Stress has been recognized as an important psycho-physiological state since the pioneering work of Hans Selye. But until quite recently it has mainly been perceived in humans as a condition generated by extreme and hostile environments such as going into battle, hospital or academic examinations. Increasingly, however, it has been identified as being a consequence of many aspects of lifestyle and the events of everyday living and that, to varying degrees, large numbers of people experience it. Indeed, from the point of view of long-term health, low-level frequent chronic stress is likely to be much more important than occasional acute episodes.

Chronic stress can hardly be studied by experimental procedures in the laboratory. It clearly needs a population approach with investigators monitoring people in the “field” as they go about their daily business. Psychologists have gained important insights by the design of questionnaires which can be applied not only to particular groups undertaking activities which are deemed to be stressful, such as air traffic controllers, but also to whole populations, experiencing a diversity of lifestyles. They have identified various elements, particularly in occupational situations, which aggravate stress, as for example absence of job control, but questionnaires are of little use outside one's own language, or at least culture. They also have questionable validity in the study of children.

For these wider studies it is necessary, or at least desirable, to have some physiological measures of the stressed state, either of the homeostatic mechanisms which are elicited to control stress or of the morbid consequences of being stressed.

It is commonplace for people to conceptualize the body's response to stress as an “adrenaline rush,” with the symptoms of this response, as described by Cannon (1915), characterizing the “fight or flight” response. This common understanding of the stress response encompasses one major component of biological stress: activation of the sympathetic adrenal medullary system (SAMS). The other main arm of the stress response, the hypothalamic pituitary adrenal cortex (HPA) axis, will be discussed in the next chapter. Activation of the SAMS has widespread effects in the body. These effects are seen as being allostatic in nature; that is, they help maintain homeostasis through the initiation of physiological change. Measures of stress have been used that identify the indirect effects of SAMS activation such as increased pulse and blood pressure. Other measures, to be examined here, focus on more direct evidence for SAMS activation.

Stress has taken on great importance for humans in the modern world. The most common ailments that threaten us, such as cardiovascular disease, diabetes, and even some cancers, apparently are influenced by chronically high stress levels. The response that has allowed our ancestors to adapt to the stressors that were part of their world now appears too often to be an inappropriate one for the kinds of stressors, from traffic to deadlines, that confront us today (Sapolsky, 1994).

The stress response represents adaptation to actual or perceived challenges from the environment, with this process termed “allostasis” (McEwen, 1998).

There is no one best way to reduce fMRI data to brain activity maps, either activation maps or connectivity maps. The fMRI investigator must select a method of choice on the basis of a number of factors including software availability, speed of computation, ease of use and taste. The factor of taste can only have legitimacy if the investigator understands how each of the various approaches work and what the trade-offs are of using one method over another. The goal of this review was to give the investigator that overview of how the various available methods work. Once a method is selected, a deeper understanding of it may be obtained from the original literature. At that point investigators may be confident that, in focusing on a small number of methods for use in their laboratory or clinic, they have not overlooked a method that may have significant impact on the interpretations of their computed brain activity maps.

All of the methods reviewed here require that the MRI data be transferred “offline” from the MRI computer to a computer dedicated to brain map computation. The only method widely available “on-line” for the computation of activation maps is a simple independent samples t-test that compares average activation in an “on” state to the average activation in an “off” state. Currently, setting up an fMRI capacity in a clinical situation is not a turn-key proposition, although a few turnkey off-line systems are now being offered commercially.

fMRI time-series contain a number of systematic sources of variability that are not due to the BOLD effect of interest. These sources of systematic variability may be removed as a preprocessing step, as outlined in this chapter, or they may be removed as a covariate at the time of activation map computation. The sources of variability include factors due to the physics of MRI, subject motion, heart beat and breathing, other physiological processes, random thermally generated noise and intersubject anatomical variability. The variability due to the physics of MRI include ghosts, geometric distortion and some signal drift which (with the exception of drift) may be removed at the pulse sequence and image reconstruction level. However, such ideal pulse sequences are not yet widely available on commercial scanners and many investigators must simply accept compromises, such as the geometric distortion of EPI images. A small amount of subject motion can be compensated for by aligning the time-series fMRI images but motion may also be dealt with at the source by using suitable restraining devices and training the subject to hold still (mock MRIs are very useful for training study subjects in this respect). It is impossible to remove the source of physiological variables so these need to be corrected at the time of data processing. Two approaches are used to remove physiologic effects from the data: one is to model heart beat and breathing, the other is to measure and remove the effect of a global signal. Again, many investigators do not account for non-BOLD physiologic variation at the expense of reduced statistical power to detect BOLD activations.

Before an fMRI time-series can be acquired, a question needs to be posed and an experiment designed. There are two main types of experimental design in fMRI, the blocked design and the event-related design. Figure 3.1 schematically shows the difference between the two designs. In the blocked design, a fixed number of multiple trials are presented in immediate succession in each block. The time between each trial is known as the stimulus onset asynchrony (SOA). Between blocks of stimulus presentation are blocks of rest. One cycle of blocked task and rest is known as an epoch. In an event-related design the tasks are presented individually, instead of in blocks, with a spacing, T, that may be variable. The time interval T may be further broken down into the interstimulus interval (ISI) and stimulus duration (SD) so that T = ISI + SD.

Each experimental design has its advantages and disadvantages. For example, a blocked design will generally be more sensitive to detecting activations while an event-related design may be better able to characterize the BOLD response. Both types of design rely on signal averaging to remove noise when the activation maps are computed. An event-related design with constant ISI may be considered as a limiting example of a blocked design. With both types of design, the hemodynamic response function (HRF) may be considered, in a linear systems approach, as the convolution of the stimulus paradigm (a step function equal to 1 during the stimulus and 0 otherwise) and the impulse response function (IRF).

You will find here a comprehensive review of all fMRI data processing methods proposed in the literature to 2005. I have endeavored, however, to produce a review that is useful to more than those already in the know. With the introduction of each major method I have additionally given an overview of how the method works in wide and hopefully intuitive terms. The overviews taken together should give the newcomer a broad idea of all the choices that can be made for transforming raw fMRI data into a brain activity map.

The term activity map is used here to include the specific forms of maps labeled as activation maps and connectivity maps. Activation maps show regions of the brain that are active in response to a task given to the person in the MRI, while connectivity maps are intended to provide information on the neural connectivity between the active regions.

All methods are described in a precise manner from a mathematical perspective. So a certain amount of mathematical and/or statistical background is assumed of the reader. However, the math is presented at a high level. You will not find here details of how to implement any of the methods reviewed. For the details you will need to consult the original literature listed in the references.

In short, this book can help the newcomer to the field of fMRI (or a seasoned researcher wanting to know about methods used by others) to become oriented via a three-step process.

The production of a brain activity map from data acquired with a volunteer or patient and magnetic resonance imaging (MRI) requires a fairly wide range of interdisciplinary knowledge and techniques. Producing brain activity maps from functional MRI (fMRI) data requires knowledge and techniques from cognitive neuropsychology, physics, engineering and mathematics, particularly statistics. The process typically begins with a question in cognitive psychology that can be answered, at least in part, by combining the knowledge obtained from brain activity maps with previous knowledge of the function of specific regions of the brain. The previous knowledge of regional brain function generally has its origin in lesion studies where disease or injury has removed a brain region, and its function, from the brain's owner. Such lesion-based knowledge has firmly established the principle of functional segregation in the brain, where specific regions are responsible for specific functions. The use of fMRI to produce activation maps allows specific questions on functional segregation to be posed and investigated without risk to the person being studied. The brain is also known to be a very complex system in which several regions, working cooperatively, are required for some tasks. This cooperation among regions is known as functional integration and may be studied using fMRI techniques that lead to connectivity maps. Methods for producing activation and connectivity maps are reviewed here with the goal of providing a complete overview of all data processing currently available to produce the brain activity maps from raw fMRI data.