Cerebral Vascular Malformations Adjacent to Sensorimotor and Visual Cortex
Functional Magnetic Resonance Imaging Studies Before and After Therapeutic Intervention
Background and Purpose It is not known how cerebral vascular malformations affect the function of the surrounding brain. Functional magnetic resonance imaging (fMRI) can provide information about normal functional neuroanatomy and its alteration by vascular lesions and therapeutic intervention.
Methods We performed fMRI studies in 24 patients harboring vascular malformations adjacent to primary somatosensory, motor, and visual cortex. The fMRI studies consisted of the acquisition of an image time series coupled with functional activation of motor, sensory, or visual cortex in both hemispheres. Activated voxels were identified using frequency domain analyses, and their number and anatomic location were compared between the affected and unaffected hemispheres.
Results Every patient capable of performing the desired task showed functional activation. Eight patients without neurological deficits showed a symmetrical pattern of activation between the hemispheres. Each had a vascular malformation located one or more gyri from the functional region imaged. Three patients showed hemispheric symmetry in the location of activated cortex but with a marked asymmetry in the number of activated voxels. Each harbored vascular malformations located within one gyrus of the functional region and showed either subtle or no neurological deficit. Eleven patients showed hemispheric asymmetry in the location of activated cortex. In 6, the anatomic displacement appeared to be due to a mass effect of the lesion. In 5, the activation occurred at a different anatomic locale, and the patients exhibited gross neurological deficit in the respective function. Posttherapeutic changes in functional activation reflected elimination of the mass effect or recovery of clinical function.
Conclusions Systematic fMRI studies are possible in patients with vascular malformations in brain regions adjacent to primary somatosensory, motor, and visual cortex. Displacement of the activated region and hemispheric asymmetry in the number of activated voxels in the functional regions appear to reflect the anatomic and physiological impact of the vascular malformation. Changes in fMRI findings after intervention reflect the consequences of therapy and parallel clinical recovery.
Vascular malformations of the brain are frequently located adjacent to primary functional areas. Proximity to eloquent brain areas correlates significantly with the functional impact of the lesion and with the increased risk of therapeutic intervention.1 2 Vascular malformations can affect brain function by mass effect, hemodynamic changes (eg, steal phenomenon, ischemia, and venous hypertension), irritative effects, and the consequences of hemorrhage.3 4 5 6
Mapping of brain function adjacent to vascular malformations has been accomplished previously using a variety of extraoperative and intraoperative techniques, including positron emission tomography, single-photon emission CT, intraoperative and extraoperative evoked potential mapping, magnetoencephalography, and direct cortical stimulation.7 8 9 10 11 12 Mapping strategies have been used to assess the prognostic risk of proposed therapeutic interventions and to plan endovascular and surgical therapeutic procedures.
fMRI provides a new modality for noninvasive mapping of brain function. fMRI maps of sensory and motor cortex have closely agreed with maps derived from the standard techniques of intracranial stimulation and somatosensory evoked potentials,13 establishing the validity of the fMRI technique. Several reports in a small number of patients have confirmed the feasibility of fMRI studies of brain vascular malformations.14 15 However, there has been no systematic study to date of the impact of cerebral vascular malformations on adjacent primary functional areas or on the changes in functional activation with therapeutic intervention and/or neurological recovery. We hypothesized that fMRI would be feasible in patients harboring cerebral vascular malformations and that the activation pattern would correlate with the clinical condition of the patient.
Here we report a study of 24 consecutive patients harboring cerebral vascular malformations adjacent to primary sensory, motor, or visual brain regions. We analyzed fMRI studies before therapeutic intervention to compare the hemispheric symmetry in location of functional activation and in the number of activated voxels. These findings were then correlated with the location of the lesion and its clinical impact. We also correlated the results of posttherapeutic fMRI studies with the patients’ clinical course.
Subjects and Methods
The study included consecutive patients with cerebral vascular malformation who underwent lesion excision (with or without preparatory embolization) by a single specialized team during a 2-year period. In each case, the lesion was within 2 gyri of primary sensorimotor or visual areas of the brain; ie, cases were excluded from study if the vascular malformation was more than 2 gyri removed from the rolandic (central) sulcus or the calcarine sulcus as identified by structural MRI. All patients provided informed consent.
Forty fMRI studies were performed in 24 subjects (22 AVMs and 2 cavernous malformations) as summarized in Table 1⇓. Baseline studies were not performed in cases 2 and 9 because of the unavailability of fMRI. In 6 patients (cases 1, 2, 3, 4, 8, and 9), fMRI studies were repeated within 1 week after presurgical embolization with acrylate and before surgical intervention. Nine postoperative fMRI studies were conducted in 7 patients (cases 1, 2, 3, 4, 7, 8, and 13) 3 to 26 weeks (mean, 12 weeks) after surgical excision of the lesion.
The fMRI studies included a hand motor task in 22 cases and a foot motor task in 2 cases (1 patient was tested with both). Hand sensory activation was completed in 12 cases, including 6 cases in which hand motor studies were also performed. In the remaining 6 patients, hand sensory studies were conducted because significant paresis prevented hand motor testing. Four visual activation studies were completed in patients with lesions adjacent to calcarine cortex.
Images were acquired with a 1.5-T Signa MR system (General Electric) with an echo planar imaging subsystem (Advanced NMR). The patient’s head was placed within a standard quadrature head coil and immobilized with a vacuum pillow, foam wedges, and a restraining forehead band.
A T1-weighted sagittal localizer series (TR, 667; TE, 13; field of view, 24; number of excitations, 1; thickness, 5 mm; interleaved) was used to select axial slices parallel to a line drawn through the anterior and posterior commissures (AC-PC line). This permitted reproducible slice selection in repeated studies. Four to seven slices were selected for functional imaging. In addition, high-resolution T1-weighted images (TR, 500; TE, 13; field of view, 20×20; number of excitations, 2; thickness, 7 mm; skip, 0 mm) were obtained for coregistration with functional images.
Functional images consisted of a time series of 128 echo planar images for each of the selected slices using a gradient echo sequence (TR, 1500; TE, 45; flip angle, 60°; slice thickness, 7 mm; in-plane resolution, 3.1×3.1 mm). The total imaging time for each run (one series of 128 images) was 199 seconds, which included 6 RF excitations per slice before image acquisition to achieve steady-state transverse magnetization.
Patients were tested in one or more of four functional tests: (1) hand motor squeeze task, (2) foot and toe flexion task, (3) passive sensory hand stimulation, and (4) visual hemifield stimulation. For the motor squeeze task, the patient held contoured sponges in each hand and performed an alternating series of left- and right-handed squeezes during the duration of each imaging run. The patient was instructed to squeeze in synchrony with a computer-generated visual indicator that was projected onto a translucent screen positioned at the patient’s feet. Left-hand squeezes were indicated by a series of left-pointing arrows positioned to the left of a fixation cross (<<<<+), while right-hand squeezes were indicated by a series of right-pointing arrows positioned to the right of the fixation cross (+ >>>>). The arrows flashed every 436 milliseconds for 10 flashes for each hand before switching to the alternate hand. Thus, patients squeezed 10 times with one hand for 8.72 seconds and then squeezed 10 times with the alternate hand for 8.72 seconds for a total cycle time of 17.44 seconds. During each run of 128 images, the patients alternated hands for 10 cycles. A test session consisted of four runs (ie, 40 total cycles). These four runs comprised two run-order pairs such that in one run of the pair the patient started the run with right-handed squeezes, and in the other run of the pair the patient began with left-handed squeezes.
The same alternation paradigm with identical timing was used in the foot motor and hand sensory tests. In the foot motor task, the patient performed ankle and toe flexion. In the hand sensory task, the experimenter brushed the thenar region of the right and left hand in synchrony with the visual indicator. In the visual hemifield task, a black and white visual checkerboard pattern alternated between left and right visual hemifields every 6.4 seconds for a total of 14 cycles per run while the patients maintained central fixation. Within each hemifield, the checkerboard reversed its black and white squares at a rate of 8 Hz.
Before analysis, the center of mass of each image in each 128-image time series was calculated. Center of mass is sensitive to small shifts in head position, as well as drift due to system instabilities or variations in susceptibility, both sources of artifact in fMRI analysis. No patient showed a shift in center of mass >0.5 voxel width for a slice within or between runs.
Voxels activated by each task were identified using a frequency domain analysis described previously in detail.16 17 18 The 128-sample time series for each voxel was transformed into the frequency domain using FFT. With a TR of 1.5 seconds (Nyquist frequency, 0.333 Hz), the FFT analysis produced estimates of power in each voxel in the 128×64 matrix at 64 equally spaced frequencies from 0 to 0.333 Hz, resulting in 64 images corresponding to each frequency. The voxel intensity in each image was proportional to the power within each voxel at the frequency of the corresponding image. Activated voxels were identified using the following procedure: (1) The frequency spectra for each voxel were averaged across the two replications of each run order. For example, in the hand motor tasks, the spectra for the two runs beginning with a right-hand squeeze were averaged, and the spectra for the two runs beginning with the left hand squeeze were averaged. (2) The spectrum for each voxel was examined separately in each of the two averaged runs to determine whether it contained a significant power “spike” at the alternation cycle frequency (eg, 0.058 Hz for the sensory and motor tasks, 0.0782 Hz for the visual task). If the power at the alternation frequency was more than 2 SD greater than the mean power at 14 adjacent frequency points in both run orders, the voxel was retained for the following step. (4) A phase decision rule was imposed. A voxel was considered activated if the phases were 180±15° different between the two run orders (eg, for runs beginning with left-hand squeezes compared with runs beginning with right-hand squeezes). The activated voxels were then counted for each task and hemisphere.
The presence or absence of neurological deficit at baseline and after therapeutic intervention was documented in the clinical record and was compared with the fMRI activation results. An examination of the baseline fMRI studies revealed three broad groupings: (1) patients with symmetrical anatomic location and similar counts of activated voxels in the functional regions of both hemispheres, (2) patients with hemispheric symmetry in the location of activation but marked decrease in the number of activated voxels in the functional regions of the affected hemisphere, and (3) patients with a displacement of the anatomic location of the activated cortex in the affected hemisphere.
After therapeutic intervention, any change in the patients’ neurological condition from baseline was compared with changes observed in fMRI. The number and location of activated voxels were compared with the baseline values for the respective case. Patients’ neurological condition at the time of imaging was used in these determinations.
Fig 1⇓ presents the results of the baseline hand motor study for case 3, in which the activated voxels for the left-hand (blue overlay) and right-hand (red overlay) squeezes are shown superimposed on a high-resolution anatomic image. The activation pattern identified by the frequency analysis (Fig 1A⇓) is closely similar to that identified by conventional t test analysis (Fig 1B⇓). The frequency spectra for the activated voxels for the left-hand squeezes (Fig 1A⇓, blue voxels enclosed by yellow oval) showed a large peak at 0.058 Hz corresponding to the alternation period for both the left-right (solid line) and right-left (broken line) run orders (Fig 1C⇓). However, the phase angles for the activated voxels from these run orders grouped in two distinct clusters, which were approximately 180° out of phase (Fig 1D⇓). The MR signal change associated with the two run orders showed a strong periodicity, with 10 peaks corresponding to the 10 periods in which the left hand was squeezed (Fig 1E⇓). The 180° phase shift introduced by the run order is clearly evident by comparison of the solid and broken lines in these activation waveforms.
The proximity of the vascular malformation to the hand motor region did not affect our ability to functionally map this region. The functional activation did not lie within the nidus of the lesion in any of the 24 baseline studies performed. Functional activations did not follow or appear within distended feeding or draining vessels in any patient.
The baseline fMRI studies were grouped into three broad categories based on differences in the number of activated voxels in the functional region in the affected and unaffected hemispheres, and on anatomic displacement of the activated region in the affected hemisphere compared with the unaffected hemisphere. The extent of activation was quantified by comparing the activated voxel counts in the motor-sensory or visual systems of each hemisphere (Table 2⇓). Activation was found in primary motor-sensory area (BA 4, 3, and 1; the central sulcus, including the posterior bank of the precentral gyrus and the anterior bank of the postcentral gyrus); premotor area (BA 6; spanning the precentral sulcus over the dorsolateral frontal lobe, incorporating anterior precentral gyrus); supplementary motor area (BA 6; medial aspect of superior frontal gyrus, bordered posteriorly by paracentral sulcus, extending anteriorly to cover approximately the caudal half of superior frontal gyrus); and parietal area (BA 7 and 40; centered in intraparietal sulcus, involving posterior banks of both postcentral gyrus and superior parietal lobule, and anterosuperior aspect of the inferior parietal lobule). For visual tasks, activation was found in the calcarine cortex (BA 17 and 18).
Symmetry in Extent of Activation With No Displacement of Functional Region
Patients 7, 12, 13, 15, 16, 18, 19, and 22 showed symmetrical baseline activation patterns between the two hemispheres. There was no significant difference (P=.44) in the total number of activated voxels in each hemisphere (Table 2⇑), and there was no displacement of activation in the hemisphere harboring the vascular malformation. All 8 patients in this group harbored a vascular malformation that was 1 or more gyri removed from the function imaged, and none of the patients manifested any neurological deficit in the respective functions at the time of baseline imaging.
Asymmetry in Extent of Activation Without Displacement of Functional Region
Each of the 3 patients in this category (cases 10, 11, and 17) demonstrated fewer activated voxels (range of 47% to 74%) in the hemisphere harboring the vascular malformation compared with the unaffected hemisphere. The difference between the voxel counts for the affected and unaffected hemisphere for these 3 patients was significantly different from that in the group with symmetrical activation described above (P<.0001). However, despite the gross difference between the hemispheres in the number of activated voxels, there was no apparent displacement of the anatomic area of activation (Fig 2⇓). All 3 patients harbored vascular malformations located within 1 gyrus of the functional region. Two patients (cases 10 and 11) had a very mild neurological deficit at the time of baseline testing, while the other patient (case 17) had no gross clinical neurological deficit.
Asymmetry in Extent of Activation and Displacement of Functional Region
The remaining patients all showed a displacement in the location of activated cortex in the affected relative to unaffected hemisphere. These patients all harbored malformations within 1 gyrus of the function imaged. In patients 5, 8, 14, 21 (hand motor study), 23, and 24, the shift of functional activation appeared to be caused by mass effect of the lesion. The activation on the lesion side was within the central sulcus; however, the location of the sulcus was shifted. None of these patients manifested gross neurological deficits in the respective function. The results for case 14, who had an asymptomatic presentation of the vascular malformation and no focal neurological clinical signs on testing, are presented in Fig 3⇓. The rolandic region of her left hemisphere was displaced posteriorly because of the mass effect of the lesion, resulting in the shift of location of sensorimotor activation compared with the right hemisphere.
In cases 1, 3, 4, 6, 20, and 21, the activated region involved an unexpected anatomic locale. Each of these patients manifested a gross clinical neurological deficit in the respective function. An extreme example of such shift of location of activation is illustrated in case 1. The patient had a large AVM involving most of her right frontal lobe and the rostral right parietal lobe, which had caused a profound hemiparesis since childhood. The patient did not have enough motor strength in her left hand (2/5 motor strength) to complete motor activation testing, so only a hand sensory task was conducted. Both right- and left-hand sensory activation resided in the left hemisphere (Fig 4⇓). The response to sensory stimulation of her left hand evoked a smaller region of activation than the right-hand activation. The patient did perceive touch and pinprick sensation in the left hand, but it was diminished compared with the right hand.
In case 21, both hand and foot motor functions were studied. This patient had a left frontal AVM near the mesial wall, lying very close to the left foot motor area. The large lesion (4 cm) also caused a mass effect, pushing the central sulcus laterally and posteriorly. The hand motor activation on the side of the AVM was displaced by mass effect, and there was no deficit in hand strength. Foot motor activation on the left side was weak and displaced laterally and posteriorly when compared with the right side. The patient exhibited weakness of the right leg.
Seven fMRI studies were conducted after embolization but before surgical resection, and another nine studies were conducted in 7 patients after lesion resection. Five patients were studied at all three stages. Of these 5, 3 patients showed no change in functional activation after therapy when compared with the baseline study. The other 2 patients showed change in activation and/or location in the posttherapy studies.
Patients With No Change From Baseline
The 3 patients in this group (cases 1, 12, and 13) differed in clinical presentation and pathology, but none exhibited a new neurological deficit after therapeutic intervention. For example, patient 1, in whom left-hand sensory stimulation produced a left-hemisphere activation at baseline (Fig 4⇑), showed an identical activation pattern after embolization and resection. Patient 12 harbored a lesion that was more than 1 gyrus removed from her primary visual cortex and showed symmetrical activation at baseline that did not change after therapy. Patient 13 had a cavernous malformation 1 gyrus removed from her hand sensory region; resection did not alter her activation pattern.
Patients With a Posttherapy Change
Two patients (cases 4 and 8) showed significant change in the posttherapy study when compared with baseline. Patient 4 presented with acute hemorrhage from a lesion within the left central sulcus and had gross hand weakness at the time of baseline fMRI studies (Fig 5⇓). The hand sensorimotor activation (to both motor and sensory tasks) appeared to be displaced laterally at baseline compared with the unaffected hemisphere. After excision of the lesion and coincident with clinical recovery of hand paresis and hypoesthesia, the hand motor activation area returned to a position similar to that of her motor function in the unaffected hemisphere.
Patient 8 had an asymptomatic presentation and remained free of neurological deficit throughout the course of therapy. Her baseline fMRI showed a superior shift of the area of activation that was likely due to mass effect from her lesion. The patient did not show a significant change in the extent of activation but did demonstrate a shift in the location of activation after excision of her large lesion because of removal of the mass effect.
Our results confirm preliminary reports by other groups that have demonstrated the technical feasibility of fMRI studies in patients with neurovascular lesions.13 14 We have identified patterns that correlate with clinical deficit and that may reflect the functional impact of the lesion on adjacent brain. The proximity of the lesion to the functional cortical area did not appear to affect our ability to map that function using fMRI, even in cases of large high-flow lesions. The area of apparent functional activation did not lie within the nidus of the lesion or in feeding or draining vessels.
Baseline fMRI Studies
The baseline studies were grouped into three broad categories on the basis of two variables: (1) the extent of cortical activation (measured by counts of activated voxels) in the affected and unaffected hemispheres and (2) the degree of anatomic displacement of the activated cortex in the affected hemisphere. No patients showed overt neurological deficit in the tested function if a comparison of the affected and unaffected hemisphere showed no asymmetry in extent of activation and no anatomic displacement of activated cortex.
Most patients showed some small differences in the extent of activated cortex, with some of these differences likely due to the positioning of the anatomic slice examined to the location of the functional areas in each hemisphere. However, the second grouping of patients manifested a marked asymmetry in the number of activated voxels in the affected and unaffected hemisphere, with only a few voxels obtained in the affected hemisphere. In each case, the lesions were within 1 gyrus of the functional cortex. These patients showed no deficit or subtle neurological dysfunction in the tested function.
The third group was more heterogeneous and included patients with displacement of the activated cortex, with or without a hemispheric difference in the counts of activated voxels. In several patients, the displacement was apparently due to the mass effect of the lesion. These patients showed little neurological deficit. However, gross neurological deficits in the tested function were invariably present when the activated cortex was displaced to a new anatomic location.
Despite a potential effect of handedness on the symmetry of somatosensory activation, every instance of statistically significant difference in functional activation between the two hemispheres was attributable to the respective vascular malformation (Table 2⇑). Among patients not manifesting neurological deficits on examination, it is possible that more specialized testing would reveal subtle somatosensory or visual disturbances correlating with lesion location and not evident on gross clinical examination. Future studies should aim to correlate quantitative measures of subtle clinical function with quantitative patterns of activation on fMRI studies. This may validate the broad categories that we defined in this study and allow a more rational correlation of functional activation with the physiological impact of the lesions.
Of greatest interest are cases with gross displacement of functional activation in association with overt neurological deficit. Such displacement of function, in apparent adaptation to the injury of primary functional areas, has been described in the setting of stroke and other chronic neurological conditions.19 20 21 22 23 24 25 Functional MRI studies will provide a unique opportunity to monitor such functional adaptation in time and space, in close relation to clinical functional recovery. It is also possible that certain patterns of functional adaptation might be prognostically more favorable or more amenable to rapid recovery or response to rehabilitation.
Effect of Therapeutic Intervention
Our limited observations after therapeutic intervention appear to correlate with the patterns of baseline fMRI changes and the patient’s clinical course. In several cases, we documented a mass-effect physical displacement of the functional activation that was relieved by lesion resection without overt clinical consequences. In other instances, there was worsening or improvement of fMRI activation in concert with respective changes of clinical function. Further experience with posttherapeutic fMRI changes may provide more accurate prognostication and objective monitoring of functional recovery.
Noninvasive imaging of brain functions with fMRI provides a powerful and noninvasive tool to map brain activity and localized dysfunction in relation to adjacent structural lesions of the brain. In the setting of cerebral vascular malformations, this tool will refine classic concepts of brain eloquence and may provide more powerful prognostic predictors than have been possible in the past. The ability to image brain function in precise anatomic register is itself extremely useful for neurosurgical planning of lesion excision and other therapeutic interventions. This information may be integrated within the stereotactic domain, thereby eliminating any need for more invasive and intraoperative mapping. Our results confirm the possibility of performing accurate and quantitative fMRI studies of brain function despite proximity to high-flow vascular lesions. Such quantitative measures might reflect subtle functional impact of the lesion beyond the resolution of gross clinical testing. Future studies will better refine the prognostic value of this instrument and define other potential clinical applications. The adaptation of fMRI methodologies to the study of more complex functions (such as language and memory) might allow other novel noninvasive investigations of the impact of vascular malformations on these complex and highly relevant tasks.
Selected Abbreviations and Acronyms
|FFT||=||fast Fourier transform|
|fMRI||=||functional magnetic resonance imaging|
This work was supported by the Department of Veterans Affairs, by National Institute of Mental Health grant MH-05286, by National Institute of Neurological Diseases and Stroke grant NS33332, and by a Howard Hughes Medical Student Fellowship. We thank A. Anderson, H. Sarofin, M. Luby, and K. McNamara for assistance.
- Received February 10, 1997.
- Revision received April 2, 1997.
- Accepted April 2, 1997.
- Copyright © 1997 by American Heart Association
Constantino A, Vinters HV. A pathologic correlate of the ‘steal’ phenomenon in a patient with cerebral arteriovenous malformations. Stroke. 1986;17:103-106.
Martin NA, Vinters HV. Arteriovenous malformations. In: Carter LP, Spetzler RF, eds. Neurovascular Surgery. New York, NY: McGraw-Hill Book Co; 1995:888-897.
Batjer HH, Devous MD, Seibert GB, Purdy PD, Ajmani AK, Delarosa M, Bonte FJ. Intracranial arteriovenous malformation: contralateral steal phenomena. Neurol Med Chir (Tokyo). 1989,29:401-406.
Ryvlin P, Mauguiere F, Sindou M, Froment JC, Cinotti L. Interictal cerebral metabolism and epilepsy in cavernous angiomas. Brain. 1995;118:677-687.
Latchaw RE, Hu X, Ugurbil K, Hall WA, Madison MT, Heros RC. Functional magnetic resonance imaging as a management tool for cerebral arteriovenous malformations. Neurosurgery. 1995;37:618-625.
Maldjian J, Atlas SW, Howard RS, Greenstein E, Alsop D, Detre JA, Listerud J, D’Esposito M, Flamm ES. Functional magnetic resonance imaging of regional brain activity in patients with intracranial arteriovenous malformations before surgical or endovascular therapy. J Neurosurg. 1996;84:477-483.
Puce A, Allison T, Asgari M, Gore JC, McCarthy G. Sensitivity of human visual cortex to faces, letterstrings, and textures: a functional magnetic resonance imaging study. J Neurosci. 1996;16:5205-5215.
McCarthy G, Puce A, Luby M, Belger A, Allison T. Magnetic resonance imaging studies of functional brain activation: analysis and interpretation. In: ‘Visualization of information processing in the human brain: recent advances in MEG and functional MRI.’ Electroencephalogr Clin Neurophysiol. 1996;47(suppl):15-31.
Farmer SF, Harrison LM, Ingram DA, Stephens JA. Plasticity of central motor pathway in children with hemiplegia cerebral palsy. Neurology. 1991;41:1505-1510.
Cao Y, Vikingstad EM, Huttenlocher PR, Towle VL, Levin DN. Functional magnetic resonance studies of the reorganization of the human hand sensorimotor area after unilateral brain injury in the perinatal period. Proc Natl Acad Sci U S A. 1994;91:9612-9616.