7 May 2012
In Part I, we looked at how images are obtained from functional magnetic resonance imaging (fMRI), as well as a history of brain-scan technology and where we got the idea that blood flow relates to brain function.
In this installment, we are going to look at some of the accomplishments and some of the limitations of fMRI. Then, we will look at some of the assumptions behind fMRI and how they affect our interpretation of brain scans.
Perhaps the biggest advantage of fMRI is that doctors can obtain an image of the brain through a non-invasive technique.
fMRI does not involve the use of radiation or invasive surgery, and the patient does not have to ingest a contrast agent. Furthermore, fMRI has very good spatial resolution, so an image can show various types of tissues at a detailed level.
However, just obtaining an image of the brain is the stuff of magnetic resonance imaging (MRI). The point of fMRI is to study brain activity. While MRIs are often used in the clinical setting, fMRI is a newer technique and is currently used mainly in the research setting. However, as fMRI technology and research continues to improve, scientists are hoping to use it for clinical diagnosis.
One clinical use of fMRI is for pre-surgery brain mapping. In cases where a tumor needs to be surgically removed, the surgeon can use fMRI data to determine the best way to remove the tumor, while removing as little of the functioning tissue as possible. In this case, the patient may be asked to tap a finger, or say a name, or undertake some other activity that would indicate which regions of the brain are the most active when the patient performs these tasks.
The idea is to see which parts of the brain are active for this particular patient, and then attempt to remove as little of the active tissue as possible when removing the tumor. The key to this technique is that it is patient-specific, because, as we will see, not all brains behave the same way.
One of the limitations of fMRI is its qualitative nature. fMRI is qualitative, not quantitative, so its measurements are based on comparisons to a baseline or to a norm. It does not provide a specific indicator of a particular mental state. This can pose difficulties in comparing one individual to another, and in measuring individuals for whom a baseline measurement cannot be taken.
In the best-case scenario, a baseline is taken of the patient without any kind of stimuli (at rest), and then measurements are taken during the same session of the same patient in response to various stimuli. This gives the most accurate information for changes in blood flow. For example, in order to study neural responses to pain, a baseline must be taken in a pain-free state. This precludes studying patients with chronic pain because a baseline measurement cannot be taken such that a response to stimuli can be measured. However, this does mean that fMRI can be used to study acute pain.
The most problematic situations for qualitative measurements are when a patient is compared to “normal” subjects for a baseline measurement. This draws on too many assumptions for what “normal” brain function is. For example, a patient with schizophrenia is compared to subjects who do not have schizophrenia. However, this presumes that the control group has no other mental issues that may cause an “abnormal” reading. Furthermore, there is no real definition of a “normal” brain scan. The best that scientists can do is to take a statistical average.
However, even this can be problematic. Studies have shown that different people have different regions of the brain that respond to different stimuli. For example, most people have a region in the left hemisphere that is activated during language use, but not all people do. Some people have this in the right hemisphere, and some show neither region.
Furthermore, from an experimental standpoint, patients that have a physical handicap or some kind of impairment may use different parts of their brain to complete the task that is meant to comprise normal stimuli for measurement (e.g. tapping a finger, or touching one’s nose). Therefore, this patient is a little more difficult to compare to a “norm” if something other than the impairment is being measured.
From Part I, we learned that the fMRI signal is produced by measuring a change in blood flow. Hemoglobin will change the environment of the magnetized nuclei in the brain, producing a change in fMRI signal. This technique is inherently slow. An fMRI measurement is made approximately six seconds after the stimulus. This is a large temporal lag, especially considering that neurons usually operate on the millisecond scale. This lag may mean that scientists are seeing an incomplete picture of the neurological response. Furthermore, age and gender have been shown to affect blood flow response, and therefore can affect the fMRI readings.
In general, fMRI is good for spatial resolution, but poor for temporal resolution. To accommodate for this, scientists will often use a combination of fMRI coupled with EEG or some other technique that can make up for this deficit.
One of the criticisms of fMRI is its assumption that only certain regions of the brain are operating at one time. It does not address the communication network throughout the brain, which may be the most important part of brain activity. While studies have shown that damage to certain areas of the brain will result in a lack of function, it is a leap to assume that performing this function was localized to only that region of the brain.
Furthermore, one of the fundamental assumptions of fMRI is that blood flow corresponds to brain activity. This is reasonable enough, but the opposite is not necessarily true. FMRI looks at a change in blood flow, and just because there was not a change in blood flow to one particular region of the brain, does not necessarily mean that region of the brain is “silent.”
On a fundamental level, there is a tendency in neuroscience to make reductionistic assumptions. Reductionism assumes that function (or a material’s essence) can be understood by breaking it down to its component parts. For example, the brain can be broken down to hemispheres, then regions, then neurons, and the assumption is that this will somehow give us a greater understanding of how the brain works.
Scientists have observed regional activity in the brain, but there is no reason to believe that a physical activity or mental state can only operate out of that region. There have been cases of patients who have had portions (or even an entire hemisphere) of their brain removed but who have maintained or regained most of their functional capabilities. This calls into question assumptions regarding a direct correlation between function and particular regions of the brain.
Functional magnetic resonance imaging is one of many tools for studying the human brain. But just like any tool, it needs to be used in the right way, for the right job.
Brain scans have their time and place, but they also have their limitations.
For background reference, see Scott H. Faro and Feroze B. Mohamed, eds., BOLD fMRI: A Guide to Functional Imaging for Neuroscientists (Springer, 2010); and Scott H. Faro, Feroze B. Mohamed, and Victor Haughton, Functional MRI: Basic Principles and Clinical Applications (Springer, 2006).
For a fairly good, readable summary of the problem of reductionism in neuroscience, see this 2008 op-ed piece in the Los Angeles Times by Jonah Lehrer, entitled “Misreading the Mind”: http://articles.latimes.com/2008/jan/20/opinion/op-lehrer20.
For a good overview of fMRI, see the University of Oxford’s FMRIB Centre web site: http://www.fmrib.ox.ac.uk/education/fmri/introduction-to-fmri/introduction.