Departments of Neurosurgery (WMM, KR), Radiology (FZY, VMH), and Neurology (TAH, GLM, SJS) and Biophysics Research Institute (RC), Medical College of Wisconsin, Milwaukee, Wisconsin
OBJECTIVE: The purpose of this study was to determine the usefulness of functional magnetic resonance imaging (FMRI) to map cerebral functions in patients with frontal or parietal tumors.
METHODS: Charts and images of patients witth cerebral tumors or vascula malformations who underwent FMRI with an echoplanar technique were reviewed. The FMRI maps of motor (11 patients), tactile sensory (12 patients), and language tasks (4 patients) were obtained. The location of the FMRI activation and the positive responses to intraoperative cortical stimulation were compared. The reliability of the paradigms for mapping the rolandic cortex was evaluated.
RESULTS: Rolandic cortex was activated by tactile tasks in all 12 patients and by motor tasks in 10 of 11 patients. Language tasks elicited activation in each of the four patients. Activation was obtained within edematous brain and adjacent to tumors. FMRI in three cases with intraoperative electrocortical mapping results showed activation for a language, tactile, or motor task within the same gyrus in which stimulation elicited a related motor, sensory, or language function. In patients with >2 cm between the margin of the tumor, as revealed by magnetic resonance imaging, aand the activation, no decline in motor function occurred from surgical resection.
CONCLUSIONS: FMRI of tactile, motor, and language tasks is feasible in patients with cerebral tumors. FMRI shows promise as a means of determining the risk of a postoperative motor deficit from surgical resection of frontal or parietal tumors. (Neurosurgery 39:515521, 1996)
Key words: Brain mapping, Cerebral neoplasms, Cortical stimulation, Functional magnetic resonance imaging, Functional mapping, Intraoperative brain mapping
Functional magnetic resonance imaging (FMRI) is a technique for mapping regions of the brain in which neurons are active in the performance of sensory, motor, and cognitive tasks (28, 10, 11, 1420, 23, 25, 27). The change in signal detected by FMRI in the brain ("activation") is presumed to result from changes in regional cerebral blood flow and deoxyhemoglobin linked to neuronal activity (19, 20). FMRI has been used experimentally to study the organization of cerebral functions in the normal brain (36, 10, 1417, 23, 25). It has also been used in brain mapping before craniotomy (13). FMRI seems to map the sensorimotor cortex accurately. Functional imaging potentially facilitates the assessment of the risk of damaging eloquent brain in the surgical resection of cerebral tumors. It may have an application in the selection of patients for craniotomy. The reliability of FMRI techniques in patients with raised intracranial pressure, cerebral edema, or neurological impairment is being studied. The purpose of this report is to describe early results of FMRI with echoplanar (EP) acquisition in a series of patients with cerebral tumors.
The functional magnetic resonance images were obtained in a commercial 1.5-T imager (General Electric Medical Systems, Waukesha, WI) equipped with a "birdcage" type head coil and three axis gradient coils (26). A series of localizer images was obtained in axial, coronal, and/or sagittal planes. On the basis of these images, parasagittal planes of section from the lateral surface of the widest portion of the brain to the midline or through the opposite hemisphere were selected for obtaining the anatomic reference images on which the functional images were superimposed. Typically, these anatomic reference images were obtained with SE acquisitions as follows: TR, 600 milliseconds; TE, 20 milliseconds; excitations, 20; fields of view, 24; matrix, 128 x 256; and slice thickness, 1 cm. A series of 140 single-shot blipped EP gradient recalled images was obtained at 1-second intervals in each of the planes selected for the anatomic imager. During the acquisition of the EP images, there were three periods of task performance (duration of each, 20 s) interspersed with four periods of rest (duration of each, 20 s). The technical parameters for EP images included the following: TR, 1000 milliseconds; TE, 40 milliseconds; excitations, 1; matrix, 64 x 64; fields of view, 20; slice thickness, 1 cm; and acquisition time, 40 milliseconds. Acquisition of the anatomic and EP images usually took ~1 hour. The series of images was viewed in cine mode to reject cases with head movement. The time course of the signal intensity over 140 seconds in each pixel was plotted and compared with a reference function by means of a cross correlation program (1). The reference function was a modified square wave with a period of 40 seconds and the first 5 seconds of the time course and the first 5 seconds of each rest and task period excluded in the correlation calculation. Pixels with a correlation coefficient of greater than or equal to 0.6 were displayed as "activated" pixels. The images of the activated pixels were then overlaid on the corresponding anatomic reference images by means of the image processing program.
Three tasks were used in this study. The motor task consisted of repeated, self-paced, rapid (greater than or equal to 2 Hz), rhythmic appositions of the right thumb and first finger in response to a cue from the investigator; the tactile stimulation task consisted of the investigator gently and rhythmically scratching the ventral surface of the patient's right or left hand with his or her finger tips during the prescribed 20-second time periods, and the language task consisted of word generation performed silently and audibly (28). To perform word generation, the patient, on cue, thinks or says aloud as many words as possible that begin with a letter that is specified by the investigator during each task period.
The location of the activation in FMRI was compared with intraoperative electrical cortical stimulation (IOECS) mapping, if available. The IOECS techniques with bipolar electrodes have been described (21). Activated pixels anatomically related to the sinuses and isolated activated pixels were disregarded. The sensorimotor cortex was defined in parasagittal functional images as the region with multiple contiguous activated pixels.
| Age (yr)/Gender | Lesion | Lesion Location | Task | Task Result | Distance from Edge of Abnormal Tissue | Intraoperative Cortical Mapping |
|---|---|---|---|---|---|---|
| 33/M | Arteriovenous malformations, reactive gliosis | R temporal | L motor, L sensory, word generation | + | >2 cm | Confirms FMRI |
| + | >2 cm | |||||
| + | >2 cm | |||||
| 17/F | Astrocytoma Grade II | L frontal | R motor, R sensory, word generation | + | <0.5 cm | Confirms FMRI |
| + | <0.5 cm | |||||
| + | <0.5 cm | |||||
| 8/M | Oligodendroglioma | L parietal | R motor, R sensory, L sensory | + | <2 cm | N/A, no intraoperative mapping |
| + | 12 cm | |||||
| + | ||||||
| 53/M | Astrocytoma Grade IV | L parietal | R motor, R sensory | + | <2 cm | N/A, no intraoperative mapping |
| + | <2 cm | |||||
| 41/M | Astrocytoma | R frontal | L sensory | + | 1.52 cm | N/A, no intraoperative mapping |
| 35/F | Glioma mixed | L frontal | R motor, R sensory | + | <2 cm | Confirms FMRI |
| + | 12 cm | |||||
| 11/M | Glioma (presumed) | L frontal | R motor, R sensory | + | 1 cm | N/A, surgery pending |
| + | 0.51 cm | |||||
| 41/M | Arteriovenous malformations and hematoma | L frontal | R sensory, R motor | + | <0.5 cm | N/A, no intraoperative mapping |
| + | <0.5 cm | |||||
| 33/M | Astrocytoma presumed | R frontal | R motor, R sensory, word generation | + | <0.5 cm | Radiation therapy |
| + | <0.5 cm | |||||
| + | 0.51 cm | |||||
| 56/F | Astrocytoma Grade IV | R parietal | R motor, R sensory | + | <2 cm | N/A, no intraoperative mapping |
| + | 12 cm | |||||
| + | 2 cm | |||||
| 48/M | Astrocytoma Grade II | R frontal | L motor, L sensory, R motor, R sensory, word generation | + | <0.5 cm | N/A, no intraoperative mapping |
| + | <0.5 cm | |||||
| + | ||||||
| + | ||||||
| + | 0.51 cm | |||||
| 23/F | Astrocytoma Grade II | L frontal | R motor, R sensory | | N/A, no intraoperative mapping | |
| + | 0 |
In 12 patients, a group of 8 to 20 activated pixels defining the sensorimotor cortex was identified with the motor and/or the sensory task (Fig. 1). Activation was obtained with the tactile tasks in 12 patients with midline shifts to 10 mm; with treatment with medications, including steroids, anticoagulants, beta receptor blocking agents, analgesics, antibiotics, and morphine; and with tumors or edema within 5 mm of the sensorimotor cortex. In 10 of 11 patients who performed the motor tasks, activation was obtained. The activated pixels for the motor and tactile tasks partially overlapped in the 10 patients for whom both results were obtained. In two patients, no motor activation was obtained. One was a 41-year-old man with left frontal arteriovenous malformations who was unable to perform finger-to-thumb apposition because of impaired attention span. The other was a 23-year-old woman with a left frontal astrocytoma who performed finger-to-thumb apposition slowly and hesitantly and had no demonstrable activation with the FMRI techniques used. In that patient, tumor and edema involved the rolandic cortex (Fig. 2). The tactile stimulation task produced activation in the sensorimotor cortex in both of those patients. Other groups of activated pixels were identified in the midline cuts in the region of the supplementary motor cortex during the performance of the motor task. In the four patients who performed word generation, activation was identified in the inferior frontal lobe. Activation was also identified in the sensorimotor cortex and in the supplementary motor area from word generation.
In 9 of the 12 patients, surgical biopsy or excision was performed. Intraoperative cortical stimulation mapping was performed in three of the patients. In those three patients, motor function in the hand or finger was identified in the same gyri and in the same locations in which FMRI demonstrated activation. In two patients with IOECS, speech arrest was observed in regions in which FMRI slowed activation in response to word generation or counting (Fig. 3). Postoperatively, in 11 patients with >5 mm of margin between lesion and sensorimotor, no worsening of sensorimotor functions was detected after surgery. One 41-year-old man who had right hemiparesis and activation contiguous to frontal arteriovenous malformations before surgery experienced worsening of the hemiparesis after surgery. In one case of a large left frontal astrocytoma, activation was identified from sensory and motor tasks within 5 mm of the tumor, and activation from language tasks was identified in the left frontal lobe near the tumor. A biopsy showed astrocytoma Grade 2, and radiation therapy was performed.
The accuracy of FMRI mapping has been verified, in some cases, by reference to other functional mapping techniques. Intraoperative electrocortical mapping and FMRI correlate well, according to investigators who have studied the congruence of the two methods (12, 30). The methodological problems are significant. FMRI measures cerebral changes between the task and resting state, whereas the intraoperative technique is based on responses to stimulation of the cerebrum. FMRI records changes in flow in capillaries or veins draining a region of brain, whereas intraoperative stimulation is based on changes caused by the polarization of neurons. Only cortical tissue on the surface of the brain can be tested with cortical stimulation, whereas activation deep in the brain can be measured with FMRI. The methods used for indexing the IOECS map to the FMRI by means of vascular landmarks are not exact. Although close correlation was found in our study between IOECS and FMRI, only three patients were studied with the two methods. More rigorous comparison methods and larger series are needed to verify the spatial accuracy and specificity of FMRI.
The motor and the sensory tasks did not activate uniquely the pre- and postcentral gyri, respectively. The two tasks typically produce overlapping regions of activation (11, 29), for reasons that have not been explained. The motor tasks necessarily produce some sensory stimulation. The sensory task may elicit a motor response that is not visible to the observer. The sensory and motor strips may not be as discrete as presently thought (26).
Eloquent brain can be mapped by means of several methods, including FMRI, magnetoencephalography (24), and positron emission tomography (9, 22). FMRI has the advantage of providing a high-resolution anatomic image for reference and greater accessibility.
The use of FMRI is limited to patients who have no contraindications to magnetic resonance imaging and have the ability to cooperate. Patients who do not hold still for each acquisition are poor candidates for FMRI. Any contraindication to magnetic resonance imaging excludes the use of FMRI. Good functional images are not obtained near large blood vessels or metallic or paramagnetic material. Motion of the head in the field of view or of objects with magnetic susceptibility outside the field of view degrade images. To date, techniques and paradigms are incompletely standardized and optimized. Optimal thresholds have not been determined. The specificity and accuracy of the technique are insufficiently studied.
This study suggests that successful FMRI mapping may be obtained routinely in patients with cerebral tumors. Despite inconsistencies in patient cooperation, in medications, and in intracranial pressures, activation is documented by FMRI in such patients (Fig. 4). The relationship of tumors to eloquent brain is difficult to determine in conventional magnetic resonance imaging when the anatomic landmarks for identifying eloquent brain are distorted (24). This study suggests that activation can be obtained from sensory or motor tasks with FMRI, even when the sensorimotor cortex has a close anatomic relationship to tumor and edema. Good test-retest precision is achieved with the FMRI methods used in this study (29). FMRI shows promise as a means to determine preoperatively the risk of a postoperative neurological deficit from resection of cerebral tumors.
Received, January 31, 1995.
Accepted, March 20, 1996.
Reprint requests: Victor M. Haughton, M.D., Department of Radiology, The Medical College of Wisconsin, John L. Doyne Hospital, 8700 West Wisconsin Avenue, Milwaukee, WI 53226.
Edward C. Benzel
Albuquerque, New Mexico
Mueller et al. obtained FMRI recordings for 12 patients with a variety of tumors or arteriovenous malformations (AVMs), all of whom were expected to undergo surgery. Sensory mapping and motor mapping were attempted in all 12 patients, and language mapping was attempted in 4. Good sensory signals were produced in all 12 patients, and good motor signals were produced in 10 of the 12 (because of motor dysfunction in the two failures). Some cortical activation was recorded in all four patients in whom language mapping was attempted. The success rate of functional mapping is very good.
Three of the patients underwent intraoperative electrical mapping during surgery, and the correlation between the FMRI and the electrical mapping in those three was good at the gyral level of precision. These results confirm the reliability of FMRI for functional localization in the superficial cortices and suggests localization accuracy of the deep cortices.
This article seems to serve as demonstration that it is possible to achieve FMRI activation mapping in patients with brain lesions. However, the realization of clinical usefulness will depend on more extensive studies to measure the accuracy of the method.
The authors found that the sensory and motor protocols produced overlapping regions of cortices, which they attributed to motor tasks producing sensory stimulation (and vice versa) and physiological overlapping of motor and sensory functions in the same gyrus. There may be other factors, such as a limitation of the spatial resolution of the method or a limitation in the ability to separately stimulate sensory or motor cortex. The latter may well result from the relatively limited time resolution of this method, related to the blood flow response times of a few seconds. The use of magnetoencephalography to localize regions of sensory and motor cortex, as described by Gallen et al. (1) and Sobel et al. (2), has a time resolution of milliseconds, which enables the separation of neuronal groupings responsible for those functions by differences in latency of response.
The authors have proven that FMRI is useful in determining the operability of tumors located in or adjacent to the sensorimotor area by indicating surgical tumor resection performed 2 cm away from the margin of the tumor, as revealed by magnetic resonance imaging; the FMRI activation resulted in no decline in motor function. In addition to functional mapping, however, the resection of AVMs in the sensorimotor area requires strict adherence to the technique specific for functional areas (4). The technique includes superselective interruption of shunting arterioles and communicating venules and cleavage formation between the venous loops of AVMs and brain tissue. This basic technique allows for the preservation of the sensorimotor cortex and cortical veins passively dilated by arteriolized blood (which will be reversed to venous blood postoperatively) and therefore avoidance of a large cerebral excavation after resection of AVMs (3). At Loma Linda University Medical Center in cooperation with Scripps Clinic, we have found magnetic source imaging to be a valuable intraoperative guide for resection of AVMs in the sensorimotor area, with satisfactory results. The successful FMRI report presented by Mueller et al. is a helpful addition to the literature, and I acknowledge the contribution of the challenging, very delicate, and painstaking studies that are applied directly to the clinical setting.
Shokei Yamada
Loma Linda, California
Return to the Neurosurgery Home Page
