wavelength-dependent differences between optically ... · tion spectra. while blasdel and salama...

11
Wavelength-Dependent Differences between Optically Determined Functional Maps from Macaque Striate Cortex Niall P. Mc Loughlin and Gary G. Blasdel Department of Neurobiology, Harvard Medical School, Boston, Massachusetts 02115 Received September 23, 1997 This study investigates the role of wavelength in determining the source and dynamic range of activity- driven reflectance changes in macaque striate cortex. By using short (600 nm) and long (720 nm) wavelengths to map ocular dominance, orientation, and position from the same region of cortex on alternate trials, we isolated wavelength-dependent differences in the con- tributions of different tissue compartments. In agree- ment with previous reports, 600-nm illumination was found to produce optical signals that were more than twice the size of those obtained with 720-nm illumina- tion. In addition, 600- and 720-nm images were found to correlate everywhere except in regions occluded by blood vessels, where the images obtained at 600 nm correlated with the overlying vasculature. Since the 720-nm images do not correlate with the vasculature, this difference suggests that differential images ob- tained under 600-nm illumination are disproportion- ately sensitive to vascular events (e.g., changes in blood flow, volume, etc.). This finding is supported by the absorption spectra of hemoglobin and its deriva- tives, which absorb 600-nm light 4–1000 times more strongly than 720-nm light. Hence, for the 40% of cortex covered by blood vessels larger than 50 mm, images obtained at 600 nm are dominated by the vascular compartment to the exclusion of signals from the neural compartment below. r 1998 Academic Press INTRODUCTION Light-based monitoring of neuronal activity has a long and convoluted history. Early on, researchers discovered that action potentials are accompanied by transient changes in both light scattering and mem- brane birefringence (Cohen, 1973; Cohen et al., 1968; Cohen and Keynes, 1971; Hill and Keynes, 1949; Tasaki et al., 1968). The development of numerous membrane- impermeant voltage sensitive dyes, which transduce membrane potential into larger optical signals, made it practical to remotely monitor action potentials using light (Cohen and Salzberg, 1978; Grinvald et al., 1977; Gupta et al., 1981; Ross et al., 1977; Salzberg et al., 1977). However, it was well established prior to these developments that even larger reflectance changes could be observed at much slower time scales, presum- ably on account of activity-driven shifts in the distribu- tion of water and electrolytes. As early as 1933, Penfield noticed reflectance changes, perceptible to the unaided eye, during localized seizure activity (Penfield, 1933). Years later it was determined that much smaller changes in the activity of normal cortex could be detected optically on account of similar, activity- associated shifts in blood flow and volume (Lassen and Ingvar, 1961; Jo ¨bsis, 1977). A number of subsequent studies observed that neuronal activity is often accom- panied by oximetry signals that can be detected opti- cally by monitoring the absorption and/or fluorescence of intrinsic chromophores (Chance et al., 1962; Jo ¨bsis et al., 1977). Chance et al. (1962) noted that changes in scattering, occasioned by the large flux of potassium ions from active neurons into the interstitial fluids, affected tissue transparency as well. Though known, these changes were not used to explore cortical organization until Blasdel and Salama (1986) introduced the strategy of differentially imaging them to map functional organizations associated with ocular dominance and orientation in monkey striate cortex. After showing that sustained activation leads to sustained changes in local reflectance, they established the practice of differentially imaging these changes to visualize cortical function. Despite their use of a volt- age sensitive dye (NK2367), Blasdel and Salama fo- cused on the slow activity-associated intrinsic reflec- tance changes that were apparent before the cortex had been stained. One of the key contributions of this work was to shift attention from the fast optical changes associated with action potentials to the approach of differentially imaging the much slower ones associated with local sustained activity. By demonstrating a close relation with electrophysi- ologically determined patterns, moreover, this work established a neuronal basis for the patterns they observed that led to the first detailed maps of orienta- tion selectivity, the first images of any response prop- NEUROIMAGE 7, 326–336 (1998) ARTICLE NO. NI980329 326 1053-8119/98 $25.00 Copyright r 1998 by Academic Press All rights of reproduction in any form reserved.

Upload: dinhdan

Post on 12-Feb-2019

215 views

Category:

Documents


0 download

TRANSCRIPT

Wavelength-Dependent Differences between Optically DeterminedFunctional Maps from Macaque Striate Cortex

Niall P. Mc Loughlin and Gary G. BlasdelDepartment of Neurobiology, Harvard Medical School, Boston, Massachusetts 02115

Received September 23, 1997

This study investigates the role of wavelength indetermining the source and dynamic range of activity-driven reflectance changes in macaque striate cortex.By using short (600 nm) and long (720 nm) wavelengthsto map ocular dominance, orientation, and positionfrom the same region of cortex on alternate trials, weisolated wavelength-dependent differences in the con-tributions of different tissue compartments. In agree-ment with previous reports, 600-nm illumination wasfound to produce optical signals that were more thantwice the size of those obtained with 720-nm illumina-tion. In addition, 600- and 720-nm images were found tocorrelate everywhere except in regions occluded byblood vessels, where the images obtained at 600 nmcorrelated with the overlying vasculature. Since the720-nm images do not correlate with the vasculature,this difference suggests that differential images ob-tained under 600-nm illumination are disproportion-ately sensitive to vascular events (e.g., changes inblood flow, volume, etc.). This finding is supported bythe absorption spectra of hemoglobin and its deriva-tives, which absorb 600-nm light 4–1000 times morestrongly than 720-nm light. Hence, for the 40% of cortexcovered by blood vessels larger than 50 mm, imagesobtained at 600 nm are dominated by the vascularcompartment to the exclusion of signals from theneural compartment below. r 1998 Academic Press

INTRODUCTION

Light-based monitoring of neuronal activity has along and convoluted history. Early on, researchersdiscovered that action potentials are accompanied bytransient changes in both light scattering and mem-brane birefringence (Cohen, 1973; Cohen et al., 1968;Cohen and Keynes, 1971; Hill and Keynes, 1949; Tasakiet al., 1968). The development of numerous membrane-impermeant voltage sensitive dyes, which transducemembrane potential into larger optical signals, made itpractical to remotely monitor action potentials usinglight (Cohen and Salzberg, 1978; Grinvald et al., 1977;Gupta et al., 1981; Ross et al., 1977; Salzberg et al.,

1977). However, it was well established prior to thesedevelopments that even larger reflectance changescould be observed at much slower time scales, presum-ably on account of activity-driven shifts in the distribu-tion of water and electrolytes. As early as 1933, Penfieldnoticed reflectance changes, perceptible to the unaidedeye, during localized seizure activity (Penfield, 1933).Years later it was determined that much smallerchanges in the activity of normal cortex could bedetected optically on account of similar, activity-associated shifts in blood flow and volume (Lassen andIngvar, 1961; Jobsis, 1977). A number of subsequentstudies observed that neuronal activity is often accom-panied by oximetry signals that can be detected opti-cally by monitoring the absorption and/or fluorescenceof intrinsic chromophores (Chance et al., 1962; Jobsis etal., 1977). Chance et al. (1962) noted that changes inscattering, occasioned by the large flux of potassiumions from active neurons into the interstitial fluids,affected tissue transparency as well.

Though known, these changes were not used toexplore cortical organization until Blasdel and Salama(1986) introduced the strategy of differentially imagingthem to map functional organizations associated withocular dominance and orientation in monkey striatecortex. After showing that sustained activation leads tosustained changes in local reflectance, they establishedthe practice of differentially imaging these changes tovisualize cortical function. Despite their use of a volt-age sensitive dye (NK2367), Blasdel and Salama fo-cused on the slow activity-associated intrinsic reflec-tance changes that were apparent before the cortex hadbeen stained. One of the key contributions of this workwas to shift attention from the fast optical changesassociated with action potentials to the approach ofdifferentially imaging the much slower ones associatedwith local sustained activity.

By demonstrating a close relation with electrophysi-ologically determined patterns, moreover, this workestablished a neuronal basis for the patterns theyobserved that led to the first detailed maps of orienta-tion selectivity, the first images of any response prop-

NEUROIMAGE 7, 326–336 (1998)ARTICLE NO. NI980329

3261053-8119/98 $25.00Copyright r 1998 by Academic PressAll rights of reproduction in any form reserved.

erty in vivo, and the first images of different responseproperties from the same cortical tissue whose subse-quent analysis indicated local interactions that werepreviously unknown (Blasdel, 1992a,b; Bartfeld andGrinvald, 1992; Obermayer and Blasdel, 1993; Blasdeland Campbell, 1997).

Following on from this work, Grinvald et al. (1986)employed an array of photodiodes to explore the spec-tral and temporal dynamics of these intrinsic reflec-tance changes in unstained cortex. However, due to thelimited resolution of the 12 3 12 photodiode arrayemployed, it was not until Ts’o et al. (1990) imple-mented a variant of differential imaging that the firsthigh-resolution maps, comparable to those of Blasdeland Salama (1986), were generated exclusively fromintrinsic reflectance changes. Since then a great manyintrinsic imaging studies have been published (forexample, Bonhoeffer and Grinvald, 1991, 1993; Bosk-ing et al., 1997; Das and Gilbert, 1995, 1997; Kim andBonhoeffer, 1994; Wang et al., 1996).

Although the maps generated by the various differen-tial imaging approaches appear simple and intuitive,the mechanisms responsible for producing them de-pend on a number of poorly understood factors that linkneuronal activity to reflectance. Apart from physiologi-cal parameters, one of the most important variables isthe wavelength of incident illumination at which thereflectance changes are monitored, as different tissuecompartments are characterized by different absorp-tion spectra. While Blasdel and Salama (1986) did notexplore imaging at wavelengths shorter than 700 nm,they did verify their optical patterns with reference tohistochemical stains as well as to physiologically deter-mined preferences at strategic locations. The prefer-ences inferred from both optical and electrode mapsagreed to a remarkable degree.

The first systematic exploration of wavelength wasconducted by Frostig et al. (1990), who reported thatdifferential images obtained under a range of narrow-band wavelengths, 480 to 940 nm, appeared similar.However, as the focus of this study was on analyzingthe temporal dynamics of their responses, no detailedspatial comparisons between patterns obtained underdifferent wavelengths were made. Malonek and Grin-vald (1996) reexamined the sources of intrinsic opticalsignals in the cat using an imaging spectroscope, whichallowed them to simultaneously collect spatiotemporaldata for a range of wavelengths (520 to 650 nm). Theyestimated contributions from oxyhemoglobin (HbO2),deoxyhemoglobin (Hb), and light scattering by collaps-ing their data over space and fitting the spatial averageof their measured reflectance changes at differentwavelengths to the textbook absorption spectra. Unfor-tunately, to estimate the spatial precision of each signalcomponent the authors collapsed their data acrosswavelength (from 520 to 650 nm), so no wavelength-

dependent differences were noted. Similarly, by averag-ing their spectral data across the cortical surface, theyfailed to report on differential contributions from thedifferent tissue compartments.

In order to address the possibility of there beingdifferences between differential images collected underdifferent illumination wavelengths, we compared twosets of images acquired under virtually identical condi-tions, using two different wavelengths: the 720 nm usedinitially by Blasdel and Salama (1986) and a shorter600-nm wavelength favored by Grinvald et al. (1986)and others since (Ts’o et al., 1990; Bonhoeffer andGrinvald, 1991, 1993). In agreement with previousobservations, the 600-nm signals were more than twicethe size of the 720-nm signals at most locations (Frostiget al., 1990), and patterns obtained at 600 nm corre-lated with those obtained at 720 nm in regions thatwere not covered by blood vessels. However, the pat-terns obtained at 600 nm did not correlate with thoseobtained at 720 nm in regions that were covered byblood vessels. In these regions the patterns collected at600 nm correlated with the superficial vasculature.While the active interference of blood vessels can mostlikely be reduced or eliminated through appropriateadditional precautions, the passive loss of signals fromcortex covered by blood vessels appears to result fromthe far greater absorption of 600-nm light by hemoglo-bin and the resulting inability of light at this wave-length to penetrate blood vessels to monitor the activi-ties of cortical neurons below.

MATERIALS AND METHODS

These results are based on general observations fromsix animals (one cat, five macaque monkeys) anddetailed comparisons of patterns obtained from twoadditional macaque monkeys (Macaca mulatta) ex-pressly for this study. In order to curtail extraneouscontributions, all comparisons were restricted to im-ages collected using identical parameters (of visualstimulation and image acquisition) that were obtainedalternately at least two or three times at 600 and 720nm. Except for the fact that voltage-sensitive dyes werenot used, and longer durations of visual stimulationand frame averaging were therefore required (to com-pensate for smaller signals), all equipment, procedures,and parameters of visual stimulation remained thesame as in previous publications (see Blasdel, 1992a,b;Blasdel et al., 1995).

On the day of recording each animal was inducedeither with a single injection of ketamine/xylazine orwith Isoflurane delivered through a mask in a 2:1mixture of N2O and O2. After the animal had beenintubated, and a catheter for infusing electrolytes anddrugs had been placed in the cephalic vein, it wasventilated with a 2:1 mixture of N2O:O2 and anesthe-

327CORTICAL ACTIVITY AND REFLECTANCE

tized with thiopental (0.2–2.5 mg/kg/h for monkeys and0.1–1.0 mg/kg/h for the cat), as the initial anestheticagent (ketamine or Isoflurane) wore off.

Following the completion of surgical procedures,each animal was paralyzed partially with vecuroniumbromide (or Tubocurarine), to stabilize the eyes, whichwere then protected with hard, gas-permeable contactlenses. Curvatures for the latter were chosen to bringboth eyes into focus on the screen of a 20-in. monitorplaced 2 m away. Throughout the duration of eachexperiment, anesthesia was maintained with thiopen-tal (0.2–2.5 mg/kg/h for monkeys and 0.1–1.0 mg/kg/hfor the cat) and verified with respect to the electrocardio-gram, blood pressure, and end-tidal carbon dioxide(CO2), all of which were monitored continuously. Inaddition to these measures, the adequacy of anesthesiawas also verified at regular intervals from the absenceof reflexes (e.g., lateral canthal), which were assessedafter verifying that the level of neuromuscular suppres-sion, if present, remained below 50%. Although thislevel of suppression eliminates all but the smallest eyemovements (Blasdel, 1992a), it allows spinal reflexes tobe elicited when and if they appear (which, under theworst scenario, would occur long before the animalawakens).

Optical Imaging

During the acquisition of differential images, lightfrom a 100-W quartz halogen lamp was focused by aparabolic cold mirror through heat and interferencefilters (720 6 20 or 600 6 20 nm) onto the end of a0.25-in. diameter fiberoptic bundle. Monochromaticlight from this bundle was collimated and deflectedthrough the edge of a 100-mm objective and thenthrough a protective coverslip onto the cortical surface.Light that was not absorbed or scattered was thenreflected through the same objective to a projection lensthat focused it onto a Newvicon camera (COHU Model5300), as described previously (Blasdel, 1989, 1992a,b;Blasdel and Salama, 1986). Visual stimuli in all casesconsisted of four, superimposed nonharmonic square-wave gratings, which moved back and forth at 1.5–3°/s,and that appeared with an average luminance andcontrast of 3 cd/m2 and 80%. Between periods of visualstimulation, these patterns were replaced with a blankgray screen with the same average mean luminance (3cd/m2). All frames recorded by the camera were grabbedat 60 Hz, with a pixel resolution of 512 3 480, whichlater was compressed, by binning, to 128 3 120, witheach pixel representing approximately 17.5 3 17.5 µmof cortex.

Ocular Dominance

Differential images of ocular dominance were ob-tained by stimulating each eye alternately with grat-

ings at a single orientation, while images of the cortexwere averaged and compared. During each of fivesuccessive trials, 180–300 60-Hz frames were sum-mated during each of five presentations to the left eye(at intervals of 10–15 s) and subtracted from the sum ofa similar number of frames that were added whilestimuli were presented to the right eye. The final tallyof 900–1500 positive and 900–1500 negative frameswas then divided by 2n, with 1 # n # 3, to produce thefinal image that was reduced from 512 3 480 to 128 3120 by binning (Blasdel and Salama, 1986).

Orientation Preference

Differential images of orientation were obtained bycomparing responses to orthogonally oriented contours.Specifically, the frames accumulated during five inter-leaved presentations of gratings at one orientationwere subtracted from the sum of frames acquiredduring a comparable period of stimulation by an orthogo-nal grating. All other aspects of frame collection andprocessing were otherwise identical to those describedfor ocular dominance.

Retinotopic Position

Differential maps of retinotopically defined strips ofspace were obtained in a similar manner. However,instead of comparing frames accumulated during thestimulation of each eye, or with orthogonal orienta-tions, we compared frames averaged during the presen-tation of monocular stimuli in one of two complemen-tary sets of strips of visual space (see Campbell andBlasdel, 1995; Blasdel and Campbell, 1997, for details).

Correlational Analysis

In order to quantify and visualize the similaritybetween images, crosscorrelation and autocorrelationanalyses were performed. Crosscorrelograms were ob-tained by incrementally shifting one image relative toanother, in two dimensions, and calculating a correla-tion coefficient at each offset location. The resultingtwo-dimensional distribution of coefficients yields thecrosscorrelogram, which depicts the degree of similar-ity and distribution of regular features. When thisoperation is performed on two identical images, theresult is an autocorrelogram, which consists of a peakat location (0,0), where the image matches itself com-pletely, surrounded by a gradual decay whose shapeimparts information about the two-dimensional struc-ture of the image. The shape of a crosscorrelogram canbe used to identify similarities between two imagesthat vary between no correlation (a flat gray distribu-tion) and exactly the same structure (an autocorrelo-gram).

Since vascular artifacts may induce either positive(dark) or negative (light) values, in contrast to actual

328 MC LOUGHLIN AND BLASDEL

blood vessels, which always appear positive (dark), itwas necessary to half-wave rectify differential imagesbefore obtaining crosscorrelograms with the vascula-ture. This was achieved by inverting each pixel valuethat was greater than the image mean around theimage mean. Thus, for each optical map, any pixelvalue that exceeded the mean by x was set to the meanvalue minus x. This results in a half-wave rectifiedimage in which all vascular artifacts appear dark withrespect to the surrounding cortex. These rectified im-ages were used to calculate the crosscorrelogramsbetween the response maps and the vasculature. Theoriginal unprocessed maps were used to calculate the

autocorrelograms and the cross-wavelength correlo-grams.

RESULTS AND DISCUSSION

General Observations

Optical maps of the surface vasculature, ocular domi-nance, and orientation preference were obtained, aselaborated under Materials and Methods, with 600-and 720-nm illumination. Except for the wavelengthand intensity of illumination (which were adjusted tocompensate for differences in camera sensitivity), all

FIG. 1. Optical images of striate cortex obtained with 600- and 720-nm wavelength light. (A) An image of the surface vasculature seenunder 600-nm light. The V1/V2 border runs approximately parallel to the bottom of this image, with the foveal region being up and to the left.The surface vasculature is clearly visible, because oxyhemoglobin and deoxyhemoglobin absorb 600-nm light efficiently. (B) The exact sameregion of cortex imaged under 720-nm light. The largest blood vessels are barely detectable. (C) A differential image of ocular dominance for thesame region obtained under 600-nm light. Video images of the striate cortex obtained during visual stimulation of the left eye (four to sixpresentations for 1.5–3.0 s) were subtracted from images obtained during stimulation of the right eye to generate this image. Vascular noiseoccludes much of this optical map. (D) The exact same map of ocular dominance obtained under 720-nm light. Dark and light stripes,corresponding to left and right eye dominance bands, run approximately perpendicular to the V1/V2 border throughout the imaged region. (E)An example of a differential image of orientation preference obtained under 600-nm light. This differential image was produced by collectingvideo images of the striate cortex when both eyes were stimulated by moving vertical square-wave gratings and subtracting images obtainedwhen the eyes were stimulated by moving horizontal gratings. Darker spots correspond to regions which preferred vertical to horizontal, whilelighter spots represent the complement. As in C, vascular components occlude the underlying orientation map. (F) The same differential mapof orientation preference obtained under 720-nm light. No obvious vascular artifacts are present.

329CORTICAL ACTIVITY AND REFLECTANCE

parameters of visual stimulation, frame acquisition,and averaging were the same for images on the rightand left sides of Fig. 1. Figures 1A and 1B in the top rowdepict the surface vasculature under 600- and 720-nmillumination. Figures 1C and 1D display the differen-tial images of ocular dominance, and Figures 1E and 1Fdisplay the differential images of orientation that wereobtained at 600- and 720-nm wavelengths.

Most of our observations follow from the absorptionspectra of hemoglobin and its derivatives, for example,the millimolar extinction coefficients of Hb and HbO2are 3.2 and 0.8 for 600 nm and 0.38 and 0.09 for 720 nm(van Assendelft, 1970) and can be understood qualita-tively by comparing the images in Figs. 1A and 1B.From Lambert–Beer’s law we note that the transmit-tance, defined as T 5 102e · c · l (where e is the millimolarextinction coefficient, c is the concentration in millimo-lars per liter, and l is the total light pathlength), of600-nm light through vessels ranging in thickness from50 to 250 µm is 4 to 1000 times less than that of 720-nmlight. Hence superficial blood vessels appear muchdarker in the image on the left, where they contraststrongly with the cortex.

This situation is further complicated by changes inthe perfusion of blood vessels that can also producesignals. Blood vessels can appear as selective for aparticular response property as the most highly tunedcortical regions (Blasdel and Salama, 1986; Grinvald etal., 1986)—presumably because they supply or drainsuch regions. Unfortunately, these changes usuallyreflect the preferences of displaced neurons and notthose of neurons located underneath. As there is no wayof determining which regions they represent, thesesignals must be considered a form of noise in intrinsicoptical imaging. For these reasons, the differentialimages of ocular dominance and orientation in Figs. 1Cand 1E exhibit larger contributions from the vascula-ture than those in Figs. 1D and 1F, even though allimages were collected from the same cortical region, inresponse to the same stimuli, and at virtually the sametime.

Distribution of Signal Strength

In order to quantify the differences between imagesobtained at different wavelengths, three independent

FIG. 2. Vascular and avascular compartments. (A) The surface vasculature imaged at high resolution under 570-nm illumination. Thisimage was processed to produce a sequence of gray-level compartments through which pixel variances were measured. (B) All regions of thevascular map whose gray scale value is less than the mean pixel value. This image corresponds roughly to the vascular regions and contains allthe larger blood vessels. (C) All regions whose pixel values are greater than or equal to the mean pixel value. This image corresponds roughly tothe avascular regions. Only small, lighter colored, vessels remain within this image. It is possible to make a sequence of such compartments byextracting image regions whose gray scale values lie within a limited range of pixel values. This can then be used to determine the effects of thesurface vascularity on the strength of the measured optical signals.

330 MC LOUGHLIN AND BLASDEL

measures were developed, the simplest of which is todivide the imaged region into vascular and avascularcompartments. This can be done most easily by apply-ing a 50% gray scale threshold to a high-resolutionimage of the vasculature. When this is done (see Fig. 2),approximately 40 and 60% of the imaged area fall intovascular and avascular compartments. By measuringthe pixel variance of ocular dominance patterns withineach compartment, it is possible to dissect the signalcomponents. When this is done for the ocular domi-nance pattern imaged at 600 nm, for example, thevariance of pixels in the vascular compartment exceedsthat in the avascular compartment by 65%. Taking thisanalysis one step further, one can easily subdivide thecortex in Fig. 2 into a series of compartments. Thelowest thresholds extract regions corresponding to thecenters of the largest and thickest blood vessels be-cause these reflect the least light. The next compart-ment, corresponding to pixels with values lying be-tween 18.0 and 21.5% of the maximum pixel value,extracts regions corresponding to the edges of largeblood vessels along with the centers of smaller ones.This progresses to the point at which the highestthreshold compartment, corresponding to the brightestpixels, represents unoccluded regions of the corticalsurface. Since 600 nm lies near an isobestic point in theabsorption of hemoglobin, moreover, arterioles andvenules are treated similarly by this procedure. Whilesimilar to the previous procedure, this technique dis-sects the cortex into 16 different compartments corre-sponding (approximately) to 16 different thicknesses ofblood through which the optical signals must pass.Figure 3 shows the distribution of pixel variance fromthe ocular dominance and orientation images in Fig. 1,with respect to each cortical compartment.

As one can see in Fig. 3A, for images collected at 600nm the pixel variance is greatest in regions correspond-ing to the centers of the largest blood vessels. On theother hand, the pixel variance for images collected at720 nm remains relatively flat across compartments, asone might expect from previous analysis (Blasdel,1992a). Figure 3B displays similar distributions for thedifferential images of orientation collected at eachwavelength in Figs. 1E and 1F.

Ocular dominance images obtained in an identicalmanner from a second animal appear in Fig. 4. As inFig. 3, the surface vasculature interferes with theimage of the underlying ocular dominance columnswhen 600-nm illumination is employed (Fig. 4B).

Figure 5 depicts the representation of vertical stripesimaged at 600 and 720 nm that take the form of lightand dark bands running parallel to the V2 border,which is aligned with the bottom of each frame. Sincethe topography of V1 is well established, there can beno doubt that these images represent the verticallydirected strips that were stimulated (Blasdel and Camp-

bell, 1997; Campbell and Blasdel, 1995; McLoughlinand Blasdel, 1996).

As in the previous examples, the images obtained at600 nm (see Fig. 5A) appear noisier with respect to thevasculature. The variance measured in each of the 16gray-level vascular compartments produces a similarresult. The shallower slope of the line corresponding to600 nm suggests that vascular artifacts are smaller inthis image than those in the ocular dominance andorientation images displayed previously. This couldpossibly be because these patterns occur at a larger

FIG. 3. The spatial distribution of intrinsic optical signals. (A)The computed variance of the pixel values (signal) for the oculardominance maps, Figs. 1C and 1D, as processed through a sequenceof vascular compartments as described in Fig. 2. The open squarescorrespond to the variances computed for Fig. 1C, while the opencircles correspond to the same measures for Fig. 1D. The horizontalaxis codes the range of gray scale values in each compartment, whichin this case ranges from 18 to 100%. The low % corresponds to thedarker pixels in Fig. 2A, while the higher ranges correspond to thebrightest pixels. There is a clear correlation between the range ofgray scale values used (vascularity) and the signal strength for theimages collected under 600-nm illumination. The recorded signalstrength is greatest from the regions corresponding to the biggestdarkest blood vessels and falls off in the more avascular regions. Incontrast, there is no significant difference between the recordedsignal strength coming from the vascular and avascular regions ofthe 720-nm optical maps. (B) The same computed metrics for theorientation preference maps depicted in Figs. 1E and 1F.

331CORTICAL ACTIVITY AND REFLECTANCE

scale that more closely approximates the scale at whichthe activity of blood vessels actually reflects the activityof the local neuropil.

Spatial Correlations

In order to identify the common and vascular compo-nents of the optical maps, crosscorrelograms werecalculated for the images collected at different wave-lengths and the vasculature. The results for oculardominance and orientation appear in Figs. 6A and 6B.The autocorrelogram for the vascular image in Fig. 6Aexhibits a typical pattern of decay for an image whichcontains little or no repeated spatial structure. Thecentral peak falls off rapidly in all directions. Theautocorrelograms for the 600- and 720-nm maps ofocular dominance display a very different pattern ofactivity. Activity falls off much more rapidly in thehorizontal direction than in the vertical direction. Thisindicates that the images contain a spatial structurethat is elongated in the vertical direction. The resur-gence of activity after the image is shifted even furtherin the horizontal direction indicates that the verticalpattern is repeated after a characteristic shift. Hence,

the autocorrelograms for the optical maps of oculardominance succinctly capture the fact that ocular domi-nance columns are elongated along the vertical axisand are periodic. Since the ocular dominance bands areirregular, the only exact match (correlation 5 1.0) oc-curs at the center, such that the intensity of theflanking bands fades over a distance of severalmillimeters.

From the autocorrelogram of the 600-nm oculardominance image, one can see that regular banding,though present, is greatly suppressed relative to that inthe 720-nm image, suggesting the presence of extrane-ous noise. However, the presence of a central peak inthe crosscorrelograms of the 600/720-nm ocular domi-nance images indicates that the optical maps correlatestrongly with each other, almost as strongly as the720-nm image correlates with itself. From their cross-correlations with the vasculature one can also see thatthe differences between the ocular dominance patternsobtained at these wavelengths derive from the fact thatthe 600-nm pattern, in contrast to the 720-nm pattern,also correlates with the image of the vasculature.

As all the correlograms are bounded by 1 and 0, their

FIG. 4. Examples of ocular dominance maps from a second animal. (A) The surface vasculature of a region of striate cortex taken from asecond animal, this time closer to the horizontal meridian. (B) A differential image of ocular dominance for the same region obtained under600-nm light. Video images of the striate cortex obtained during visual stimulation of the left eye (four to six presentations for 1.5–3.0 s) weresubtracted from images obtained during stimulation of the right eye to generate this image. Similar to our previous examples, vascularcontributions are obvious in much of this optical map. (C) The exact same map of ocular dominance obtained under 720-nm light. Dark andlight stripes, corresponding to left and right eye dominance bands, run approximately parallel to the V1/V2 border throughout the imagedregion, although they are less regular than those seen in Fig. 1.

332 MC LOUGHLIN AND BLASDEL

similarity can be estimated by subtracting the averagedifference between them from 1. If the remainder is 1.0,the correlograms are identical, since their averagedifference is 0.0. A remainder of 0.0 indicates that theyare exact opposites or anticorrelated with an averagedifference of 1.0, while a value of 0.5 indicates theabsence of any correlation at all. When this metric isapplied between the vascular/600-nm ocular domi-nance crosscorrelogram and the vascular/vascular auto-correlogram, the remainder is 0.586. This indicatesthat 17.2% of the 600-nm ocular dominance patternderives from vascular elements. A similar calculation

applied to the 720-nm ocular dominance pattern pro-duces a value of 0.507 and suggests that only 1.4% ofthe map’s spatial structure is derived from vascularelements.

The autocorrelograms for the 600- and 720-nm orien-tation maps depicted in Fig. 6B exhibit a very differentpattern of activity. The central peak decays equally inall directions and is surrounded some distance out by aring of activity. The slight elliptical shape of this ringreflects the local orthogonality of ocular dominanceboundaries and iso-orientation contours, resulting in aslightly exaggerated distance of repetition perpendicu-

FIG. 5. Intrinsic maps of retinotopy. (A) A differential image of a series of elongated strips in space. Video images of the animal’s cortexwere collected while it was visually stimulated in strips of space. Images collected while the animal was visually stimulated in complementarystrips were then subtracted to produce this retinotopic map. Elongated dark and bright regions, corresponding to two sets of complementaryvertical strips, run approximately parallel to the V1/V2 border, which is aligned with the bottom of each frame. Once again this differentialimage, obtained under 600-nm light, has obvious vascular components. (B) The optical map for the same stimulus collected under 720-nmlight. The position of the light and dark bands are shifted relative to those in A, indicating that the animal made an eye-movement betweenthese presentations. (C) The variance of the pixel values measured through the same gray scale compartments as in Figs. 3A and 3B, this timecomputed for the images shown in A and B.

333CORTICAL ACTIVITY AND REFLECTANCE

FIG. 6. Spatial correlations. (A) A table of all possible spatial correlograms between the two optically derived maps of ocular dominanceand the image of the overlying vasculature. The autocorrelograms run diagonally top–left to bottom–right while the crosscorrelograms appearoff-diagonal. Complete images were shifted up to 25 pixels in each direction for both the autocorrelograms and the crosscorrelograms. Edgeeffects were avoided by scaling the difference metric by the number of contributing points. An arbitrary scale was used to plot each correlogramto aid visualization of salient structures within the data. As the vasculature noise can appear in the optical maps as either lighter or darkerthan the background, the ocular dominance maps were rectified around their mean for comparison to the vascular map (Materials andMethods). (B) A table of all possible spatial correlograms between the two optically derived maps of orientation preference and the image of theoverlying vasculature. As in A, the autocorrelograms run down diagonally from the top–left while the crosscorrelograms appear off-diagonal.

334

lar to the trajectories of the ocular dominance columns,shown previously by Obermayer and Blasdel (1993).

Once again the autocorrelogram of the pattern im-aged at 720 nm exhibits much greater regularity thanthat imaged at 600 nm, even though the correspondingcrosscorrelograms show that the images correlatestrongly with one another. Crosscorrelations with thevasculature additionally show that the 600-nm orienta-tion pattern, in contrast to 720-nm pattern, correlateswith the vasculature. Applying the similarity estimateto the correlograms, one finds that vascular elementscontribute 25.1% of the spatial structure to the 600-nmimage, while they contribute only 5.2% to the 720-nmpattern.

In contrast to previous studies that concentrated onthe spectrotemporal aspects of intrinsic reflectancechanges in cortex (Frostig et al., 1990; Malonek andGrinvald, 1996), this study reports the first high-resolution spatial comparison between optical mapsobtained at different wavelengths and the overlyingvasculature through which these maps are recorded. Inagreement with the present findings, Frostig et al.(1990) reported that blood vessel artifacts were signifi-cantly larger for shorter wavelengths of light. Simi-larly, Malonek and Grinvald (1996) noted significantinvolvement of the superficial vasculature in all imagescollected at wavelengths shorter then 620 nm, prompt-ing them to exclude these regions from further analysis(Fig. 2, p. 552).

Mayhew et al. (1996) reported a pronounced 0.1-Hzoscillation in the reflected light from cortex. Thisoscillation correlated strongly with laser Doppler flowmeasurements of cerebral blood flow (CBF). Our imag-ing procedure minimizes any such contribution, be-cause we synchronize our data collection to the animal’srespiration and heart beat, subtract data collected atalternating time slices, and limit our collection times toshort periods during visual stimulation. In addition,thiobarbiturate anesthesia has recently been shown todramatically reduce CBF (Lindauer et al., 1993). Thissuggests that the effects we report are, if anything,reduced by our choice of anesthesia.

The current findings suggest that the changes inreflectance measured with 720-nm wavelength lightare not due to the selective responses of surface bloodvessels. While the mechanisms responsible for produc-ing such changes are not well understood at present,they are most likely derived from a number of differentevents, including glial cell swelling due to potassiumrelease (MacVicar and Hochman, 1991), dilation of thecapillary beds (Pawlik et al., 1981), and cell swellingdue to increased chlorine transport (Holthoff and Witte,1996). Malonek and Grinvald (1996) also noted, bytheir signal decomposition procedure, that light scatter-ing signals tend to dominate (.70%) optical responsesmeasured with wavelengths greater than 660 nm. In

short, our analyses indicate (1) that 600- and 720-nmoptical maps are strongly correlated in tissue compart-ments not underlying the surface vasculature, (2) thatwhile 720-nm optical maps exhibit a flat distribution ofsignal strength across the various tissue compart-ments, the regions associated with the surface vascula-ture exhibit much stronger signals in all the 600-nmimages examined, and (3) that the signals recordedfrom these vascular compartments are spatially corre-lated with the surface vessels when 600-nm illumina-tion is employed.

ACKNOWLEDGMENTS

This work was supported in part by a McDonnell-Pew trainingfellowship and funded by the NIH (EY06586 and EY10862) and theHuman Frontiers Foundation. We thank Lawrence Sincich, JonathanLevitt, and Wade Regehr, for useful comments on the manuscript,and Darlene Campbell for her excellent technical assistance.

REFERENCES

Bartfeld, E., and Grinvald,A. 1992. Relationship between orientation-preference pinwheels, cytochrome oxidase blobs, and ocular-dominance columns in primate striate cortex. Proc. Natl. Acad. Sci.USA 89:11905–11909.

Blasdel, G. G. 1989. Visualization of neuronal activity in monkeystriate cortex. Annu. Rev. Physiol. 51:561–581.

Blasdel, G. G. 1992a. Differential imaging of ocular dominance andorientation selectivity in monkey striate cortex. J. Neurosci. 12:3115–3138.

Blasdel, G. G. 1992b. Orientation selectivity, preference, and continu-ity in monkey striate cortex. J. Neurosci. 12:3139–3161.

Blasdel, G. G., and Campbell, D. 1997. Symmetry of cortical maps inold and new world primates. Submitted for publication.

Blasdel, G. G., Obermayer, K., and Kiorpes, L. 1995. Organization ofocular dominance and orientation columns in the striate cortex ofneonatal macaque monkeys. Visual Neurosci. 12:589–603.

Blasdel, G. G., and Salama, G. 1986. Voltage-sensitive dyes reveal amodular organization in the monkey striate cortex. Nature 321:579–585.

Bonhoeffer, T., and Grinvald, A. 1991. Iso-orientation domains in catvisual cortex are arranged in pinwheel-like patterns. Nature353:429–431.

Bonhoeffer, T., and Grinvald, A. 1993. The layout of iso-orientationdomains in area 18 of cat visual cortex: Optical imaging reveals apinwheel-like organization. J. Neurosci. 13:4157–4180.

Bosking, W. H., Zhang, Y., Schofield, B., and Fitzpatrick, D. 1997.Orientation selectivity and the arrangement of horizontal connec-tions in tree shrew striate cortex. J. Neurosci. 15:2112–2127.

Campbell, D., and Blasdel, G. G. 1995. Optical measurements ofcortical magnification factors in new and old world primates. Soc.Neurosci. Abstr. 21:771.

Chance, B., Cohen, P., Jobsis, F., and Schoener, B. 1962. Intracellularoxidation–reduction states in vivo. Science 137:499–508.

Cohen, L. B. 1973. Changes in neuron structure during actionpotential propagation and synaptic transmission. Phys. Rev. 53:373–418.

Cohen, L. B., and Keynes, R. D. 1971. Changes in light scatteringassociated with action potential in crab nerves. J. Physiol. 212:259–275.

335CORTICAL ACTIVITY AND REFLECTANCE

Cohen, L. B., Keynes, R. D., and Hille, B. 1968. Light scattering andbirefringence changes during action potential propagation andsynaptic transmission. Physiol. Rev. 53:373–418.

Cohen, L. B., and Salzberg, B. M. 1978. Optical measurement ofmembrane potential. Rev. Physiol. Biochem. Pharmacol. 83:35–88.

Das, A., and Gilbert, C. D. 1995. Long-range horizontal connectionsand their role in cortical reorganization revealed by optical record-ing of cat primary visual cortex. Science 375:780–784.

Das, A., and Gilbert, C. D. 1997. Distortions of visuotopic map matchorientation singularities in primary visual cortex. Nature 387:594–598.

Frostig, R., Lieke, E. E., Ts’o, D. Y., and Grinvald, A. 1990. Corticalfunctional architecture and local coupling between neuronal activ-ity and the microcirculation revealed by in vivo high-resolutionoptical imaging of intrinsic signals. Proc. Natl. Acad. Sci. USA87:6082–6086.

Grinvald, A., Lieke, E., Frostig, R. D., Gilbert, C. D., and Wiesel, T. N.1986. Functional architecture of cortex revealed by optical imagingof intrinsic signals. Nature 324:361–364.

Grinvald, A., Slazberg, B. M., and Cohen, L. B. 1977. Simultaneousrecording from several neurones in an invertebrate central nervoussystem. Nature 268:140–142.

Gupta, R. K., Slazberg, B. M., Grinvald, A., Cohen, L. B., Kamino, K.,Lesher, S., Boyle, M. B., Waggoner, A. S., and Wang, C. H. 1981.Improvements in optical methods for measuring rapid changes inmembrane potential. J. Membr. Biol. 58:123–137.

Hill, D. K., and Keynes, R. D. 1949. Opacity changes in stimulatednerve. J. Physiol. 108:278–281.

Holthoff, K., and Witte, O. W. 1996. Intrinsic optical signals in ratneocortical slices measured with near-infrared dark-field micros-copy reveal changes in extracellular space. J. Neurosci. 16:2740–2749.

Jobsis, F. F. 1977. Noninvasive, infrared monitoring of cerebral andmyocardial oxygen sufficiency and circulatory parameters. Science198:1264–1267.

Jobsis, F. F., Keizer, J. H., LaManna, J. C., and Rosenthal, M. J. 1977.Reflectance spectrophotometry of cytochrome aa3 in vivo. Appl.Physiol. Respir. Environ. Exerc. Physiol. 43:858–872.

Kim, D., and Bonhoeffer, T. 1994. Reverse occlusion leads to a preciserestoration of orientation preference maps in visual cortex. Nature370:370–372.

Lassen, N. A., and Ingvar, D. H. 1961. The blood flow of the cerebralcortex determined by radioactive krypton. Experientia 17:42–43.

Lindauer, U., Villringer, A., and Dirnagl, U. 1993. Characterization of

CBF response to somatosensory stimulation: Model and influenceof anesthetics. Am. J. Physiol. 264:H1223–H1228.

MacVicar, B. A., and Hochman, D. 1991. Imaging of synapticallyevoked intrinsic optical signals in hippocampal slices. J. Neurosci.11:1458–1469.

Malonek, D., and Grinvald, A. 1996. Interactions between electricalactivity and cortical microcirculation revealed by imaging spectros-copy: Implications for functional brain mapping. Science 272:551–554.

Mayhew, J. E. W., Askew, S., Zheng, Y., Porrill, J., Westby, G. W. M.,Redgrave, P., Rector, D. M., and Harper, R. M. 1997. Cerebralvasomotion: A 0.1-Hz oscillation in reflected light imaging of neuralactivity. NeuroImage 4:183–193.

McLoughlin, N. P., and Blasdel, G. G. 1996. Differential opticalimages of visual space in monkey striate cortex. Invest. Ophthamol.Visual Sci. (Suppl.) 37:939.

Obermayer, K., and Blasdel, G. G. 1993. Geometry of orientation andocular dominance columns in monkey striate cortex. J. Neurosci.13:4114–4129.

Pawlik, G., Rackl, A., and Bing, R. J. 1981. Quantitative capillarytopography and blood flow in the cerebral cortex of cats: An in vivomicroscopic study. Brain Res. 208:35–58.

Penfield, W. 1933. The evidence for a cerebral vascular mechanism inepilepsy. Ann. Intern. Med. 7:303–310.

Ross, W. N., Salzberg, B. M., Cohen, L. B., Grinvald, A., Davila, H. V.,Waggoner, A. S., and Wang, C. H. 1977. Changes in absorption,dichroism, and birefringence in stained giant axons: Optical mea-surement of membrane potential. J. Membr. Biol. 33:141–183.

Salzberg, B. M., Grinvald, A., Cohen, L. B., Davila, H. V., and Ross,W. N. 1977. Optical recording of neuronal activity in an inverte-brate central nervous system: Simultaneous monitoring of severalneurons. J. Neurophysiol. 40:1281–1291.

Tasaki, I., Watanabe, A., Sandlin, R., and Carnay, L. 1968. Changesin fluorescence, turbidity and birefringence associated with nerveexcitation. Proc. Natl. Acad. Sci. USA 61:883–888.

Ts’o, D. Y., Frostig, R. D., Lieke, E. E., and Grinvald, A. 1990.Functional organization of primate visual cortex revealed by highresolution optical imaging. Science 249:417–420.

van Assendelft, O. W. 1970. Spectrophotometry of HaemoglobinDerivatives. Royal VanGorcum, Assen, The Netherlands.

Wang, G., Tanaka, K., and Tanifuji, M. 1996. Optical imaging offunctional organization in the monkey inferotemporal cortex. Sci-ence, 272:1665–1668.

336 MC LOUGHLIN AND BLASDEL