Skip Navigation

This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (12)
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Turner, R.
Right arrow Articles by Jones, T.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Turner, R.
Right arrow Articles by Jones, T.
Related Collections
Right arrow Neurology
Right arrow Neurophysiology
Right arrow Radiology
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

British Medical Bulletin 65:3-20 (2003)
© 2003 The British Council

Techniques for imaging neuroscience

Robert Turner* and Terry Jones{dagger}

* Wellcome Department of Imaging Neuroscience, Institute of Neurology, University College London, London, UK
{dagger} Academic Department of Radiation Oncology, Christie Hospital NHS Trust, Manchester, UK


   Abstract
Top
 Abstract
Functional magnetic resonance...
 References
 
In the last 20 years, a number of non-invasive spatial mapping techniques have been demonstrated to provide powerful insights into the operation of the brain during task performance. These are, in order of their emergence as robust technologies: positron emission tomography, source localization with EEG and MEG, and functional magnetic resonance imaging. The imaging neuroscience study areas represented in this volume use the first or last of these – PET and fMRI. The physical principles underlying both of these techniques are outlined, and the important assumptions and limitations are made explicit. The range of applications for each is briefly indicated.


   Functional magnetic resonance imaging (fMRI)
Top
 Abstract
 Functional magnetic resonance...
References
 
Basic mechanism

As its name implies, functional magnetic resonance imaging (fMRI) is a technique based on MRI1. This provides images of the distribution of hydrogen nuclei (protons) in tissue water. When the tissue is immersed in a very uniform, large, steady magnetic field, the protons can be resonantly excited by a radiofrequency magnetic field of the appropriate frequency. The tiny voltage induced in a receiver coil by the precessing protons is detected with sensitive radiofrequency electronics, and converted into a high resolution image of the pre-selected slice of the tissue. The MR image intensity, depends mainly on the density of water protons, and is modulated by intrinsic local properties of the tissue such as biochemical composition and magnetic heterogeneity. The latter effect is of particular importance for imaging neuroscience.

In 1990, a Bell Laboratories physicist, Seiji Ogawa, noticed that the visibility of veins in MR images of rat brain at the high field of 4.7 T was enhanced when the rat's blood oxygenation was reduced2. Veins appear as dark lines in T2* weighted images, sensitive to local variations in the static magnetic field. He realized that this enhancement was due to the greater magnetic susceptibility of deoxyhaemoglobin as compared with tissue and oxyhaemoglobin, previously noted as early as 19363, which thus acts as an endogenous paramagnetic MRI contrast agent. Ogawa's results were followed up in a study by Turner4, using a feline model, which showed that such changes in vascular contrast could be monitored in real time and without motion-related artefact by means of a very fast MRI technique known as echo-planar imaging (EPI), capable of capturing moderately high resolution images of an entire human brain in 3–5 s.

These two studies led directly to corresponding work with human subjects5,6, in which visual stimulation took the place of respiratory challenges in modifying vascular oxygenation. It had been known from PET and other studies7,8 that focal increases in neuronal activity give rise to increases in blood flow that surpass changes in oxygen consumption, so that the net oxygenation of blood leaving a neuronally activated area is increased. Thus, this venous blood becomes more similar in its magnetic properties to the surrounding tissue, and causes a local rise in MR image intensity. At typical MRI static field strengths of 1.5 T, this increase is about 4% of the total image intensity, and is thus quite readily observable9. Because the arteriolar control of blood flow is spatially well matched to the areas of increased neuronal activity, and changes in oxygen demand also correlate well to these electrical changes, this technique has the potential to provide quite accurate localization of neuronal activity. The change in signal has a vascular origin, and for this reason its time-course (termed the haemodynamic response function) is slow compared with that of the underlying neuronal activity. Typically, the changes take 4–8 s to build up and to decay, as compared with neuronal time constants of a few tens of milliseconds (Fig. 1).



View larger version (12K):
[in this window]
[in a new window]
  		Fig. 1 Time course of signal changes from one voxel in visual cortex of data set of Plate I (see end of file p.*21), after de-trending and correction for effects of head motion. Also shown is model of task effects used for evaluation of correlation (box-car convolved with haemodynamic response function).

 


View larger version (31K):
[in this window]
[in a new window]
  		Plate I Statistical parametric map of BOLD contrast brain activations produced by visual stimulation of a volunteer subject using a reversing checkerboard. Brain coverage 48 slices, voxel resolution 3 mm x 3 mm x 3 mm, echo time 40 ms. MRI sequence was T2*-sensitive gradient-echo EPI. Activation thresholded at P < 0.0001, uncorrected for multiple comparisons. Data obtained with a 2.0 T Siemens Vision MRI scanner at the Functional Imaging Laboratory, Queen Square, London. Thanks to Dr Chloe Hutton for providing the analysed data.

 
The change in signal increases with field strength9, and so also does the intrinsic image signal-to-noise ratio (SNR). Thus, many fMRI researchers are turning to scanners of 3 T or greater, although certain image artefacts also increase with field strength, as will be discussed later.

Equipment

Excellent MRI hardware is required to obtain functional MRI data of acceptable quality. Until the early 1990s, commercial MRI scanners were not designed to perform the very fast imaging method, EPI, that makes fMRI relatively straightforward.

By 1995, all major MRI manufacturers offered scanners with adequate specifications. There is now increased interest in 3 T scanners, which have higher fMRI sensitivity, and research sites have whole-body systems at field strengths up to 8 T, requiring drastic measures for dealing with image quality problems. Improvements in specifications can result in improved image quality, improved sensitivity, and/or improved data capture rate. Simply stated, the specifications required to obtain useful fMRI data are summarized in Table 1.


View this table:
[in this window]
[in a new window]
  		Table 1 Specifications required to obtain useful fMRI data

 
System stability
 Image intensity should be stable to within 0.5% short-term, 1% long-term. Because very small fractional signal changes are observed in BOLD contrast, there is a requirement for excellent scanner stability. While global changes of image intensity, which might be caused by fluctuations of radiofrequency transmitter power, for instance, can be removed by image intensity normalization, this decreases by one the degrees of freedom in the data for subsequent analysis. More seriously, hardware-related spatial variations in image intensity can dramatically increase the variance of a data set, and thus reduce the detectability of brain activation. What is now achievable with good quality MR hardware is an image intensity stability of 0.1% over 15 min, for echo-planar images repeated at the rate of 10 slices/s.

An illustration of the brain coverage that can be obtained with modern equipment is shown in Plate I (see end of file p.*21), where activation maps produced by visual stimulation are shown. Each brain volume took 4 s to acquire, and the total run time was 5 min. A Siemens Vision MRI scanner was used for this work, operating at a field strength of 2 T.



View larger version (39K):
[in this window]
[in a new window]
  		Plate II Images of regional cerebral glucose utilisation using [18F]-fluorodeoxyglucose in a normal subject, co-registered with the MRI, using the currently most advanced positron emission tomograph for the human brain-ECAT HRRT39. Supplied by kind permission of Prof. Klaus Weinhard, Max-Planck-Institute for Neurological Research, Cologne, Germany.

 
Data acquisition

For fMRI studies of human brain function, subjects normally lie on a bed within the bore of the MRI scanner magnet. During each experimental run of 5–30 min, brain images are acquired continually while the subject performs tasks related to the aspects of brain function under consideration. Each brain volume takes 2–5 s to acquire, and a spatial resolution of 2–3 mm is feasible with most modern MRI scanners.

The most popular type of acquisition is echo-planar imaging11, as mentioned earlier, which has several variants implemented by different scanner manufacturers. These include spiral EPI12 and PRESTO13. While advantages have been claimed for these variants, they also have their problems, and the conceptual simplicity, ready availability, and ease of interpretation of EPI data makes it widely used. A typical data set consists of up to several hundred brain volumes, with a single voxel SNR of perhaps 50–100, and amounting to hundreds of kbytes of data.

Task presentation

Tasks relating to neuropsychological and psychophysical hypotheses are presented using a selection of devices from a battery of ‘cognitive interface’ equipment, that allows the experimenter free rein in designing experimental protocols. This equipment can include apparatus for high quality visual, auditory, tactile, and olfactory presentation, and devices for response monitoring, such as keypads, joysticks and keyboards. It is designed to be compatible with the very powerful magnetic fields used in MRI, and is appropriately shielded to prevent radiofrequency interference with the tiny signals from the nuclear spins that allow reconstruction of MR images. Such equipment is often home-built, although there are now a small number of manufacturers that have developed such specialized devices. It is also possible to monitor EEG signals continuously during scanning14, which is useful for studies of such areas as sleep and epilepsy.

Types of experiment
 fMRI provides quite good temporal resolution, with rise and fall times for functionally-related signal of a few seconds, as already mentioned. Commonly, tasks are presented in ‘block design’ mode, in which experimental conditions have durations of 20–30 s, or in ‘event-related’ mode15, where single presentations of the various tasks of interest are separated by a few seconds. Often, event-related paradigms utilise the strategy of randomizing the inter-trial interval, which allows designs of increased statistical power16. Experimental designs involving comparisons of conditions each lasting longer than a minute are not recommended, because slow physiological and instrumental signal drifts create significant low frequency noise. However, designs where alternated tasks are repeated in different physiological states are feasible, and a number of studies of brain processes involved in learning have been successfully performed (e.g. Karni et al17).

Areas that have been studied with fMRI range from fundamental psychophysics of various sensory modalities (e.g. Sereno et al18) to investigation of cerebral representation of emotion (e.g. Dolan et al19). The effects of psychoactive drugs on brain activity have also been studied (e.g. Coull et al20) although special care must be taken to avoid the confounding effects of time of scan.

Sources of error
 The signal changes used in fMRI are small, less than 10% of the static signal, at conventional MRI field strengths of up to 3 T. Most of the signal comes from water protons in brain tissue that are not affected by changes in brain activity. Thus the largest potential source of error in fMRI is head motion21. If uncorrected, a movement of as little as 0.5 mm can result in a 40% change of signal near a contrast boundary, swamping functionally-related intensity changes. However, this motion sensitivity also allows very precise characterizations of head movement, assuming rigid body motion during the inter-scan interval, and using the image data themselves. A number of efficiently-implemented algorithms22,23 have become publicly available, which re-align time series of volume images of the brain and remove almost all the error arising from this source. Additional sources of error are thermal noise, which becomes progressively more important for smaller voxel sizes and lower magnetic field strength, and low frequency noise arising from naturally-occurring slow fluctuations of blood flow (vasomotion)24 and pulsatile cerebral motions deriving from heartbeat and respiration. These can be largely removed by appropriate experimental design (see above) and high-pass filtering of the image data time series.

Another type of error is the mislocalization of functional signals, in regard to underlying neuronal activity. This can arise from two quite different origins. The first is image distortion and drop-out. Creation of images sensitive to magnetic field changes produced by variations in blood oxygenation entails their sensitivity to endogenous static variations in magnetic field associated with the geometry of the head and the differing magnetic susceptibilities of the head and the surrounding air. If the head and brain were spherical, a uniform external magnetic field would engender a uniform field within the brain. However, the head cavities close to the brain, such as the nasal sinuses and the ear canal, cause perturbations to this desired uniformity which create distortions and even complete loss of signal in EPI brain images. Thus, a functional signal may be displaced as much as 6 mm from its correct location in an undistorted structural brain image25,26, or may even be unobservable because there is no signal from the brain at that point27,28. These problems, which are exacerbated with increasing field strength, fortunately affect a minority of brain image voxels. Drop-out can be overcome by means of acquiring additional images with different imaging sequence parameters, that compensate locally for the effects of head cavities29, and image distortion can be corrected by the rapid acquisition of a magnetic field map of the head, which allows rewarping of the image data to its appropriate positions25,26,30.

The second cause of mislocalization arises from the cortical vascular physiology itself. During neuronal activation, regional CBF increases and thus increases the oxygenation of venous blood leaving the activated area. This increase of oxygenation causes an increase in MRI signal wherever it occurs, potentially in large draining veins distant from the site of increased electrical activity. While there has been some empirical research on this issue31, it was not until recently that the implications of typical venous geometry have been explored, and limits to the degree of spatial mislocalization have been estimated32.

In practice, imaging neuroscience researchers have largely ignored both of these problems, which can be unimportant, particularly in multisubject studies where successful averaging over brains normalized to a common template involves spatial smoothing by as much as 10 mm, a distance large compared with the spatial errors caused by distortion and draining veins. However, as single-subject and clinical studies gain in popularity, greater attention will need to be paid to these concerns.

Data analysis
 It is beyond the scope of this article to review in detail methods for data analysis. This is often performed using a software package known as SPM (Statistical Parametric Mapping, <http://www.fil.ion.ucl.ac.uk/spm/>), a sophisticated implementation of general linear model analysis techniques, which removes known confounds from the data and uses F- and t-tests to estimate the significance of apparent brain activity. Similar packages have been developed in a number of laboratories, and such techniques as cluster analysis33 and independent component analysis34 have been explored for the purpose of data-led, rather than hypothesis-led, analysis.

Limitations
 The main limitations of fMRI arise from the vascular origin of the signal changes that are correlated with neuronal activity. This imposes physiological constraints on temporal and spatial resolution. The haemodynamic response takes place over several seconds, and it varies somewhat across brain tissue. Thus, a temporal resolution of at best 250 ms can be obtained, and that only in specially designed experiments with a great deal of signal averaging35. Normally, a resolution of no better than a few seconds can be expected, much greater than the temporal response of neurons, in the millisecond range.

The spatial resolution is limited by three factors: (i) the imprecision of the vascular control of blood flow; (ii) the resolution of the MR images (in turn limited by available signal-to-noise ratio); and (iii) the contaminating effect of blood leaving activated neuronal tissue with changed oxygenation, which progresses down the venules and draining pial veins to give apparent activation outside of the neurally activated area31. For well-designed experiments, in which the differential areas of brain activation are expected to be small, this does not constitute a major problem. A typical value for the spatial resolution of fMRI is about 3 mm. This can be improved using special techniques, to allow visualization of ocular dominance columns, spaced by less than 1 mm36.

A final limitation, previously mentioned, is due to the high sensitivity of the images to slight variations in the magnetic field used for scanning, required to observe BOLD contrast. Because the magnetic susceptibility of tissue is slightly different from that of air, re-entrant air-filled cavities in the head close to the brain, such as the nasal sinuses and the ear canal, can cause distortion and loss of signal in some brain regions, notably orbitofrontal cortex and inferior temporal cortex. While remedies exist for these problems25,29, they can result in slower and more complex scanning techniques.

Potential
 The scope of fMRI is great, as is evident from the range of studies using this technique in this volume of the British Medical Bulletin. Because it is uniquely non-invasive, has good sensitivity, and gives relatively high spatial and temporal resolution, it has replaced [15O]-PET in many research areas where localization of function is of primary interest. Once some remaining problems, such as dealing with large head movements in uncompliant subjects, have been resolved, there is great potential for longitudinal studies of brain functional development and deficit in childhood, and learning of knowledge and skills at any age. There has been much interest in fMRI for psychiatric studies, although the need for precisely characterized disease conditions, adequate numbers of subjects and controls, and appropriate analysis methods has not always been appreciated. As understanding and methods improve, this is likely to be an area of research that is very productive in new insights regarding localization of brain areas and connections associated with behavioural and cognitive abnormalities. Studies of large numbers of subjects of research interest are feasible, because of the short duration (< 30 min), non-invasiveness, and good subject acceptability of fMRI.

Positron emission tomography (PET)

Basic mechanism
 PET
This imaging modality rests on being able to measure the distribution of positron emitting radioisotopes in the body. Although a positron can travel no more than a few millimetres in tissue, it is possible to locate its presence within the body. This is because when the positron stops, it is captured by an electron, and two photons are emitted. These annihilation photons leave the body at approximately 180° to each other and, by using radiation detectors that operate in coincidence, lines through the body are defined along which the positron capture occurred. If sufficient lines of events and angles are sampled, it is possible to reconstruct the three dimensional, tomographic distribution of where the positrons are being captured in the body and the distribution of radioactivity. Hence the term positron emission tomography: PET37.

Tracer principle
The reader will be aware of the tracer principle, where a marker is mixed with a material, the fate or distribution of which needs to be determined within a physical or biological matrix. The key factor is that the marker has to follow faithfully the same fate as the substance being traced. Hence, it needs to be in the same physical or molecular form. One answer to this has been to use radioactive forms of the parent material to act as tracers, the presence of which is clearly identifiable and measurable with high sensitivity. This takes the form of detecting the specific radiation emitted by the radioisotope and its rate of radioactive decay, against a low natural level of background radiation.

Tracing biological systems in vivo
In order to trace physiological/biochemical organic molecules, radioactive tritium (3H1) as a marker substitute for hydrogen (1H1) and carbon-14 (14C6) as a marker of carbon (12C6) have been used extensively. However, these radioisotopes emit low energy ß-rays, which mean that biological assays rest on destructive sampling of tissues. To avoid this, there is a need for radioactive forms of hydrogen or carbon which emit {gamma}-rays. Since these rays penetrate through tissue, they can be recorded with radiation detectors placed external to the body. By suitably collimating and calibrating the detectors, the regional tissue concentration and time course of tracer can be measured. From such measurements, the fraction of administered tracer distributed to the tissue of interest and its rate of flux or sequestration into that tissue can be quantified. Using appropriate kinetic models that define the biological fate of the tracer, it is then possible to analyse the recorded data to derive absolute rate constants of exchange or turnover of the substrate. It is these measurements which can be used to investigate the in vivo physiology of normal tissue and the pathophysiology of focal disease, as well as providing objective, functional assessment of treatment.

Positron emitting radio-isotopes
When considering what {gamma}-ray emitting radio-isotopes could be used to radiolabel organic molecules in a way similar to that of tritium and carbon-14, we have the positron emitters of carbon-11 and fluorine-18. The later can substitute in some cases for hydrogen and still retain the molecule's biological activity. Oxygen-15 is also a positron emitter. It has a half-life of only 2.1 min and, therefore, it is difficult to build it into a complex molecule and still have sufficient activity to study. However, it can be used to label water or molecular oxygen for human studies.

Molecular imaging
Hence PET, which not only stands for both positron emission tomography but also positron emitting tracers, provides a means to measure in the body, three-dimensional distributions of tracers radiolabelled with positron emitting radio-isotopes. Carbon-11 and fluorine-18 have relatively short radioactive half-lives of 20.1 and 110 min, respectively. However, rapid radiochemical techniques have been developed to incorporate them into biochemicals and pharmaceuticals to provide sufficient material for imaging their uptake in the tissues of the body following intravenous administration38. This results in a form of molecular imaging in that, by being able to label specific molecules and image their distribution and time course within the tissues of the body, it is possible to study molecular interactions and pathways. These include neurotransmitter receptor binding sites and metabolic substrate transport

Equipment
 Cyclotrons
In order to produce positron-emitting radio-isotopes, in most cases stable elements need to be bombarded with a high-energy charged particle beam. This is provided for with a particle accelerator that is usually a cyclotron. When using such radioisotopes as oxygen-15 and carbon-11, since their half-lives are so short, the cyclotron needs to be located near to where they are to be used. At one time, such equipment, because it was specialised and needed to be heavily shielded, was quite a rare installation within a hospital environment. However, today, there are numerous hospital or distribution centre-based cyclotrons.

Radiochemistry
Shielded lead enclosures known as hot cells, are used to house the chemistry apparatus needed to build the high levels of radioactive carbon or fluorine into the biochemical or pharmaceutical of interest. This process has to be fully automated to avoid exposure of staff to radiation. It is accompanied by a procedure for rapidly determining chemical and radiochemical purity of the labelled compound prior to its administration into humans. In the case of 20.1 min half-life carbon-11, both procedures have to take place within an hour following the production of the radio-isotope. Equipment to carry out a number of the more standard radiochemical procedures are now available commercially. Gradually, for standard preparations, this is becoming slightly less specialised. In the case of one of the most simplest of tracers, H215O, it is prepared in the PET scanning room adjacent to the camera by the catalytic burning of 15O2, piped from the cyclotron and hydrogen.

PET camera
Because data are collected by electronic (coincidence) counting, the use of physical collimation is circumvented. As a consequence, the arrays of detectors viewing the brain are able to effect a relatively large solid angle for accepting counting events. This provides the high sensitivity that is necessary to be able to reconstruct the spatial resolution, accurately follow the time course of a tissue's concentration of a tracer over time, and detect low concentrations (sub-picomolar) of molecular binding. The latest camera for the brain is able to survey the whole head and has a spatial resolution of < 2.5 mm across the human brain (Plate II, see end of file p.*22)39.

Peripheral monitoring
For studies in which the time course of tracer in the arterial blood also needs to be measured, in order to solve the kinetic uptake of tracer into a tissue of interest, a radial artery is usually cannulated and blood continuously withdrawn over an external detector. In addition, when using a physiological substrate or drug that becomes metabolised within the body, it is necessary to analyse serial blood samples into their constituent parent compound and metabolite components. This requires dedicated chromatographic analysis and high sensitivity detectors.

Data acquisition
 Quantitative data collection
With the subject positioned within the PET camera and prior to administering the tracer, a transmission scan is carried out. This involves rotating a point or rod source of activity, usually the long-lived positron emitting source germanium 68/gadolinium 68. When these scan data are ratioed with a blank scan, recorded when there is no subject in the camera, factors are produced for the attenuation of signal that occurs in the subject's head, along each coincidence line of response. The net result of this is that the PET image data can be accurately corrected for attenuation. Hence, the camera's voxel-element response can be calibrated against laboratory detectors used for counting blood samples and the dose of activity administered. This means that concentrations of tracer measured in brain tissue are recorded in the same units as those of blood or plasma samples. Data expressed in common units enable kinetic models, using input functions derived from blood samples, to be operated to derive measures of tracer flux or binding. It also allows a measure of the percentage of the administered dose that is present in a unit volume of brain tissue.

Recording kinetic PET data
When using tracers or ligands labelled with short-lived radioisotopes, co-ordination of timing is of the essence. In particular, the subject has to be prepared and lying in the PET camera in good time, before the rapidly decaying probe arrives from the radiochemists. Blood sample timing has to align with that of the scan times. Following the injection of tracer, the data from the PET camera are collected either within pre-defined time frames or on an event-by-event bases against time (i.e. list mode data). In the later case, the data can be subdivided post-acquisition into any combination of time frames. With the more advanced cameras, a 90-min time-course study can accumulate up to 10 Gbytes of data. In many cases, arterial blood is continuously withdrawn and monitored, and a number of discrete samples taken to determine the partitioning between plasma to red cell and between parent compound and labelled metabolite. From the total blood data collected, the time course of the plasma concentration of parent tracer or ligand can be derived.

Less invasive PET procedures
What has been described is the data acquisition needed to implement a full quantitative approach. However, in the interest of being less invasive and complicated, procedures have been adopted which avoid blood sampling. This is possible by using an internal reference tissue that only contains non-specific signal such as the cerebellum and which substitutes for an input function40. In most cases, the simplification of the more invasive and complicated protocols is only possible after having recorded the full data set and demonstrating that analysis using the reference tissue produces the same accuracy for detecting focal change in function as that using the blood input data. A further example of how a more qualitative data collection is sufficient is in the case of studying regional cerebral activation using H215O as a marker of cerebral blood flow. Following earlier protocols which effected the full quantitative approach, it was realised that if the interest was mainly in identifying focal changes, then an analysis of the raw integral, flow-dependent, cerebral uptake of the tracer would suffice.

Task presentation
 The task presentations itemised for fMRI studies in the previous section are very similar to those when PET is used for regional cerebral activation studies whether they involve measures of focal changes in blood flow or ligand displacement. If anything, the logistics of task presentation are simpler with the PET scanner than within the confines of the bore of a MRI scanner where extra precautions need to be taken because of the high magnetic field. For non-activation studies, the subject just has to lie in the PET scanner and not get too distracted or bored over a typical 2-h period.

Types of experiment
 Cerebral activation studies apart, a major area of research is that of making observations of the pathophysiology of focal brain disease either in neurological or psychiatric patients and how this changes over time or in response to treatment. Hence, not only does molecular imaging with PET provide a clinical scientific tool, but also a means to help in the development of new drugs. Here, considerable emphasis is placed on deriving quantitative values for across-group and time comparisons. This is based either on measurements of receptor binding or substrate flux. A good example of the latter is the following up, over a number of years, of patients who have undergone striatal transplants for Parkinson's disease.

Sources of error
 The tracer and ligand
Bearing in mind that one is effecting the tracer principle, it is important, when using high affinity ligands to image pharmacologically active binding sites, that the level of cold ligand is kept as low as possible. The goal is to use high specific activity (ratio of radioactive to cold compound) preparations. Otherwise, the cold ligand will block the binding site and perturbate the apparent receptor density. Hence, great care is needed to assay each preparation for the specific activity prior to administration to the subject. If the radiochemical yield is for any reason low, care is needed not to compromise the statistical quality of the data by administering lower than normal levels of radioactivity.

The recorded PET data
Three sources of error can occur in the recording of the primary PET data. One concerns the registration of scattered coincidences. These occur when one or both of the emitted annihilation photons are scattered within the head before they are incident on a detector. As a result, the line through the point of the positron's annihilation will be wrongly defined. In a typical PET brain study, some 35% of the registered coincident events will have been scattered. There are a number of proven successful ways for accurately correcting for the scatter background within the PET data set. Care is needed when implementing to avoid noise amplification and to achieve data that can be quantified. That also is the case when correcting for the registration of random coincidences. Finally, care is needed not to exceed the response of the PET camera when administering high levels of tracer. The error here is the misunderstanding that if more radioactivity is given that there is a proportionate gain in signal. That does not follow since there are dead-time effects which mean that at higher signal rates, less signal is accumulated per unit of radiation dose received by the subject to whom the radioactivity is administered. This becomes of particular concern when studying young patients and especially their age-matched controls.

The biological kinetic model
As has been mentioned earlier, the recorded PET data are analysed within pre-defined compartment models. These essentially define the biological spaces occupied by the molecular probe and its rates of exchange between these spaces. Most of the models are derived from animal studies, usually of normal tissue. Hence, in the first instance, the model may not hold up in humans because of species' differences. However, of greater concern is how valid is the model for defining the fate of the tracer or ligand in diseased tissue? A compartment model developed in normal, well-regulated brain tissue would have little resemblance to say tissue recovering from a stroke, with gradients of pathophysiology ranging from necroses to ischaemia and hypoxia. Another gross difference would be in transplanted tissue and, of course, neoplasia. The brain tissue of patients, such a schizophrenics who have received chronic neuroleptic medication over many years, could well handle the tracer or ligand in quite a different way from that of normal tissue. Hence, care is needed in interpreting the results of processing PET data through models derived from normal tissue and that changes seen in disease could be due to the inappropriate model. When converting the recorded PET regional brain data into absolute concentrations of tracer, it needs to be appreciated that, if the anatomical structures are smaller than the physical resolution of the PET camera, the data will be diluted. This partial volume effect41 can be corrected for using the structural MRI scan image.

Non-invasive kinetic studies
While it is tempting to introduce non-invasive methods and use a reference tissue to normalise data or derive input functions, assumptions usually have to be made about the levels of specific binding and rates of exchange within that reference. This does represent a potential source of error that is often difficult to examine in human tissue.

Patient-related error
There are two main sources of patient-related error over and above concerns as to the effects of the disease states. One is head movement during the study and the other is the dietary state of the patient. This is especially important when studying the transport of metabolic precursors (e.g. glucose and L-dopa. Both of these are transported by systems that can be competed for by naturally circulating substrates.

Data analysis
 The analytical packages mentioned in the fMRI section used to identify the presence and statistical significance of focal changes in cerebral function have also been used for analysing PET brain data. The goal is to process a kinetic time sequence of data to produce quantitative, parametric images of entities of the tissue's function. These include rate constants for substrate exchange and sequestration, receptor density, etc. The challenge is to derive these images by operating the analysis on a voxel-by-voxel bases, thereby maintaining the spatial resolution and yet still minimise noise amplification. As a result, attention is being centred on signal averaging techniques and those that are more data, rather than model, led. These include principle components, cluster, wavelet and spectral analysis42. The later makes no assumptions on a radioactive probe's compartmentalisation in the tissue, but identifies what are the significant components that are present in the time course data. Carson has recently written an extensive review on analytical techniques used in PET43.

Limitations
 Lack of chemical resolution
Although PET is the most sensitive means for molecular imaging of the human brain44, it has the disadvantage that the recorded signal has no chemical resolution. This follows, since one is detecting the single energy pairs of photons emitted from a positron's annihilation with an electron. Hence, to use the specificity of PET, the signal has to be interpretable as being the administered tracer or ligand with or with out some well-defined, background levels which are low enough to be corrected for. These include radiolabelled metabolites and non-specific binding.

Lack of PET tracers and ligands
A second limitation stems from the fact that although there are numerous molecular probes that work well for in vitro assays, there are no more than a dozen or so which work in vivo. This is due to either rapid in vivo metabolism, poor penetration across the blood–brain barrier or high levels of non-specific binding. Hence, at this time, there is a need for fundamental research to understand further what makes for a good probe with respect to circumventing the above restrictions. Further insight into the discovery of novel probes should lead on to realising the full potential of molecular imaging with PET. One strategy to consider for wide-spread clinical application is to focus on the development of more generic tracers such as [18F]-fluorodeoxyglucose for cerebral glucose metabolism and H215O for blood flow that have found broad use. A foretaste of such a generic probe is that reported in the accompanying chapter by Banati who describes the use of a PET marker of microglia activation in the brain, a process that accompanies neurological damage.

Simplification of PET procedures
The use of short-lived radioisotopes such as carbon-11 is quite complex and logistically difficult which limits broad application. This is further complicated by the need to handle large sets of kinetic data and computationally intensive data analysis. One approach to avoid these hurdles is to develop probes labelled with the 110 min half-life fluorine-18. This would allow it to be synthesised at a central distribution point. Secondly, to simplify the recording and analysis of data, more trapped probes, usually in the form of enzyme inhibitors, need to be developed where the kinetic process is frozen to effect static imaging. A good example of such a probe is fluorodeoxyglucose which is transported into the brain in a way similar to natural glucose but is phosphorylated and hence trapped by the enzyme hexokinase45. Labelled with fluorine-18, this tracer is used to image regional cerebral glucose utilisation and also is used extensively in oncology for imaging the presence in the body of metabolically active cancer deposits. These uses are interesting given that this molecule was originally developed as an anti-cancer agent46.

Potential
 PET has not realised its full potential for molecular imaging of the brain, either as a research or clinical diagnostic tool. It is upon the research success of PET that the more advanced use of PET in more routine clinical medicine will evolve. Methodological developments are required throughout the whole of the imaging science of PET47. However, in particular, it is through the development of more specific probes for receptor and pathway imaging that there is potential to gain further insight into neurochemical perturbations in such diseases as dementia, schizophrenia or depression as well as broadening the support for drug development in general. As the post-genome era matures, such studies would ideally parallel the characterisation of the genetic profiling of patients by measuring the functional phenotype.


   Footnotes
 
Correspondence to: Prof. Terry Jones, Academic Department of Radiation Oncology, Christie Hospital NHS Trust, Wilmslow Road, Withington, Manchester M20 4BX, UK


   References
Top
 Abstract
 Functional magnetic resonance...
 References
 

  1. Hashemi RH, Bradley WG. MRI: The Basics. London: Williams & Wilkins, 1996
  2. Ogawa S, Lee TM, Kay AR, Tank DW. Brain magnetic resonance imaging with contrast dependent on blood oxygenation. Proc Natl Acad Sci USA 1990; 87: 9868–72[Abstract/Free Full Text]
  3. Pauling L, Coryell CD. The magnetic properties and structure of hemoglobin, oxyhemoglobin and carbonmonoxyhemoglobin. Proc Natl Acad Sci USA 1936; 22: 210–6[Free Full Text]
  4. Turner R, Le Bihan D, Moonen CTW, Despres D, Frank J. Echo-planar time course MRI of cat brain deoxygenation changes. Magn Reson Med 1991; 22: 159[Web of Science][Medline]
  5. Kwong KK, Belliveau JW, Chesler DA et al. Dynamic magnetic resonance imaging of human brain activity during primary sensory stimulation. Proc Natl Acad Sci USA 1992; 89: 5675–9[Abstract/Free Full Text]
  6. Ogawa S, Tank DW, Menon R et al. Intrinsic signal changes accompanying sensory stimulation: functional brain mapping with magnetic resonance imaging. Proc Natl Acad Sci USA 1992; 89: 5951–5[Abstract/Free Full Text]
  7. Fox PT, Raichle ME. Focal physiological uncoupling of cerebral blood flow and oxidative metabolism during somatosensory stimulation in human subjects. Proc Natl Acad Sci USA 1986; 83: 1140–4[Abstract/Free Full Text]
  8. Frostig RD, Lieke EE, Ts'o DY, Grinvald A. Cortical functional architecture and local coupling between neuronal activity and the microcirculation revealed by in vivo high-resolution optical imaging of intrinsic signals. Proc Natl Acad Sci USA 1990; 87: 6082–6[Abstract/Free Full Text]
  9. Bandettini PA, Wong EC, Hinks RS, Tikofsky RS, Hyde JS. Time course EPI of human brain function during task activation. Magn Reson Med 1992; 25: 390–7[Web of Science][Medline]
  10. Turner R, Jezzard P, Wen H et al. Functional mapping of the human visual cortex at 4 Tesla and 1.5 Tesla using deoxygenation contrast EPI. Magn Reson Med 1993; 29: 277–9[Web of Science][Medline]
  11. Schmitt F, Stehling MK, Turner R. (eds) Echo-Planar Imaging. Berlin: Springer, 1998
  12. Meyer CH. Spiral echo-planar imaging. In: Schmitt F, Stehling MK, Turner R. (eds) Echo-Planar Imaging. Berlin: Springer, 1998
  13. Liu G, Sobering G, Duyn J, Moonen CT. A functional MRI technique combining principles of echo-shifting with a train of observations (PRESTO). Magn Reson Med 1993; 30: 764–8[Web of Science][Medline]
  14. Allen PJ, Josephs O, Turner R. A method for removing imaging artifact from continuous EEG recorded during functional MRI. Neuroimage 2000; 12: 230–9[CrossRef][Web of Science][Medline]
  15. Buckner RL, Bandettini PA, O'Craven KM et al. Detection of cortical activation during averaged single trials of a cognitive task using functional magnetic resonance imaging. Proc Natl Acad Sci USA 1996; 93: 14878–83[Abstract/Free Full Text]
  16. Josephs O, Henson RN. Event-related functional magnetic resonance imaging: modelling, inference and optimization. Philos Trans R Soc Lond B Biol Sci 1999; 354: 1215–28[Abstract/Free Full Text]
  17. Karni A, Meyer G, Jezzard P, Adams MM, Turner R, Ungerleider LG. Functional MRI evidence for adult motor plasticity during motor skill learning. Nature 1995; 377: 155–8[CrossRef][Medline]
  18. Sereno MI, Dale AM, Reppas JB et al. Borders of multiple visual areas in humans revealed by functional magnetic resonance imaging. Science 1995; 268: 889–93[Abstract/Free Full Text]
  19. Dolan RJ, Morris JS, de Gelder B. Crossmodal binding of fear in voice and face. Proc Natl Acad Sci USA 2001; 98: 10006–10[Abstract/Free Full Text]
  20. Coull JT, Nobre AC, Frith CD. The noradrenergic alpha2 agonist clonidine modulates behavioural and neuroanatomical correlates of human attentional orienting and alerting. Cereb Cortex 2001; 11: 73–84[Abstract/Free Full Text]
  21. Hajnal JV, Myers R, Oatridge A, Schwieso JE, Young IR, Bydder GM. Artifacts due to stimulus correlated motion in functional imaging of the brain. Magn Reson Med 1994; 31: 283–91[Web of Science][Medline]
  22. Woods RP, Cherry SR, Mazziotta JC. Rapid automated algorithm for aligning and reslicing PET images. J Comput Assist Tomogr 1992; 16: 620–33[Web of Science][Medline]
  23. Richard S. Frackowiak JCM, Friston KJ, Frith CD. (eds) Human Brain Function. London: Morgan Kaufmann, 1997
  24. Biswal B, Yetkin FZ, Haughton VM, Hyde JS. Functional connectivity in the motor cortex of resting human brain using echo-planar MRI. Magn Reson Med 1995; 34: 537–41[Web of Science][Medline]
  25. Jezzard P, Balaban RS. Correction for geometric distortion in echo planar images from B0 field variations. Magn Reson Med 1995; 34: 65–73[Web of Science][Medline]
  26. Hutton C, Bork A, Josephs O, Deichmann R, Ashburner J, Turner R. Image distortion correction in fMRI: a quantitative evaluation. Neuroimage 2002; 16: 217–40[CrossRef][Web of Science][Medline]
  27. Ojemann JG, Akbudak E, Snyder AZ, McKinstry RC, Raichle ME, Conturo TE. Anatomic localization and quantitative analysis of gradient refocused echo-planar fMRI susceptibility artifacts. Neuroimage 1997; 6: 156–67[CrossRef][Web of Science][Medline]
  28. Lipschutz B, Friston KJ, Ashburner J, Turner R, Price CJ. Assessing study-specific regional variations in fMRI signal. Neuroimage 2001; 13: 392–8[Web of Science][Medline]
  29. Deichmann R, Josephs O, Hutton C, Corfield DR, Turner R. Compensation of susceptibility-induced BOLD sensitivity losses in echo-planar fMRI imaging. Neuroimage 2002; 15: 120–35[CrossRef][Web of Science][Medline]
  30. Jezzard P, Clare S. Sources of distortion in functional MRI data. Hum Brain Map 1999; 8: 80–5[CrossRef][Web of Science][Medline]
  31. Lai S, Hopkins AL, Haacke EM et al. Identification of vascular structures as a major source of signal contrast in high resolution 2D and 3D functional activation imaging of the motor cortex at 1.5 T: preliminary results. Magn Reson Med 1993; 30: 387–92[Web of Science][Medline]
  32. Turner R. How much cortex can a vein drain? Downstream dilution of activation-related cerebral blood oxygenation changes. Neuroimage 2002; In press
  33. Scarth G, Somorjai R, Alexander M, Wowk B, Wennerberg ABA, McIntyre MC. Cluster analysis of functional brain images. Hum Brain Map 1995; Suppl 1: 158
  34. McKeown MJ, Makeig S, Brown GG et al. Analysis of fMRI data by blind separation into independent spatial components. Hum Brain Map 1998; 6: 160–88[CrossRef][Web of Science][Medline]
  35. Menon RS, Kim SG. Spatial and temporal limits in cognitive neuroimaging with fMRI. Trends Cogn Sci 1999; 3: 207–16[CrossRef][Web of Science][Medline]
  36. Menon RS, Ogawa S, Strupp JP, Ugurbil K. Ocular dominance in human V1 demonstrated by functional magnetic resonance imaging. J Neurophysiol 1997; 77: 2780–7[Abstract/Free Full Text]
  37. Ter-Pogossian MM, Phelps ME, Hoffman EJ, Mullani NA. A positron-emission transaxial tomograph for nuclear imaging . Radiology 1975; 114: 89–98[Abstract]
  38. Stocklin G, Pike VW. (eds) Radiopharmaceuticals for Positron Emission Tomography. Amsterdam: Kluwer Academic, 1993
  39. Wienhard K, Schmand M, Casey ME et al. The ECAT HRRT: performance and first clinical application of the new research tomograph IEEE. Trans Nucl Sci 2002; 45: I104–10[CrossRef]
  40. Gunn RN, Lammertsma AA, Hume SP, Cunningham VJ. Parametric imaging of ligand-receptor binding in PET using a simplified reference region model. Neuroimage 1997 6: 279–87
  41. Hoffman EJ, Huang SC, Phelps ME. Quantitation in positron emission tomography: 1. Effect of object size. J Comput Assist Tomogr 1979; 3: 299–308[Web of Science][Medline]
  42. Cunningham VJ, Jones T. Spectral analysis of dynamic PET studies. J Cereb Blood Flow Metab 1993; 13: 15–23[Web of Science][Medline]
  43. Carson RE. Tracer kinetic modelling in positron emission tomography. In: Valk, Bailey, Townsend, Maisey (eds) Positron Emission Tomography: Principles and Practice. London: Springer, 2002
  44. Jones T. The role of positron emission tomography within the spectrum of medical imaging. Eur J Nucl Med 1996; 23: 207–11[CrossRef][Web of Science][Medline]
  45. Sokoloff L, Reivich M, Kennedy C et al. The (14C)-deoxyglucose method for the measurement of local cerebral glucose utilisation: theory, procedure and normal values in the conscious and anesthetized albino rat. J Neurochem 1977; 28: 897–916[Web of Science][Medline]
  46. Bessell EM, Courtnay VD, Foster AB, Jones M, Westwood JH. Some in vivo and in vitro antitumour effects of deoxyfluoro-D-glucopyranoses . Eur J Cancer 1973; 9: 463–70
  47. Jones T. The imaging science of positron emission tomography. Eur J Nucl Med 1996; 23: 807–13[CrossRef][Web of Science][Medline]

Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?


This article has been cited by other articles:


Home page
 Am J Sports MedHome page
E. Kapreli, S. Athanasopoulos, J. Gliatis, M. Papathanasiou, R. Peeters, N. Strimpakos, P. Van Hecke, A. Gouliamos, and S. Sunaert
Anterior Cruciate Ligament Deficiency Causes Brain Plasticity: A Functional MRI Study
Am. J. Sports Med., December 1, 2009; 37(12): 2419 - 2426.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (12)
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Turner, R.
Right arrow Articles by Jones, T.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Turner, R.
Right arrow Articles by Jones, T.
Related Collections
Right arrow Neurology
Right arrow Neurophysiology
Right arrow Radiology
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?