7], building on the work of Ingvar and Lassen[8], used the radioisotope Xe-133, which emits gamma rays, to measure the regional cerebral blood flow. The development of computed tomography in the 1970's allowed mapping of the distribution of the radioisotopes in the brain, and led to the technique now called SPECT[9].

The radiotracer (or radiopharmaceutical) used in SPECT emits gamma rays, as opposed to the positron emitters used in PET. There are a range of radiotracers that can be used, depending on what is to be measured. For example I-123-3-quinuclidinyl 4-isodobenzilate is a neurotransmitter agonist which can be used for imaging receptors. For rCBF measurements Xe-133 can be introduced into the blood stream by inhalation. Detection is carried out using a gamma camera - a scintillation detector consisting of a collimator, a NaI crystal, and a set of photomultiplier tubes as shown in Figure 3.1.

Figure 3.1 Schematic diagram of a gamma camera used in SPECT imaging

By rotating the gamma camera around the head, a three dimensional image of the distribution of the radiotracer can be obtained by employing filtered back projection (as described in section 2.3.4). The radioisotopes used in SPECT have relatively long half lives (a few hours to a few days) making them easy to produce and relatively cheap. This represents the major advantage of SPECT as a brain imaging technique, since it is significantly cheaper than either PET or fMRI. However it lacks good spatial or temporal resolution, and there are safety aspects concerning the administration of radioisotopes to the subject, especially for serial studies.

Positron emission tomography[10] has two major advantages over SPECT, namely better spatial resolution and greater sensitivity. This comes from using positron emitters such as O-15 and F-18 as the radionuclide. When such nuclei decay they emit a positron, that is a particle with the same rest mass as an electron but with charge +1. Once the positron is emitted it travels a short distance before colliding with an electron. The annihilation of the two particles creates two photons each with energy 511 keV. In order to conserve momentum, the two photons are emitted at virtually 180 degrees to each other, and it is these photons that are detected in a ring of scintillators and photomultiplier tubes surrounding the head (Figure 3.2).

Figure 3.2 Schematic diagram of a PET scanner. A positron is emitted from a radioisotope in the brain. The positron annihilates with an electron, producing two photons emitted at 180 degrees to each other

Opposite pairs of detectors are linked so as to register only coincident photons, thus defining a set of coincidence lines. Reconstruction of these lines by filtered back projection gives an image of the source of the annihilation. Since the detectors only record the site of the annihilation, resolution in a PET scanner is limited by the distance travelled by the positron through the tissue before it meets an electron. This fundamentally restricts the resolution of the scanner to 2 - 3 mm at best.

The positron emitters used in PET have short half lives, of the order of 2 - 100 minutes. This means that the isotopes must usually be made at the site of the scanner, using an expensive cyclotron. However, this short half life means that dynamic studies of brain function can be carried out using the technique.

Functional imaging studies using PET first appeared in 1984 with a study using C15O 2. Usually two cognitive states are imaged, one active and one resting, and by subtracting these two states a map of the regions of the brain responsible for that task made. PET is widely used as a tool for cognitive function, and much of the literature on brain function published in the last 10 years has used the technique. Positron emitting neurotransmitters such as F-18 labelled DOPA, mean that dopamine function can be followed in patients with Parkinson's disease. The main drawback of PET is it's use of radioisotopes, and its very high cost, however it's neurotransmitter mapping ability means that it will retain a role in the face of the less invasive, and more available fMRI.

3.1.2 EEG and MEG

Measuring the electrical signals from the brain has been carried out for several decades [11], but it is only more recently that attempts have been made to map electrical and magnetic activity. The electroencephalogram (EEG) is recorded using electrodes, usually silver coated with silver chloride, attached to the scalp and kept in good electrical contact using conductive electrode jelly. One or more active sites may be monitored relative to a reference electrode placed on an area of low response activity such as the earlobe. The signals are of the order of 50 microvolts, and so care must be taken to reduce interference from external sources, eye movement and muscle activity. Several characteristic frequencies are detected in the human EEG. For example, when the subject is relaxed the EEG consists mainly of frequencies in the range 8 to 13 Hz, called alpha waves, but when the subject is more alert the frequencies detected in the signal rise above 13Hz, called beta waves. Measurements of the EEG during sleep have revealed periods of high frequency waves, known as rapid eye movement (REM) sleep which has been associated with dreaming.

The EEG can also be measured in response to some regularly repeated stimulus such as the pattern reversal of a projected checkerboard, or more complicated task such as number memorising and recall. The signal, from each electrode, in response to the stimulus is recorded and averaged together over a number of trials. The response is characterised by the delay, or latency, of the peak of the signal from the presentation of the stimulus. This is of the order of milliseconds for brain stem responses and 100's of milliseconds for cortical responses. For example a commonly detected response to an oddball stimulus, the P300, has a latency of 300 ms. Having found an electrical signal of interest, the magnitude of the peak can be mapped across the scalp giving an idea of the source. The major limitation of studying brain function using EEG is that the signals measured are recorded on the scalp, which may not represent the activity in the underlying cortex. However since the technique is very cheap and safe, it has many uses and can be involved in studies where the scanning techniques could not, such as continuous monitoring during sleep.

Since the magnetic signals from the firing of the neurons do not need to be conducted to the scalp, measuring them gives much greater signal localisation than measuring the currents themselves. This is the basis of the much newer technique of magnetoencephalography (MEG). The first successful measurements of these magnetic fields were made by Cohen in 1972 [12]. The magnitude of the magnetic signals, resulting from the electrical firing of the neurons is of the order of 10-13 Tesla, for 1 million synchronously active synapses. Such small signals are picked up using a superconducting quantum interference device (SQUID) magnetometer (Figure 3.3).

Figure 3.3 Schematic diagram of a magnetometer for measuring MEG's

Interference from external sources is a major problem with MEG, and the magnetometer must be sited in a magnetically shielded room. Current magnetometers have as many as 120 SQUIDS covering the scalp.

MEG experiments are carried out in much the same way as their EEG counterpart. Having identified the peak of interest, the signals from all the detectors are analysed to obtain a field map. From this map an attempt can be made to ascertain the source of the signal by solving the inverse problem. Since the inverse problem has no unique solution, assumptions need to be made, but providing there are only a few activated sites, close to the scalp then relatively accurate localisation is possible, giving a resolution of the order of a few millimetres.

MEG has the advantage over EEG that signal localisation is, to an extent, possible, and over PET and fMRI in that it has excellent temporal resolution of neuronal events. However MEG is costly and its ability to accurately detect events in deeper brain structures is limited.

3.1.3 Functional MRI and MRS

Since functional magnetic resonance imaging (fMRI) is the subject of this thesis, little will be said in this section as to the mechanisms and applications of the technique, as this is covered later in the chapter. The purpose of this section is to compare fMRI to the other modalities already mentioned, and also to consider the related, but distinct technique of magnetic resonance spectroscopy (MRS).

During an fMRI experiment, the brain of the subject is scanned repeatedly, usually using the fast imaging technique of echo planar imaging (EPI). The subject is required to carry out some task consisting of periods of activity and periods of rest. During the activity, the MR signal from the region of the brain involved in the task normally increases due to the flow of oxygenated blood into that region. Signal processing is then used to reveal these regions. The main advantage of MRI over its closest counterpart, PET, is that it requires no contrast agent to be administered, and so is considerably safer. In addition, high quality anatomical images can be obtained in the same session as the functional studies, giving greater confidence as to the source of the activation. However, the function that is mapped is based on blood flow, and it is not yet possible to directly map neuroreceptors as PET can. The technique is relatively expensive, although comparable with PET, however since many hospitals now have an MRI scanner the availability of the technique is more widespread.

FMRI is limited to activation studies, which it performs with good spatial resolution. If the resolution is reduced somewhat then it is also possible to carry out spectroscopy, which is chemically specific, and can follow many metabolic processes. Since fMRS can give the rate of glucose utilisation, it provides useful additional information to the blood flow and oxygenation measurements from fMRI, in the study of brain metabolism.

3.1.4 Comparison of the Functional Brain Imaging Modalities

The brain imaging techniques that have been presented in this chapter all measure slightly different properties of the brain as it carries out cognitive tasks. Because of this the techniques should be seen as complementary rather than competitive. All of them have the potential to reveal much about the function of the brain and they will no doubt develop in clinical usefulness as more about the underlying mechanisms of each are understood, and the hardware becomes more available. A summary of the strengths and weaknesses of the techniques is presented in Table 3.1.

Technique Resolution Advantages Disadvantages
SPECT 10 mm Low cost
Limited resolution
PET 5 mm Sensitive
Good resolution
Metabolic studies
Receptor mapping
Very expensive
EEG poor Very low cost
Sleep and operation monitoring
Not an imaging technique
MEG 5 mm High temporal resolution Very Expensive
Limited resolution
for deep structures
fMRI 3 mm Excellent resolution
Limited to activation studies
MRS low Non-invasive
metabolic studies
Low resolution
Table 3.1 Comparison of modalities for studying brain function

Previous Section | Contents | Next Section