Denis Le Bihan, MD, PhD; Peter Jezzard, PhD; James Haxby, PhD; Norihiro Sadato, MD, PhD; Linda Rueckert, PhD; Venkata Mattay, MD
Le Bihan D, Jezzard P, Haxby J, Sadato N, Rueckert L, Mattay V. Functional Magnetic Resonance Imaging of the Brain. Ann Intern Med. 1995;122:296-303. doi: 10.7326/0003-4819-122-4-199502150-00010
Download citation file:
Published: Ann Intern Med. 1995;122(4):296-303.
This conference reviewed the potential scope of application for recently developed techniques for functional magnetic resonance imaging (fMRI) of the brain.The most successful technique is based on the sensitivity of magnetic resonance imaging (MRI) to magnetic effects caused by the modulation of the oxygenation state of hemoglobin, which is induced by local variations in blood flow during task activation. Typically, the MRI signal increases by a few percentage points during brain activation because blood flow and oxygen supply sharply increase. Brain activation images with excellent combined spatial and temporal resolution have been obtained noninvasively using visual, sensorimotor, or auditory stimuli, or during higher-order cognitive processes such as language or mental imagery. Although sensitive to misregistration artifacts and macroscopic vessels, MRI permits both the direct correlation of function with underlying anatomy and repeated studies on the same person. It may become the method of choice for studies of mental and cognitive processes, presurgical mapping, monitoring recovery from stroke or head injuries, exploration of seizure disorders, or monitoring the effects of neuropharmaceuticals.
Dr. Denis Le Bihan (Diagnostic Radiology Department, The Warren G. Magnuson Clinical Center, National Institutes of Health [NIH], Bethesda, Maryland): We review the basic principles and potential applications of functional magnetic resonance imaging (fMRI) of the brain. In fMRI, scientists and clinicians from many disciplines have a new tool at their disposal that enables them to discover, in vivo and noninvasively, where in the brain normal or abnormal cognitive processes take place.
Like positron emission tomography (PET), fMRI is based on the tight relation among neuronal synaptic activity, energy metabolism, and blood circulation, as was suggested at the end of the last century by Roy and Sherrington . However, some of the limitations PET has for brain activation studies are overcome by MRI. First, MRI is completely noninvasive and does not use ionizing radiation; thus, it allows for repeated studies on individual persons. Second, PET facilities are few, but many MRI units are available for patient scanning. Finally, MRI can be used to obtain images with a resolution of better than 1 mm in less than one tenth of a second. Fast imaging methods, such as echoplanar imaging (EPI) , can now run on clinical MRI units and provide images of the working brain almost in real time. Activation images obtained with the Oxygen-15-labeled water PET technique take at least 40 seconds and have a resolution of better than 4 mm.
Belliveau and colleagues  were the first to show that MRI could be used for activation studies. One can obtain images of blood volume by injecting a bolus of paramagnetic contrast agent, such as gadolinium diethylene triamine pentaacetic acid, into the bloodstream. Using this approach, Belliveau and colleagues obtained images of blood volume before and during stimulation with a flashing light and showed that the blood volume of the primary visual cortex increased during stimulation.
This approach constituted a significant breakthrough but had several limitations, one of which was that it required the use of an external contrast agent. On the basis of earlier works describing the magnetic properties of hemoglobin  and the effect of these properties on the magnetic resonance signal , it was recently suggested that deoxyhemoglobin (which is paramagnetic, although oxyhemoglobin is not) in erythrocytes could be used as an endogenous source of contrast material [6, 7]. Kwong and colleagues , soon followed by others [9, 10], showed that MRI could be used to monitor in real time the local modulation of the level of blood oxygenation associated with brain activity.
After a brief overview of the principles of MRI and fMRI, we review the first applications of fMRI in the visual and sensorimotor systems and in some cognitive tasks, such as language, memory, and behavior. Finally, we summarize the potential clinical applications of fMRI and review the current limitations of this technique and current issues in its research.
Dr. Peter Jezzard (Laboratory of Cardiac Energetics, National Heart, Lung and Blood Institute, NIH, Bethesda, Maryland): The nuclear magnetic resonance effect, first shown in the 1940s, is the result of the magnetic properties of some atomic nuclei, especially hydrogen nuclei from water molecules. When a person is placed in an MRI scanner, the magnetic fields from the hydrogen nuclei orient themselves along the strong magnetic field of the scanner. This orientation may be disturbed by exciting those nuclei with a burst of electromagnetic energy in the form of radiofrequency pulses. The nuclei then realign themselves by transmitting a radiofrequency signal that can be detected in a receiver coil placed around the patient's head. The frequency of the signal depends on the strength of the scanner's magnetic field; by causing the strength of this magnetic field to vary spatially, the resulting signal reflects the magnetization of water molecules at each point in space, thus producing an image .
An increasingly popular MRI technique, used extensively for brain scans, is EPI , which can efficiently sample the spatial information of the object after a single excitation of the water molecules. In this way, all data needed for an image can be obtained in less than 100 milliseconds, whereas a more conventional imaging sequence would take several tens of seconds (images resulting from the latter process would, however, have greater spatial resolution). This type of fast sequence is particularly useful for sampling large portions of the brain many times at short intervals to monitor brain activity.
The most commonly used fMRI technique relies on a subtle increase in signal intensity in the regions activated during neuronal stimulation, which can be enhanced in a carefully designed experiment. The cause of this small increase in intensity is ascribed to a local change in the blood oxygenation balance that results from the neuronal stimulation. It has been known for many years that deoxyhemoglobin is magnetically different from oxyhemoglobin [4, 5] and that, in the case of blood passing through tissue, the presence of deoxyhemoglobin creates point magnetic inhomogeneities within the blood vessels that result in microscopic field distortions around the vessels. The coherence of the signal from hydrogen nuclei in water in surrounding tissue is partially destroyed by those microscopic inhomogeneities, and, thus, the signal intensity is lower than it would be if they did not exist. During neuronal activity, local metabolism increases and local blood flow increases substantially, more than compensating for the demand for extra oxygen caused by the neuronal activity . As a result, the local concentration of deoxyhemoglobin decreases, and the signal in those regions increases in intensity.
Experiments have been done at various scanner field strengths (most clinical MRI systems range from 0.5 to 1.5 tesla). At NIH and numerous other sites, researchers have access to even higher field strengths for studies of humans (>3 tesla), which allows field strength to be compared. Because the MRI effect during neuronal stimulation has been shown to increase linearly and quadratically with field strength, high field strength is desirable, although it causes more technical difficulties.
In a typical fMRI study, many images are collected at intervals of a few seconds over a period of several minutes to improve the statistical reliability of the data. An idealized time plot from a region of the brain in which signal changes occur is shown in Figure 1. The signal changes correspond to the applied stimulus but are somewhat delayed and smeared by the hemodynamic response of the patient. Figure 1 also shows an actual time course from a point located in the primary visual cortex during a visual stimulation study in which a flashing light stimulus was presented to the patient for periods of 30 seconds alternated with 30 seconds of rest. If statistical correlation or other statistical techniques are used, the data can be analyzed point by point to map regions of brain function.
Schematic time course from an idealized region of tissue in response to alternating periods of stimulus and rest. Time course from an experiment (visual cortex) in which a visual stimulus was presented to the participant. Note the delay and smear of the data relative to the input condition; this is caused by the slow (on the neuronal timescale) hemodynamic response.
Dr. James Haxby (Section on Functional Brain Imaging, Laboratory of Psychology and Psychopathology and Pathology, National Institute of Mental Health [NIMH], NIH, Bethesda, Maryland): An extensive body of literature on visual neuroanatomy and neurophysiology in nonhuman primates provides a detailed guide for directing functional brain imaging investigations of human vision [13, 14]. Functional brain imaging studies of the human visual cortex, done using PET, have shown it to have substantial homology with the visual cortex in nonhuman primates. Fox and colleagues  showed the retinotopic organization of receptive fields in the human primary visual cortex. In the extrastriate cortex, PET studies have identified an area in the lateral occipitotemporoparietal cortex that responds selectively to visual motion [16-18] and that may be the human homolog of the middle temporal/middle superior temporal complex (MT/MST, also called V5) in the macaque (Macaca mulatta). In the lingual gyrus, an area that is selectively activated by viewing colored stimuli or selectively attending to color has also been identified [16-18].
Many of the early fMRI studies were done on the human visual cortex [3-8]. More recent studies have taken advantage of superior spatial resolution to generate more detailed maps of the retinotopic organization of the primary visual cortex . Additional representations of the visual field obtained by mapping retinotopy in individual brains have suggested the locations of early extrastriate regions. The area that responds selectively to visual motion has also been identified using fMRI . This study showed that stimuli with low luminance contrast were as effective as high-contrast stimuli in activating this area; striate cortex, however, showed different responses to low- and high-contrast stimuli. This finding is consistent with the neurophysiology of area MT (middle temporal area) and further corroborates the homology between the human and nonhuman primate visual motion areas. This study also shows that MRI can continually measure the strength of response, thereby allowing studies of the effects of varying stimulus measures that may distinguish visual areas from one another.
Studies with PET have also shown that the human visual extrastriate cortex, like that of the M. mulatta, appears to be organized into a ventral, occipitotemporal stream associated with the visual identification of objects and a dorsal, occipitoparietal stream associated with perception of the spatial relation among objects and with visual guidance of movements toward objects . My colleagues and I have shown [22, 23] that selectively attending to face identity in a test of visual perception selectively activates ventral occipitotemporal areas extending, in an anterior direction, to the mid-fusiform gyrus (Figure 2). Selectively attending to the location of pictures of faces, on the other hand, selectively activates occipitoparietal areas in the dorsolateral occipital cortex, the posterior superior parietal cortex, and the intraparietal sulcus. Independent PET studies of face perception and spatial vision have corroborated these results [24-26].
Direct comparison of areas activated by selective attention to face identity and a sensorimotor control task in a positron emission tomography (PET) (regional cerebral blood flow) study of nine participants (left; adapted from reference 23) and a functional MRI study of a single participant ( ). The areas shown in color showed a significant difference between the face-matching and control tasks ( = 0.001 for both). The PET results are shown on coronal sections taken from the Talairach and Tournoux stereotactic brain atlas, 85 mm and 60 mm posterior to the anterior commissure. The functional MRI results are shown on coronal sections of a high-resolution MRI scan of the same participant. Data were obtained from echoplanar imaging volume sets of eight coronal sections, each 6 mm thick, collected every 6 seconds in the right posterior extrastriate cortex.
We have replicated these findings using fMRI and the same stimuli and tasks (Figure 2). Participants alternated a visual matching task (face or location matching) with a baseline, sensorimotor control task and changed tasks every 30 seconds for 2 minutes. Because PET studies are limited by the amount of radiation to which normal control participants can be exposed, data are usually averaged across participants to achieve acceptable signal-to-noise ratios. Preliminary results from the fMRI study show activation in the same general areas identified by the PET study, but activations in individual participants are more discrete than the group average areas shown by PET. Moreover, the neuroanatomical locations of functional areas relative to sulcal landmarks differ among participants. For example, the more anterior focus of activation associated with face perception in Figure 1 was in the occipitotemporal sulcus. Other participants, however, had a focus of activation on the crest of the fusiform gyrus.
Dr. Norihiro Sadato (Human Motor Control Section, Medical Neurology Branch, National Institute of Neurological Disorders and Stroke [NINDS], NIH, Bethesda, Maryland): In the human motor pathways, the neurons that regulate the activity of individual body parts are topographically arranged. Damage to a particular subdivision will produce specific deficits of motor function . Functional MRI done during simple motor tasks, such as apparent finger tapping, could provide a useful way to map these pathways and, thus, to localize lesions. Initially, those maps were obtained using electrical stimulation of the cortex . More recently, noninvasive methods such as PET [29-31] and transcranial magnetic stimulation [32, 33] have been used, but with limited spatial resolution.
Surgery for epilepsy done near functionally essential cortical areas is one situation in which direct, invasive, electrical brain mapping is currently used to avoid postsurgical deficit. Recently, the Mayo Clinic group found a correlation between mapping done with fMRI and invasive cortical mapping  in a 24-year-old woman with simple partial motor seizures that had resisted medication for 17 years. Conventional MRI showed a left frontoparietal tumor close to the central sulcus; fMRI images—obtained during bilateral opponent finger tapping and movement of the lower face and coregistered on high-resolution MRI images—showed that the motor cortical representation of the face and hand was superficial to the tumor. Strips of electrodes surgically placed on the brain surface adjacent to the tumor confirmed the relation between the motor representation of the face and hand and the tumor, suggesting the potential use of fMRI for presurgical mapping of the primary motor cortex.
On the other hand, it is known that about 10% of the corticospinal tract does not cross in the medulla [35, 36]. Ipsilateral innervation is of particular interest because of its potential role in functional recovery after stroke . Further, some asymmetry is apparent; left hemispheric lesions result in motor dysfunction of the left hand, whereas lesions on the right side leave the motor function of the right hand relatively unaffected [38, 39]. Kim and colleagues  reported hemispheric asymmetry in the functional activation of the human motor cortex during contralateral and ipsilateral finger movements. The right primary motor cortex was activated primarily by contralateral opposition finger movements, whereas the left motor cortex was activated by both ipsilateral and contralateral movements. The latter effect was more pronounced in right-handed than in left-handed persons, suggesting a relation between ipsilateral motor control and degree of hemispheric specialization.
At the cognitive level, the positive effect of mental practice on motor skill performance has been noticed [41, 42]. We used fMRI to investigate the participation of the primary motor cortex during mental imagery . Figure 3 shows a parasagittal fMRI of the left cerebral hemisphere during sequential opponent finger movement of the right hand with activation of the central sulcus by the actual finger movement. The signal time course Figure 3 in this region of the primary cortex acquired during the paradigm (ideation of the motor task-rest-actual movement-rest-ideation) shows that the primary motor cortex is activated during ideation, although less prominently than during actual finger movement, suggesting that processes used in the primary motor areas during motor performance may also be used during mental imagery.
The activated area of the left primary motor cortex by self-paced sequential opponent finger movement of the right hand (approximately 2 Hz). The activated area in orange showed a z-score greater than 2. Time intensity curve of the left primary motor cortex, which followed the periodicity of the activation task (rest, ideation, rest, movement, rest, ideation, and rest). Activation by actual movement was more prominent (percentage increase of signal intensity compared with rest condition was 4.14 ±1.37 [mean ±SE]; = 0.000001, rank analysis of variance) than that by ideation (1.70 ±1.08; = 0.000005, rank analysis of variance), which was still significant.
Nonprimary motor cortices (the supplementary motor area and the premotor cortex) are supposed to play a role as higher-order centers involved in the generation and programming of complex movement [44, 45]. Studies using PET have shown that the changes in cerebral blood flow in the nonprimary motor cortical areas caused by motor tasks are usually less prominent than those in the primary motor area [46-48]. Rao and colleagues  showed that fMRI was also able to detect activation of the nonprimary motor cortex. Although simple simultaneous opponent finger movement of the right hand caused activation of the left primary motor area only, complex sequential opponent finger movement also caused the activation of the supplementary motor area and the bilateral premotor areas.
Dr. Linda Rueckert (Cognitive Neurosciences Section, Medical Neurology Branch, NINDS, NIH, Bethesda, Maryland): Over the past several years, great progress has been made in the study of the cerebral basis of language. Researchers using PET [50-52] have suggested a complex network that goes far beyond that found in the early studies of Broca and Wernicke. Many have found some degree of right hemisphere activation during verbal tasks, suggesting that this hemisphere plays a greater role in language than was originally indicated by lesion data and the results of Wada tests. Because the use of PET often necessitates averaging over several patients to observe significant effects, important differences among individual persons could be obscured, as is suggested by electrical stimulation studies . Because fMRI allows for studies in individual persons, it is potentially useful for studying the wide range of variation in verbal ability and strategies for doing verbal tasks.
The results reported here are an extension of a fMRI word-generation study done by the Cognitive Neuroscience Section of NINDS . Seven normal, right-handed participants were tested using a silent word-generation paradigm. They were required to generate as many words beginning with a given letter as they could, but silently, to avoid motion artifacts. Three different control conditions were used. During the “night sky” condition, participants were asked to relax and imagine a night sky, in order to avoid verbal thinking. During the “automatic speech” condition, they were asked to silently repeat the names of the days of the week. During a third control condition, “phoneme articulation,” participants were required to silently repeat the syllable “ta.”
The location of the four 5-mm slices was chosen on the basis of previous PET studies and a pilot fMRI study. The left and right hemispheres were scanned in separate sets. Significance of activated regions was assessed using statistical z-scores calculated for each point of each slice from the average difference in signal intensity between the activated and control states .
Representative activation maps superimposed on anatomical MRI images from the left and right hemisphere of one participant are shown in Figure 4. No matter which of the three control conditions was used as a baseline, no systematic differences in activation patterns were seen. All participants showed significant activation in the left frontal operculum of the left inferior frontal gyrus below the inferior frontal sulcus (Brodmann areas 44, 45, or both; Broca area). Activation was also found in the region between the left central and precentral sulci and in two sites in front of the precentral sulcus, corresponding to the motor and premotor cortex, respectively (areas 4 and 6). All participants showed additional activation elsewhere in the left hemisphere as well as in the right hemisphere, but the location of that activation was not consistent across participants. The degree of right hemisphere activation was less than for the left hemisphere.
42.5 mm left of midline. 42.5 mm right of midline. Orange areas are those for which mean signal intensity during the word-generation periods was at least 1 standard deviation higher than the mean during the control condition. Green areas are those for which word-generation intensity was at least 1 standard deviation below that for the control condition. The isolated region to the left is the eye and shows decreased signal because of eye movements. Data were obtained using a 4-tesla magnetic resonance imaging scanner with echoplanar imaging. The image resolution is 2.5 mm. Four contiguous 5-mm slices were acquired every 3 seconds.
Several other studies have now shown the utility of fMRI in mapping cortical regions involved in language. McCarthy and colleagues , using axial EPI images, found dorsolateral prefrontal cortex activation during word generation. Other investigators have also obtained similar results using conventional MRI equipment without EPI capability [57, 58]. Other studies have examined verbal reception and semantic categorization and obtained results comparable to those reported using similar tasks with PET [59-62]. Rao and colleagues  found activation along the superior temporal gyrus during passive word presentation; Benson and colleagues  found activation in the Broca area and in the left inferior parietal lobule during a same or different word-matching decision task. In some cases, fMRI has detected activation in some participants in regions not previously implicated by PET . Some rarely activated regions might disappear in PET studies when results from participants are averaged together.
Dr. Venkata S. Mattay (Laboratory of Diagnostic Radiology Research, Office of Intramural Research, and Clinical Brain Disorders Branch, NIMH, NIH, Bethesda, Maryland): The capacity of functional neuroimaging to integrate structure with function enables us to identify neural systems that subserve complex cognitive tasks and behaviors. As such, it has proved uniquely valuable in neuropsychiatric research. Because fMRI is based on single individual analyses, it is potentially uniquely capable of highlighting differences among individual persons, whereas nuclear medicine techniques tend to look for group tendencies. Thus, fMRI can be used to predict phenomenologic variance in normal persons and in patients with neuropsychiatric disorders.
Although fMRI is still at the stage of technical validation, neuroscientists are striving to use it to replicate results obtained earlier using PET during various cognitive tasks and in neuropsychiatric disorders.
The various cognitive tasks that have been studied with different fMRI techniques include the Wisconsin Card Sorting Task , the Continuous Performance Test , the Spatial Working Memory Task , and the Stroop Conflict Paradigm .
The Wisconsin Card Sorting Task is a “gold standard” neuropsychologic task for the prefrontal cortex and involves using “working” or “scratch pad” memory. Positron emission tomography studies using Oxygen-15-labeled water  showed activation in the dorsolateral prefrontal cortical region while participants did the task. Using fMRI , we have shown similar activation in the dorsolateral prefrontal cortex in participants doing the Wisconsin Card Sorting Test. Cohen and colleagues  used fMRI to study another working memory condition, the Continuous Performance Test. Using a conventional 1.5-tesla scanner, these investigators could show activation of the dorsolateral prefrontal cortex in five of six participants, a finding consistent with previously reported PET results. McCarthy and colleagues  used EPI on a 2.1-tesla scanner to show activation of the dorsolateral prefrontal cortex with a spatial working memory task. Consistent with PET findings, Casey and colleagues  showed activation in the right anterior cingulate region during color and neutral word conditions of the Stroop Conflict Paradigm.
Researchers have also used fMRI to study clinical populations with schizophrenia and obsessive-compulsive disorder. Over the past few years, Weinberger and colleagues  have used PET to consistently show hypofrontality in schizophrenic patients doing the Wisconsin Card Sorting Test. Using fMRI, we could show similar findings in a schizophrenic patient (Figure 5).
Activation is seen in the left dorsolateral prefrontal cortex region in the normal volunteer. Lack of similar activation shows hypofrontality in the schizophrenic patient.
Breiter and colleagues  studied 11 patients with obsessive-compulsive disorders and 6 normal volunteers during resting and during provoked symptomatic states. The stimulus given to the patients was tailored depending on the obsessive symptoms; the controls were given stimuli tailored to elicit disgust. Using fMRI, these researchers had findings similar to those obtained with PET; their patients showed activation in the orbital gyri and dorsolateral prefrontal cortex; controls showed no such activation.
Although the results from these studies are preliminary, they suggest that fMRI will emerge as a useful tool in neuropsychiatry.
Dr. Le Bihan: This conference shows that fMRI has tremendous potential as a means for understanding how the human brain works. Most studies have already replicated many of the key findings of PET studies and have taken advantage of the superior spatial resolution of fMRI to provide more detailed activation maps. We were able, for instance, to clearly show that the primary visual cortex was involved not only during visual perception but also during short-term memory recall . Studies using fMRI have not yet, however, taken advantage of the technique's superior temporal resolution, although the usefulness of this resolution may be limited by the smoothness of the hemodynamic response function underlying the signal change detected by MRI .
Because this method is noninvasive and does not involve exposure to radiation, it confers the advantage of repeatability. Repeated testing of a single patient will allow clinicians to follow changes in cerebral activity over the course of a progressive disease, during recovery from injury or stroke, and in response to treatment. With language studies, various groups of patients could also be evaluated, including those with progressive aphasia or children with disorders such as dyslexia or intractable complex seizures. In this example, fMRI could be used for presurgical mapping, if language is to be spared during lobectomy [72, 73]. Until now, hemispheric language dominance has been determined from invasive, risky tests, such as the Wada test, which is based on intracarotid amobarbital injection, and electrical cortical recording using electrodes positioned intraoperatively. It is clear that fMRI language studies can potentially replace such invasive tests .
The effects of psychopharmacologic treatments could be monitored on individual persons across various cognitive conditions, and within the same naive as opposed to practiced condition. At another level, if PET remains the gold standard for receptor-binding imaging, fMRI could be used to visualize regions of the brain where receptors have been activated or inhibited through effects on blood flow and blood oxygenation, as we recently suggested in the monkey brain using neurochemical activation by arecoline .
Functional MRI, of course, has limitations. First, as with conventional MRI, patients with pacemakers and magnetic implants must be excluded. The cooperation of patients is also particularly necessary: Not only is their active participation needed for the activation tasks, but they are also required to maintain absolute immobility during scanning to avoid misregistration artifacts. Re-registration algorithms and sophisticated statistical analysis packages are being evaluated. Other areas of active research involve more basic issues, such as understanding the coupling between neuronal activity and blood flow or oxygenation, and the localization, in terms of microvessels or macrovessels, of the magnetic effect responsible for the fascinating images that fMRI can now produce.
The In the Clinic® slide sets are owned and copyrighted by the American College of Physicians (ACP). All text, graphics, trademarks, and other intellectual property incorporated into the slide sets remain the sole and exclusive property of the ACP. The slide sets may be used only by the person who downloads or purchases them and only for the purpose of presenting them during not-for-profit educational activities. Users may incorporate the entire slide set or selected individual slides into their own teaching presentations but may not alter the content of the slides in any way or remove the ACP copyright notice. Users may make print copies for use as hand-outs for the audience the user is personally addressing but may not otherwise reproduce or distribute the slides by any means or media, including but not limited to sending them as e-mail attachments, posting them on Internet or Intranet sites, publishing them in meeting proceedings, or making them available for sale or distribution in any unauthorized form, without the express written permission of the ACP. Unauthorized use of the In the Clinic slide sets will constitute copyright infringement.
Copyright © 2016 American College of Physicians. All Rights Reserved.
Print ISSN: 0003-4819 | Online ISSN: 1539-3704
Conditions of Use
This PDF is available to Subscribers Only