Journal of Neuro-Ophthalmology, December 1998, Volume 18, Issue 4

Variability in Visual Cortex Activation During Prolonged Functional Magnetic Resonance Imaging

This study was conducted to test whether cortical activation varies across successive epoques during functional magnetic resonance imaging (fMRI) studies. Ten normal adult volunteers were studied with a 1.5-T MR scanner. Pseudocoronal study planes were chosen perpendicular to the tentorium cerebelli, at two thirds the distance from the posterior edge of the splenium of the corpus callosum to the transverse sinuses. Functional images were acquired with a T2*-weighted spoiled gradient echo sequence. The visual cortex was stimulated by goggles flashing at 8 Hz. Each study consisted of 82 sequential scans, lasting 15 seconds each for a total of 20.5 minutes. Two scans without stimulation were alternated with two scans of visual stimulation. Scans 3 through 83 were divided into five sequences of 16 scans. For each sequence, the number of pixels within a predefined rectangular region of interest that showed increased activity during stimulation were counted. Least squares regression models of straight lines were fit to the data. The initial level of visual cortex activation in the region of interest, as measured by the y-intercept, varied substantially from subject to subject (range: 4-68, p < 0.001). There was sufficient evidence of systematic change with time to reject the hypothesis of constant activation with the same stimulus over time (p=0.02). The observed visual cortex activation with single-plane fMRI varied both with time over successive epoques and among subjects. Possible factors responsible for the variation may include head movement, eyelid position, attention, and physiologic fatigue. These factors must be accounted for in experimental design and in data analysis and interpretation.

Journal of Neuro- Ophtlwlmology 18( 4): 258- 262, 1998. © 1998 Lippincott Williams & Wilkins, Philadelphia Variability in Visual Cortex Activation During Prolonged Functional Magnetic Resonance Imaging Grant T. Liu, M. D., Douglas W. Fletcher, B. S., Rachel J. Bishop, B. S., Maureen G. Maguire, Ph. D., Graham E. Quinn, M. D., Phyllis Hendy, R. T. M. R., Robert A. Zimmerman, M. D., and John C. Haselgrove, Ph. D. This study was conducted to test whether cortical activation varies across successive epoques during functional magnetic resonance imaging ( fMRI) studies. Ten normal adult volunteers were studied with a 1.5- T MR scanner. Pseudocoronal study planes were chosen perpendicular to the tentorium cerebelli, at two thirds the distance from the posterior edge of the splenium of the corpus callosum to the transverse sinuses. Functional images were acquired with a T2*- weighted spoiled gradient echo sequence. The visual cortex was stimulated by goggles flashing at 8 Hz. Each study consisted of 82 sequential scans, lasting 15 seconds each for a total of 20.5 minutes. Two scans without stimulation were alternated with two scans of visual stimulation. Scans 3 through 83 were divided into five se­quences of 16 scans. For each sequence, the number of pixels within a predefined rectangular region of interest that showed increased activity during stimulation were counted. Least squares regression models of straight lines were fit to the data. The initial level of visual cortex activation in the region of interest, as measured by the y- intercept, varied substantially from subject to subject ( range: 4- 68, p < 0.001). There was sufficient evidence of systematic change with time to reject the hypothesis of constant activation with the same stimulus over time ( p = 0.02). The observed visual cortex activation with single- plane fMRI varied both with time over successive epoques and among subjects. Possible factors responsible for the variation may include head movement, eyelid position, at­tention, and physiologic fatigue. These factors must be ac­counted for in experimental design and in data analysis and interpretation. Manuscript received February 10, 1998; accepted June 18, 1998. From the Neuro- ophthalmology Service ( G. T. L.), Radiology Depart­ment ( R. J. B., D. W. F., P. H., R. A. Z., J. C. H.), and Division of Ophthal­mology ( G. E. Q.) of The Children's Hospital of Philadelphia, and the Science Eye Institute ( G. T. L., M. G. M., G. E. Q.); and the Departments of Neurology ( G. T. L.), Ophthalmology ( G. T. L., M. G. M, G. E. Q.), and Radiology ( R. A. Z.), University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania, U. S. A. Supported, in part, by The Children's Hospital of Philadelphia Ethel Foerderer Fund for Excellence ( G. T. L., R. A. Z., J. C. H.), the University of Pennsylvania Research Foundation ( G. T. L.), and Grant R21 EY10964 ( M. G. M.) from the National Eye Institute, Bethesda, Mary­land, U. S. A. Address correspondence to Grant T. Liu, M. D., Division of Neuro-ophthalmology, Department of Neurology, Hospital of the Univer­sity of Pennsylvania, 3400 Spruce Street, Philadelphia, PA 19104, U. S. A. Key Words: Functional magnetic resonance imaging- Visual cortex. Functional magnetic resonance imaging ( fMRI) is ri­valing positron emission tomography ( PET) and single photon emission computed tomography ( SPECT) in the ability to provide functional neuroanatomic information. Additionally, fMRI is nonionizing and noninvasive, al­lowing repeated testing without exposing patients to ra­dioactive substances ( 1). fMRI has already proved to be a highly valuable research and clinical tool. Several in­vestigators have used it to study striate ( 2- 6), extrastriate ( 7), motor ( 8) and auditory cortices ( 9), and language lateralization ( 10) in humans. Gadolinium ( 1 1) was used in one of the first success­ful fMRI studies of visual cortex, but subsequent inves­tigators have used blood oxygen level- dependent con­trast ( 12) to identify areas of cortical activity in gradient-echo MR images. Neuronal activation is associated with an increase in local blood flow ( 13) and volume with little or no change in oxygen consumption ( 14). The resultant increase in venous oxygenation causes an in­crease in MR signal ( 14). During fMRI, blood oxygen level- dependent studies, two sets of magnetic resonance images are acquired, one set during a functional activity (" on"), and the other with the patient resting (" off"). The sets are compared to identify any changes suggestive of an increase in local blood flow associated with brain activation ( 15). Ordinarily, on and off sets are alternated during a single fMRI study, which usually lasts just a few min­utes. In more complex studies ( cognitive studies, e. g.), several functional activities may be performed over a more extended period. One group of investigators ( 16) demonstrated the " feasibility and value of conducting multiple functional paradigms in single sessions." Be­cause of the important methodologic implications, sev­eral investigators have studied visual cortex activation during prolonged sustained stimulation, and the results have been inconsistent ( 17- 19). We studied the stability 258 FUNCTIONAL MRl OF VISUAL CORTEX 259 of visual cortex activation in a different manner, over successive epoques, because ordinarily, on and off sets are alternated during single fMRI studies. METHODS Ten normal adult volunteers were studied ( age range, 22- 35 years). Two were women, and eight were men. Subjects were placed supine in a 1.5- T MR scanner ( Si­emens Magnetom Vision; Erlangen, Germany). Closely fitting foam pads were placed around their heads to dis­courage movement. Task Visual stimulation was obtained by using light- proof binocular goggles with a 5 x 5 array of red light- emitting diodes flashing at 8 Hz ( model S10VSB: Grass Instru­ments; Quincy, MA). Subjects performed the passive task of lying in the MR with goggles over the eyes. They were instructed to keep the eyelids open during the stimulation periods, but eyelid position was not moni­tored. The subjects were told to stay awake throughout the session. To ensure compliance, they were given the active task of squeezing a rubber ball each time the lights were turned on. The ball was connected by a rubber tube to a bell in the monitor's room. In this way, it was con­firmed by the monitor that all subjects remained awake throughout the study. Imaging Method Several research groups have used a plane through the calcarine fissure to study visual cortex activation ( 11). We found this to be an unreliable and inconsistent method for several reasons: First, the calcarine fissure is rarely flat and commonly bends at least once ventrally or dorsally. Second, frequently, there is interhemispheric asymmetry in the position of the left and right fissures. Third, there is wide intersubject difference in the first and second variables. The combination of these factors made imaging the visual cortex in one plane a difficult task, and the amount of visual cortex included in the plane varied from person to person. Instead, in an attempt to obtain anatomically consis­tent views across volunteers, each was studied with pseudocoronal planes. Tl- weighted parasagittal scout images ( recovery time [ TRj, 300 msec; echo time [ TE], 15 msec; alpha, 90°; matrix, 256 x 256; slice thickness = 5 mm) were obtained for anatomic localization. An­other Tl- weighted image was obtained in a pseudocoro­nal plane perpendicular to the tentorium cerebelli, two thirds the distance from the posterior edge of the sple-nium of the corpus callosum to the transverse sinuses and was labeled the pseudocoronal anatomic image ( Fig. I A). This method insured that the plane was posterior to the parieto- occipital sulcus. For the activation studies, an oblique 5- mm thick axial slice was aligned in the same position as the pseudocoro­nal anatomic image. The functional images were ac­quired with a T2*- weighted spoiled- gradient echo se­quence ( TR, 100 msec; TE, 60 msec; alpha, 30°; matrix, 128 x 128; and field of view [ FOV], 22- 24 cm). Shim­ming of the scanner occurred before the functional stud­ies for each subject. Each study consisted of 82 sequential scans, lasting 15 seconds each, for a total of 20.5 minutes. The sequence alternated two scans ( 30 seconds) of no activation ( rest­ing baseline) with two scans of activation ( flashing lights in both eyes simultaneously), for 20 periods each of ac­tive and resting states. To study the possibility that our results could be caused by instrumental factors, the ex­periment was also performed with a phantom. Data Analysis All images were transferred to a commercial system ( Sun SPARC station 1; Sun Microsystems; Mountain View, CA). The first two images were discarded to in­sure that a steady state had been achieved. For each subject, five activation maps were calculated using suc­cessive sets of 16 images. Visual inspection of the im­ages displayed rapidly was performed. In addition, the centroid of each image was calculated and plotted versus time to monitor movement in the x- and y- planes. The area of each image was calculated and plotted versus time to gauge through- plane movement. No motion cor­rection was performed on the data set. An image processing routine written in IDL ( Interac­tive Data Language; Research Systems; Boulder, CO) created a statistical parametric map. To calculate the sta­tistical parametric map, the baseline images and the ac­tivated images of the time course of the functional im­ages were divided into two groups and subjected to Stu­dent's /- test on a pixel- by- pixel basis. The threshold value was chosen by calculating the 1% significance level for all 80 images of each patient using the process­ing strategy described by Requardt ( 20). Within the pseudocoronal anatomic Tl images, a rect­angular region of interest ( ROI) containing calcarine cor­tex in both hemispheres was selected for each subject ( Fig. IB). The ROIs were determined anatomically with­out reference to activation. Control ROIs of identical size were selected from the cerebellum. For each statistical parametric map, the number of pixels with /- values above the threshold of p = 0.01 were counted within the ROI ( Fig. 1C). Least squares regression models of straight lines were fit to the data to evaluate the relation­ship between the level of activated pixels in the ROI and the time since the beginning of the testing session. The regression models that were evaluated included i) a separate intercept and a separate slope for each subject, ii) a separate intercept for each subject and a common slope for all sub­jects, and iii) a common intercept and common slope. F-tests comparing the residual variance from alternative mod­els were used to test whether additional parameters indi­vidual slopes or intercepts) should be included ( 21). RESULTS Review of plots of the level of activated pixels versus time showed substantial variation among volunteers ( Fig. 2). Comparison of the three statistical models of the data showed two important features of the response pattern of the 10 participants: First, the overall level of activation, ./ Neum- Ophllwhiml, Vol. 18, No. 4, 1998 260 G. T. LIU ET AL. characterized by the ^- intercept, varied significantly among participants ( p < O. OOl), ranging from 4 to 68 pixels in the ROI with a mean of 34. Second, a unique slope for each person did not provide a significantly improved fit of the data compared with the fit of a com­mon slope for all participants ( p = 0.29). In some vol­unteers, the slope was positive, and in others it was nega­tive. In all 10 volunteers, there was little or no activation in the cerebellum. These comparative data suggest that the blood oxygen level- dependent signal changes in the visual cortex truly represented stimulus- related cortical activation and were not artifactual. In the phantom experiment, no consistent grouping of the ( approximately 1%) of pixels above the threshold occurred. This confirmed that our data genuinely re­flected a change in activation of visual cortex and that there was no machine variability that could account for it. Slight head motion was detected in all volunteers. The average range of motion of the centroids in the x direc­tion was 0.83 mm ( range, 0.53- 1.12 mm), whereas the average range of motion of the centroids in the y-direction was 2.1 mm ( range, 1.03- 2.84 mm). The aver­age change in image area, reflecting through- plane movement, was 4.2% ( range, 1.5- 8.8%). DISCUSSION In our single- plane study, visual cortex activation var­ied across successive epoques when volunteers stimu­lated with flash goggles were studied using fast low-angle shot ( FLASH) fMRI. Results of the studies of pre­vious groups have been contradictory. Hathout et al. ( 17) performed uninterrupted visual stimulation by flash goggles for 18 to 24 minutes and found a decrease in signal intensity in visual cortex 2 to 5 minutes after the initial peak in activation. In another study by Kranda et al., ( 18) when subjects viewed flashing vertical gratings for 4 minutes, there was an immediate increase in signal measured in visual cortex, followed by an exponential decline for the remaining stimulation period. In contra­distinction, Howseman et al. ( 19) found stable activation by checkerboard stimuli using echo- planar and FLASH techniques, and they suggested that discrepancies in re­sults may be caused by the use of different stimuli and not by the imaging method. Although the use of flash goggles, rather than check­erboard stimuli, may have accounted for a change in < FIG. 1. A: Sagittal T1 - weighted image from one of the volunteers. A line was drawn through the tentorium cerebelii ( 1). At two thirds the distance from the posterior edge of the splenium of the corpus callosum ( 2) to the transverse sinuses ( 3), a pseudocoronal plane was selected ( 4). B: Pseudocoronal image, T1 - weighted corre­sponding to plane described in ( A). A rectangular region of inter­est ( ROI) is drawn around visual cortex. C: Activation of visual cortex within the pseudocoronal plane, blood oxygen level-dependent technique, displayed over T1- anatomic image with ROI ( B). In this case, the statistical parametric map was calcu­lated using the first 16 images. Activated pixels represent those with a /- test value above the threshold of p = 0.01. ./ Ncuio- Ophihulmol. Vol. IN, No. 4. 1998 FUNCTIONAL MRI OF VISUAL CORTEX 261 FIG. 2. Data for all 10 volunteers. V- axis: number of pixels activated within the pseudocoronal plane in a rectangular region of interest containing visual cortex; X- axis: time ( minutes). Each diamond represents the data point for each activation map, of which there were five for each volunteer during the study period. activation, there are several other explanations that must be considered, including head movements, habituation of cortical neurons, varied attention to the stimulus, and eyelid position. Head movements may have accounted for the signal variation in our study. Although the cine loop can detect large head movements within the plane of the image, small movements may be missed. Our analysis of the centroid position and image area over time suggested that some slight head movements occurred. Motion cor­rection algorithms currently exist, but these are more applicable to three- dimensional data sets. Retinal photoreceptors ( 22) and neurons in the lateral geniculate nucleus ( 23) do not fatigue during continuous light stimulation. However, in animals, intracellular re­cordings have demonstrated that striate cortical neurons can adapt during prolonged viewing of high- contrast si­nusoidal gratings ( 24). In visual evoked potential studies in humans performed over several minutes or hours, the PI00 latency and amplitude can remain relatively stable ( 25), but in some instances a decrease in the amplitude can be observed with passing time ( 26). Alternatively, changes in level of activation may be explained by variable attention to the visual stimulus. Hypervigilance may have accounted for the increase in activation seen in three of our volunteers, whereas de­creased attention during the task in the other volunteers may have resulted in a decrease in activation. Partici­pants in other studies ( 19) tested with a checkerboard pattern, perhaps a more interesting stimulus, may have had stable activation with passing time because they were more attentive during the prolonged study. Changes in eyelid position may also have accounted for some of the signal variation. Although subjects were instructed to keep their eyes open during the visual stimulation periods, we had no way of monitoring this. In addition, the interpersonal variability in the amount of visual cortex activation was striking. It is possible that variable activation is a real phenomenon, but anatomic variation in striate cortex is likely a contributory factor. Our pseudocoronal planes produced more consistent im­ages through visual cortex than axial planes would have allowed. However, we still could not guarantee that ex­actly the same portion of striate cortex was imaged from subject to subject. Ross et al. ( 27) demonstrated differ­ences in visual cortex activation in younger versus older J Neuro- Ophllmlmol, Vol. IS, No. 4, I99S 262 G. T. LIU ET AL. subjects. However, our volunteers were all in a younger age group. Variability in visual cortex activation, for which there are several possible reasons, must be accounted for in fMRI experimental design and in data analysis and in­terpretation. Quantitative intersubject comparisons and those of visual tasks performed at the beginning and end of long study sessions may not be valid. Every attempt must be made to reduce head movements. Acquiring the data in three- dimensions is more desirable to allow the application of motion correction programs. Interleaving the visual tasks may be desirable to account for possible fatigue during a lengthy study session. Future investiga­tion should include comparison of flash and checker­board stimuli during prolonged stimulation periods, and whether variation in visual attention may alter the level of cortical activation. Finally, a method for insuring that the eyelids remain open during the stimulation periods should be applied. REFERENCES 1. Turner R. Magnetic resonance imaging of brain function [ edito­rial |. Ann Neurol 1994; 35: 637- 8. 2. Kwong KK, Belliveau JW, Chester DA, et al. dynamic magnetic resonance imaging of human brain activity during primary sensory stimulation. Proc Natl Acad Sci USA 1992; 89: 5675- 9. 3. Schneider W, Noll DC, Cohen JD. Functional topographic map­ping of the cortical ribbon in human vision with conventional MRI scanners. Nature 1993; 356: 150- 3. 4. Hedera P, Lai S, Haacke EM, et al. Abnormal connectivity of the visual pathway in human albinos demonstrated by susceptibility-sensitized MR1. Neurology 1994; 44: 1921- 6. 5. Miki A, Nakajima T, Fujita M, et al. Functional magnetic reso­nance imaging in homonymous hemianopia. Am .1 Ophthalmol 1996; 121: 258- 66. 6. Miki A, Nakajima T, Takagi M, et al. Detection of visual dysfunc­tion in optic atrophy by functional magnetic resonance imaging during monocular visual stimulation. Am J Ophthalmol 1996; 122: 404- 15. 7. Tootcll RB, Reppas JB, Kwong KK, et al. Functional analysis of human MT and related visual cortical areas using magnetic reso­nance imaging. J Neurosci 1995; 15: 3215- 30. 8. Rao SM, Binder JR, Bandettini PA, et al. Functional magnetic resonance imaging of complex human movements. Neurology I993; 43: 23l 1- 8. 9. Binder JR, Rao SM, Hammcke TA, et al. Functional magnetic resonance imaging of human auditory cortex. Ann Neurol 1994; 35: 662- 72. 10. Binder JR, Rao SM, Hammeke TA, et al. Lateralized human brain language systems demonstrated by task subtraction functional magnetic resonance imaging. Arch Neurol 1995; 52: 593- 01. 11. Belliveau JW, Kennedy DN, McKinstry RC, et al. Functional map­ping of the human visual cortex by magnetic resonance imaging. Science 1991; 254: 716- 9. 12. Ogawa S, Lee TM, Kay AR, Tank DW. Brain magnetic resonance imaging with contrast dependent on blood oxygenation. Proc Natl Acad Sci USA 190; 87: 9869- 72. 13. Fox PT, Raichle ME. Stimulus rate determines regional brain blood flow in striate cortex. Ann Neurol 1985; 17: 303- 5. 14. Cohen MS, Bookheimer SY. Localization of brain function using magnetic resonance imaging. Trends Neurosci 1994; 17: 268- 77. 15. DeYoe EA, Bandettini P, Neitz J, et al. Functional magnetic reso­nance imaging ( FMRI) of the human brain. ./ Neurosci Methods 1994; 54: 171- 87. 16. Savoy RL, Tootell RBH, O' Craven KM, Reppas JR. Cortical lo­calization of fMRI response to stereo disparity [ abstract]. Hum Brain Mapping 1995; l( Suppl 1): 57. 17. Hathout GM, Kirlew KA, So GJ, et al. MR imaging signal re­sponse to sustained stimulation in human visual cortex. J Magn Reson Imaging 1994; 4: 537- 43. 18. Kranda K, Schilling A, Villringer K, Blankenburg F. Neural ad­aptation in the human visual cortex to patterned stimulation [ ab­stract]. Hum Brain Mapping 1995: l( Suppl 1): 52. 19. Howseman AM, Josephs O, Porter D, et al. Sustained activation in visual cortex: a comparison between echo- planar and FLASH fMRI methods ( abstract). ISMRM 1996; 4: 1855. 20. Requardt M. Evaluation strategies for MRI of human brain acti­vation: individual analysis of correlational imaging and cluster detection [ abstract]. Proc Soc Magn Reson 1994; 2: 625. 21. Draper NR, Smith H. Applied regression analysis. 2nd ed. New York: John Wiley & Sons, 1981. 22. Gouras P, MacKay CJ. Growth in amplitude of the human cone electroretinogram with light adaptation. Invest Ophthalmol Vis Sci 1989; 30: 625- 30. 23. Sclar G, Lennie P, DePriest DD. Contrast adaptation in striate cortex of macaque. Vision Res 1989; 29: 747- 55. 24. Allison JD, Casagrandc Va, Debruyn E. I, Bonds AB. Contrast ad­aptation in striate cortical neurons of the nocturnal primate bush baby ( Galago crassicaudatus). Vis Neurosci 1993; 10: 1129- 39. 25. Chiappa KH. Reproducibility- reliability. In: Chiappa KH, ed. Evoked potentials in clinical medicine. New York: Raven Press, 1990: 86- 90. 26. Sappey- Marinier D, Calabrese G, Fein G, et al. Effect of photic stimulation on human visual cortex lactate and phosphates using IH and 3 IP magnetic resonance spectroscopy. J Cereh Blood Flow Metab 1992; 12: 584- 92. 27. Ross MH, Yurgelun- Todd DA, Renshaw PF, et al. Age- related reduction in functional MRI response to photic stimulation. Neu­rology 1997; 48: 173- 6. J Neiiro- Oplilliahiml. Vol. IS, No. 4. 1998