Neuroimaging expands with functional MRI

Technique provides vital information for presurgical mapping and evaluation of lesions near eloquent cortex

By Joseph A. Maldjian, M.D., and Jonathan H. Burdette, M.D.

We have all seen the beautiful pictures in the lay press. Multicolored blobs of brain activity “lighting up” on exquisite, high–resolution MR images. These pictures are generated using functional MRI (fMRI), a method of identifying regions of the brain that are active in response to a specific task. Functional MRI has been around for more than a decade and has revolutionized the field of neuroscience. It can provide important clinical information for presurgical evaluation of patients with lesions near the eloquent cortex.^1,2

While it has been predominantly a research tool so far, as familiarity with the technique grows among referring clinicians and the lay public, radiologists will find themselves increasingly asked to perform fMRI studies. Thus, it is important to know what fMRI is and what it is not, what the potential clinical indications are, and how it is performed in a reasonable manner.

Functional MR identifies active regions of the brain, but does not identify neuronal firing. It relies on blood flow changes occurring over several seconds.

When I tap my index finger, blood flow increases to the primary motor cortex responsible for the movement, along with a concomitant increase in oxygen delivery. While local oxygen extraction increases simultaneously, the increase in oxygen extraction is less than the increase in oxygen delivery. This mismatch results in a relative decrease in deoxyhemoglobin at the post–capillary level, and this is the basis for the fMRI response. The sequence of events has been called the blood oxygen level–dependent (BOLD) effect.³

Deoxyhemoglobin is paramagnetic and tends to diminish signal intensity on a T2*–weighted image. Since there is a slight decrease in the deoxyhemoglobin in the activated state, there will be a slight increase in signal intensity on a T2*–weighted image. It is this small change in signal intensity that generates functional MR maps. The change in signal is very small and requires the acquisition of many images to be detected with confidence.

How It Is Done

In a typical fMRI examination, rapid imaging is performed—typically, an entire volume is acquired through the brain every two seconds—while the subject cycles through blocks of rest and task. For a motor task, the process can take several minutes, during which the subject alternates tapping the index finger for 20 seconds and resting for 20 seconds. The images, several hundred at each slice location, are subjected to a statistical test comparing the “on” images with the “off” images. The resulting statistical map is thresholded for significance, and any surviving voxels are color–coded and overlaid onto a high–resolution MR image of the subject. The process can involve additional steps to improve the quality of the data, including motion correction and various forms of noise filtering or smoothing. Although the final images are overlaid onto high–resolution T1 images, the resolution of the actual images used to generate the areas of activation is often on the order of 4 × 4 × 5 mm.

As to paradigms, the task paradigm is the activity that the subject is to perform in the scanner. The only limitations are that there should be no head motion, and there should be a properly balanced baseline condition. Ideally, the task would differ from the baseline condition only in terms of the activity we want to identify.

A routine, clinical 1.5–tesla scanner is adequate for these studies. The de facto standard for performing fMRI is some variety of echo–planar imaging, including spiral, which is the only imaging method that permits the necessary temporal resolution and slice coverage for whole–brain functional studies. With the advent of diffusion imaging for stroke evaluation, many scanners already have echo–planar capabilities. Although field strengths of 3 tesla and higher can provide increased sensitivity to the BOLD effect,⁴ the vast majority of studies performed to date have been at 1.5–tesla, and this is expected to continue for at least the next several years.

Dealing With The Data

Vendors are addressing many of the numerous technical issues involved in performing fMRI. In the past, performing a functional imaging study required the acquisition of raw data, transfer of the data from the MR computer to another workstation, off–line reconstruction, and extensive computer work to generate an fMRI map (motion correction, normalization, statistical mapping, display). Solutions are becoming available to remove many of the technical hurdles for clinical fMRI. Although such programs should provide an easier path to the end result, the usefulness of the functional image will depend on the quality of the study and the robustness of the task paradigm.

Several vendors, among them Avotech and Resonance Technology, sell equipment for administering paradigms within the magnet environment. For the clinical paradigms described, auditory stimulation equipment alone can suffice, and sound capability is already available for the patient’s comfort at many clinical MR installations.

For more complex, visual paradigms, a generic LCD projector and screen at the foot of the gantry are suggested. The subject views the screen through a prism within the head coil. The output of a laptop computer at the console is fed to the projector, and responses can be recorded using the same computer.

To accommodate visually impaired subjects, we have purchased plastic–framed eyeglasses with various diopter levels. Higher end systems provide LCD goggles with eye–tracking capabilities, magnet–compatible headphones, and a variety of stimulus response devices (button pads, track balls, glove sensors, etc.). The price tag for a complete system can range upwards of $80,000. The response device and auditory system are best purchased through a vendor that specializes in the systems. MR vendors can also provide stimulus presentation hardware and software as part of a turnkey solution.

Clinical Indications

The main clinical indication for performing fMRI is preoperative mapping of eloquent cortex near a brain tumor or arteriovascular malformation. Although a number of paradigms exist, only two are usually required for clinical purposes: a motor paradigm and a language paradigm. The neurosurgeon is concerned about the location of eloquent cortex in relation to a brain tumor, and this means motor function and language function. Motor paradigms are fairly robust, and reproducible activation can be achieved with simple paradigms (Figure 1).

The most basic of these involves self–paced finger tapping, which can be cued via the system intercom (20 seconds on, 20 seconds off for five minutes). We play a CD to the patient through MR–compatible headphones with a metronome beating at a tick per second throughout the paradigm. The subject paces the tapping to the metronome. Every 20 seconds, the subject is instructed to begin or end tapping. Typically, with motor paradigms, activation can be expected in both contralateral motor and sensory regions, since there is a sensory component to any motor task. In addition, the ipsilateral cerebellum and the contrateral supplementary motor area (responsible for planning motor movements) can occasionally be identified. The supplementary motor area tends to be demonstrated with more complex motor tasks. If the lesion is medially located, a toe-tapping paradigm can be substituted for the finger tapping.

Many language paradigms exist, all of which have been shown effective in activating either frontal and/or temporal lobe language areas. These include word generation (prompted by a letter), semantic decision–making (e.g., Is the displayed word a vegetable?), and even simple auditory presentation of words. One of the problems in many of these paradigms, however, is the appropriate baseline condition, which needs to be matched in word length, word content, symbolic representation, etc. Also, these paradigms vary in their ability to activate frontal and temporal lobe areas, as well as their ability to lateralize function.

We have begun using a paradigm for language function that involves listening to text played forward (active condition) and backward (baseline condition). Such a paradigm can be easily constructed from audio tracks of a story played forward and backward using a generic CD authoring program. This paradigm has demonstrated robust activation of frontal and temporal lobe language areas, as well as excellent lateralization of function in these lobes (Figure 2). Other paradigms, such as those that provide visual stimulation using a flashing checkerboard, are also fairly simple to administer (Figure 3). A variety of useful paradigms are listed in the table.

Another potential indication for fMRI is noninvasive Wada testing. The Wada test, also known as the sodium amobarbital test, involves invasive carotid arteriography to identify hemispheric dominance prior to seizure surgery (i.e., temporal lobectomy). A catheter is placed within the internal carotid artery, and sodium amytal is infused to anesthetize one hemisphere of the brain. The patient then performs tasks (language and memory) to determine any deficit. The procedure is repeated on the contralateral side and the degree of lateralization of function is determined. A similar determination can be performed using fMRI, and a variety of tasks have been described, including language and memory.

In general, laterality indices compare the ratio of activated pixels in one hemisphere with the other to determine hemispheric dominance. The forward–backward language task described above has demonstrated excellent hemispheric lateralization and can be sufficient to demonstrate hemispheric dominance. At this juncture, however, clinical decisions for temporal lobectomy should not be made solely on the basis of an fMRI examination.

Keeping Referrers Happy

When referring clinicians order a clinical fMRI study, they expect to view the results painlessly, obtain an interpretation, and go over the results with a radiologist who is knowledgeable about both the functional aspects of the study and the clinical implications. We have adopted the practice of generating a color print of the results of the fMRI study, overlaid onto the patient’s anatomic images, which is sent to the referring clinician with a typed report of the results. We always include a disclaimer about the investigational nature of fMRI, and warn that it should not be used as the sole determinant for clinical decisions. With PACS, the fMRI images will also be available for viewing on any image–review workstation.

As our neurosurgical colleagues become involved in ordering fMRI studies for preoperative mapping, the question of integration with intraoperative neurosurgical workstations arises. These neuronavigational workstations provide the neurosurgeon with a probe–guided MR display of the brain in the operating room as the procedure is being performed. In the past, we have accomplished the integration by “reverse engineering” the proprietary vendor data formats specific to the neuronavigational workstation.^5–7

Current–generation neuronavigational workstations will accept DICOM images. The fMRI solutions provided by the MR vendors should include PACS integration of fMRI images fused with high–resolution anatomic images. These images can then be downloaded from the PACS directly onto neuronavigational devices for intraoperative guidance (Figure 4). Manufacturers that come up with turnkey solutions will obviate many technical issues, even as they recognize the demand for clinical fMRI.

Dr. Maldjian and Dr. Burdette are neuroradiologists at Wake Forest University in Winston–Salem, NC.

References

1. Maldjian J, Atlas S, Howard R, et al. Functional magnetic resonance imaging of regional brain activity in patients with intracerebral arteriovenous malformations before surgical or endovascular therapy. J Neurosurg 1996;84:477-483.
2. Atlas S, Howard R, Maldjian J, et al. Functional MRI of regional brain activity in patients with intracerebral gliomas: findings and implications for clinical management. Neurosurgery 1996;38:329-338.
3. Ogawa S, Lee T, Kay A, Tank D. Brain magnetic resonance imaging with contrast dependent on blood oxygenation. Proceedings of National Academy of Science, U.S.A. 1990;87:9868-9872.
4. Maldjian J, Gottschalk A, Patel R, et al. The sensory somatotopic map of the human hand demonstrated at 4 Tesla. Neuroimage 1999;10:55-62.
5. Maldjian J, Schulder M, Liu W, et al. Intraoperative functional MRI using a real-time neurosurgical navigation system. JCAT 1997;21:910-912.
6. Schulder M, Maldjian JA, Liu WC, et al. Functional MRI-guided surgery of intracranial tumors. Stereotact Funct Neurosurg 1997;68:98-105.
7. Schulder M, Maldjian J, Liu W, et al. Functional image-guided surgery of intracranial tumors in or near sensorimotor cortex. J Neurosurgery 1998;89:412-418.

Back to Top