Images of brain anatomy
Anotomical images of the brain
You are all familiar with CT, MRI, and other devices that radiologists use to take pictures of the structure or anatomy of the brain. There are some relatively new techniques that are now being used to get pictures of brain activity in addition to brain anatomy.
Functional magnetic resonance imaging (fMRI)
Functional magnetic resonance imaging (fMRI) is a novel, non-invasive method for measuring activity in the human brain, and for investigating the relationship between brain activity and behavior. The technique is similar to conventional MRI, which is used to take pictures of the anatomy of the brain, but can generate images of brain activity in addition to brain anatomy.
The movie shows a demonstration of an fMRI experiment. A human subject was lying still in the MRI scanner while watching dynamic visual stimuli. The visual stimuli are shown in the top row. The one on the right alternates between a flickering checkerboard pattern and a blank (uniform gray). The visual stimulus on the top-left alternates between moving dots and stationary dots. The bottom row shows an image of the anatomy of the subject's brain. The back of the brain is at the bottom of the image. Superimposed on the anatomical picture are fMRI measurements of brain activity, in response to the two visual stimuls above. Flickering checkboards evoke activity at the very back of the brain in an area called the primary visual cortex (V1). Moving dots also evoke activity in V1. But in addition to that, the moving dots evoke activity on each side of the brain (sort of behind the ears) in an area of the brain called visual cortical area MT.
The location of area MT is indicated in green on this computer graphics rendering of the brain. As we will see later in the semester, the neurons in cortical area MT respond selectively to visual motion, and are believed to be responsible for the perception of motion.
fMRI works by measuring blood flow, taking advantage of the coupling between neuronal activity and hemodynamics (the local control of blood flow and oxygenation) in the brain to allow the non-invasive localization and measurement of brain activity. Briefly, here's how it works:
The ultimate success of fMRI as a measurement of brain function depends on the relationship between the fMRI signal and the underlying neuronal activity. The vascular source of the fMRI signal places important limits on the usefulness of the technique. Although we know that the fMRI signal is triggered by the metabolic demands of increased neuronal activity, the details of this process are only partially understood. Consequently, this issue has emerged as an important question in neuroscience. There are two parts to the story: 1) The physics is very well understood, but 2) The physiology is not so well understood. If you are interested in learning more, see Physics and Physiology of fMRI.
fMRI is, therefore, an indirect measure of the underlying neural activity. But there are numerous demonstrations that the fMRI measurement is tightly linked with underlying neural firing rates. The above figure shows one example. This graph plots the average firing rate, averaged across a large number of individual neurons in primary visual cortex (V1) of the monkey brain, in response to series of visual stimuli with different stimulus contrasts. The graph also shows that fMRI measurements of human brain activity increase in the same way with contrast. This is important because it shows that the fMRI measurements are tightly linked with the underlying firing rates of the neurons.
Cerebral control of blood flow
In fact, we've known about the control of blood flow in the brain for many years...
Angelo Mosso, a late 19th century physiologist fascinated with pulsations in the brain (easily observable in newborn infants) that keep pace with the heartbeat. He worked with patients who had suffered head injuries that left them with permanent defects in the skull over the frontal lobes of the brain. He noticed that pulsations increased in magnitude during periods of mental activity. Mosso's recordings taken of the forearm (dark curves labeled A in each panel) and the head (light curves labeled C) show stronger brain pulsations after the events marked by the arrows have stimulated brain activity. Top panel, "resting quietly". Second panel, "clock strikes noon and bells of a church are heard". Third panel, "asked if Ave Maria should have been said". Bottom panel, "what is 8x12?"
Walter K recordings
Another extraordinary example is patient Walter K, observed in the 1920's. Congenitally abnormal blood vessels serving his visual cortex yielded turbulent flow, creating a brief rushing sound with each heartbeat. The sound was audible to the patient and to his physicians. Dr. John Fulton (a famous neurosurgeon) recorded these sounds. The patient would open his eyes and start to read a newspaper. A few seconds later there was a noticeable increase in the intensity of the sound. Recordings taken of Walter K's skull show little activity while his eyes are closed (top), but much more activity while he is reading (bottom 3).
The relationship between brain function and blood flow was first characterized in 1890 by Roy and Sherrington (Cambridge University). Based on experiments with animals, they suggested that there exists an automatic mechanism that regulates the blood supply in response to brain activity.
Recently, blood flow response has been measured carefully in animal studies. One opens a hole in the skull, points a video camera at it, and collects images that reflect the relative amount of oxygenated ("red") versus deoxygenated ("blue") blood.
This method of optical imaging the brain has been used to visualize fine details of the organization of brain function. The example above shows ocular dominance columns in primary visual cortex, regions of the visual cortex that receive dominant inputs from either the left or right eye. We'll learn all about this in the lecture on LGN and V1 later in the semester.
There is actually an overcompensation in the blood flow response. When the brain is active, an oversupply of oxygenated blood is delivered. This has the effect of making the fMRI images slightly brighter in the vicinity of active neurons.
Positron emission tomography (PET)
PET = positron emission tomography. A patient lies in a PET scanner after being injected with radioactive-labeled substance (typically water or glucose). Because more blood goes to active brain regions, there is more radioactivity in those brain regions. The PET scanner detects and localizes these pockets of higher radioactivity. Specifically, the scanner detects the unique radioactive decay of positrons, the anitmatter equivalent of electrons. These tiny, positively charged, radioactive particles emerge from the radioactive water (or glucose) that was injected. After being emitted they are attracted to nearby negatively charged electrons. When positrons and electrons come together, they are annihilated, and energy is released in the form of two photons (particles of light, no charge) that leave the point of annihilation in exactly opposite directions. The PET scanner is set up to detect the coincidental arrival of pairs of photons. The location of the positron-electron annihilation is determined by which pair of detectors are simultaneously active.
Test condition: flickering visual stimulus. Control condition: blank screen. Subtract the two PET images. This gives a picture of where in the brain there was a greater response to the test condtion.
Often, PET studies average results across several/many subjects. This is a problem because individual human brains can be quite different from one another in shape (just because different human heads and skulls are different in size and shape). One would like to repeat the measurement in the same brain, and average to reduce noise. But, one can't do that because of too much exposure to radioactivity.
Electroencephalogram (EEG) and the event-related potential (ERP)
EEG (electroencephalograph) is also being used to study brain function. At each electrode, the electrical activity is recorded at fixed intervals following a stimulus presentation - say, every 4 milliseconds. The electrical values at each interval are taken from many trials and averaged together so that electrical activity not caused by the stimulus averages to zero and the resultant signal shows only the activity produced by the stimulus. The result is called an event-related potential (ERP). Usually 10 to 100 presentations of the stimulus suffice to produce a reliable measurement that reflects characteristics of both the individual brain and the particular stimulus.
Magnetoencephalogram (MEG)
MEG measures the tiny magnetic fields evoked by electrical currents in neurons. The magnetic fields are generated when electrical currents travel along the dendrites of a large enough number of nearby neurons simultaneously.
Comparing fMRI, ERP, MEG, and PET
Both fMRI and PET depend on regional control of blood flow in the brain. But, current fMRI methods go way beyond the conventional PET subtraction methodology. Critically, fMRI is non-invasive; no radioactive stuff need be injected in your blood stream. So one can measure fMRI in the same subject as many times as you like. This is important when doing science: there is repeatability of the measurements, averaging out the noise in the measurements, and one can do proper control experiments. But mainly, it is important for experimental design. One can redo a study in a slightly different way based on preliminary results. This is really the way science is done. Scientists don't think up "the right'' experiment and go into work the next day and do it. Rather, they have a concept for what experiment they want to do. They try it. It typically doesn't work so well. Perhaps the data are too noisy (or not repeatable) to be conclusive. Perhaps they realize that there is a critical confound. Then, they decide how to do better and they try it again.
Spatial and temporal resolution of fMRI and PET are limited by blood flow. FMRI response depends on the average activity of neurons in a little chunk of brain, perhaps a couple of mm on a side (recall that there are about 50,000 neurons per cubic mm), and averaged over time. Temporal resolution is better with fMRI than PET because you get a single PET image from an entire (e.g., 30 sec) scan whereas with fMRI you collect a time-series of images (every second or two). Spatial resolution with fMRI is typically better than PET, in part because of the basic differences in the technology but also because we can do experiments on individual subjects. Averaging data from PET experiments across subjects tends to blur the results because there are individual differences in the shapes of our brains.
The main advantage or EEG/ERP over fMRI is temporal resolution. With EEG, you get millisecond time resolution. One HUGE disadvantage of EEG is localization. One doesn't really know exactly where these electrical signals are coming from. The EEG signal recorded with each electrode on the scalp reflects the pooled (average) electrical responses of large numbers of neurons throughout the brain. The strength of the EEG signal at each electrode depends on how far the electrode is from the neural source so the array of electrodes on the scalp can be used to infer roughly where the neural source is located. This works well if there are only a few neural sources. But if there are a large number of brain regions active simultaneously then the locations of the different neural sources get confused with one another and there is no way to pull them apart.
Like EEG, MEG has the advantage of millisecond time resolution. But like EEG, it has limited spatial resolution and localization. Given the importance of spatial localization of activity in the brain, fMRI has become the method of choice.
Another feature of fMRI is that it is widely available. Above are some pictures taken in front of the Meyer building at NYU when our fMRI scanner was delivered.
The neuroimaging revolution
Neuroimaging, particularly functional magnetic resonance imaging, has revolutionized neuroscience over the past decade. It is allowing a new era of research, complementary to more invasive techniques for measuring neuronal activity in animal models, to explore the function and dysfunction of the human brain. The technique of fMRI was invented only 14 years ago. Because of this technical advance, we are now faced with an unprecedented opportunity to develop an understanding of complex, human behavior in terms the mechanistic operation of the brain. There are over 12,000 papers published each year in scientific journals that report results from fMRI experiments. fMRI is being used to study schizophrenia, depression, autism, dyslexia, attention deficit disorder, and a host of other human neurological, psychiatric, and developmental disorders. New interdisciplinary fields have been founded because of this technological advance including neuroeconomics (the study of how the brain makes economic decisions) and social neuroscience (the study of how the brain guides human social interactions). Even so, the full potential of this technique has yet to be realized. MRI technology, experimental design, and data analysis techniques are all evolving rapidly. It is clear that fMRI provides a new and different picture of brain function. This presents both challenges and opportunities.
Neuroethics
One of the challenges involves the possible misuse or premature application of fMRI technology. Along with the revolution in neuroimaging, a new field of neuroethics has evolved. Neuroethics is the study of the ethical, legal and social questions arising when scientific findings about the brain are carried into medical practice, legal interpretations, and health and social policy. There are a host of ethical concerns including privacy (reading someone's mind with fMRI) and culpability (should someone be held responsible for a crime if an fMRI measurement can show that there is something wrong with their brain), etc. The Dana Foundation is a good place to start for finding information about neuroethics.
A pressing example of neuroethics concerns lie detection. Two companies (Cephos and NoLieMRI) have announced new lie detection technologies based on fMRI. The ACLU held a press briefing on this topic entitled "Mining the Mind", which can be downloaded from the ACLU web site. The ACLU has also issued request under the Freedom of Information Act for information about the government use of brain scanners in interrogations (see ACLU press release). The issue has been covered by articles in the scientific journals Nature and Science, as well as USA Today, the SF Chronicle, and a number of other newspapers and magazines.
This paper by Davatzikos et al is one of a series of recent brain imaging studies on lie detection, the most impressive to date. The technique that they use relies on sophisticated statistical analysis of brain imaging measurements. They conclude 88% accuracy (90% sensitivity, 86% specificity). This corresponds to 14% false alarm rate and 10% miss rate. Put another way, about 1 person in 7 will be incorrectly identified as lying. This approach is very different from, and potentially much more powerful than, traditional polygraph. Polygraph measures physiological responses that indirectly reflect the activity of the autonomic nervous system: pulse, blood pressure, sweat (electrical conductance of the skin). The autonomic nervous system regulates essential bodily functions (heart, breathing, digestion). This happens automatically and subconsciously. Changes in the polygraph measures can reflect arousal during deception but also general anxiety. fMRI lie detection and EEG "brain fingerprinting" record neural activity in the brain including neural activity that is related to conscious mental state. This is very different from the polygraph which is an indirect physiological measures of automatic and unconscious regulation of bodily function.
The anterior cingulate cortex has often been reported to be active when people lie. But it has not been established that there is a causal relationship between activity in this part of the brain and lie detection (see above).
Many of the concerns about reliability might be mitigated by improving our understanding of how lying works in the brain. Even the issue of individual differences and different strategies for lying might become well understood. But that leaves us with another dilemma: should state-of-mind be considered private? We should be concerned about misapplication and preadoption of a technique that is not reliable. We should also be concerned about invading the privacy of someone’s mind. Does a person have a right to keep his or her subjective thoughts private, the right to cognitive freedom?