Journal of Neuroscience Methods 91 (1999) 135 – 145
www.elsevier.com/locate/jneumeth
A focusing image probe for assessing neural activity in vivo
David M. Rector a, Robert F. Rogers b, John S. George a,*
a
Biophysics Group, P-21, MS-D454, Los Alamos National Laboratory, PO Box 1663, Los Alamos, NM 87545, USA
b
Central R&D, E.I. Du Pont De Nemours & Co., Inc., Wilmington, DE 19880 -0328, USA
Received 8 October 1998; received in revised form 7 June 1999; accepted 13 June 1999
Abstract
We describe a compact, focusing image probe to record rapid optical changes from neural tissue. A gradient index (GRIN) lens
served as a relay lens from tissue to a microscope objective which projected an image onto a CCD camera. The microscope
objective and camera assembly was adjusted independently from the GRIN lens, allowing focus changes without disturbing the
probe/tissue interface; firm contact minimized movement and specular reflectance. Fiber optics around the probe perimeter
provided diffuse illumination from a 780 nm laser, or 660 and 560 nm light emitting diodes. To characterize depth-of-field, we
imaged a black suture through increasing tissue thicknesses. Light modulation by the suture remained detectable down to 900 mm
using 780 nm illumination. We acquired images from cardiorespiratory areas of the rat dorsal medulla, at different depths and
illumination wavelengths. Images illuminated at 560 nm were dominated by vasculature flow patterns, while 660 nm illumination
revealed different spatial patterns which preceded vascular flow by 40 ms and may represent cardiac-related neural activity. Using
780 nm light, image sequences triggered by the cardiac R-wave showed vascular perfusion changes with delayed and broader
responses at deeper levels. Electrical stimulation within the vagal bundle caused fast optical changes which track the electrical
response, with a different spatial distribution from hemodynamic signals. © 1999 Elsevier Science B.V. All rights reserved.
Keywords: Blood flow; Depth viewing; Optical imaging; Light scattering
1. Introduction
Reflected light images from neural structures disclose
several correlates of neural activity (Cohen, 1973; Grinvald et al., 1988; Rector et al., 1993, 1997a). Depending
on the illumination wavelength used, indications of
hemodynamic changes (e.g. 560 nm), metabolic protein
conformational changes (B 400 nm), or neuronal
swelling (\ 600 nm) can serve as measures of local
neural activity. Light reflectance techniques typically
illuminate the brain surface with monochromatic light,
then detect back-scattered light from the tissue using a
charged-coupled device (CCD) camera. Changes in the
amount of back-scattered light are calculated as differences or ratios across time on a pixel-by-pixel basis,
thereby forming a parametric map of light scattering or
absorption changes from the tissue of interest. Such
procedures for recording from large neural populations
* Corresponding author. Tel.: +1-505-665-2550; fax: + 1-505-6654507.
E-mail address: jsg@lanl.gov (J.S. George)
provide insight into neural interactions and neural network properties. The need to study many neurons
simultaneously has driven significant advances in optical measurements of neural activation.
Most imaging studies using intrinsic optical signals
have employed slow scan imaging technologies and
steady-state stimulation to visualize hemodynamic
changes associated with neural activation. Although
some components of the hemodynamic signals are comparatively fast (B 1 s), long integrated signal acquisition is used to average over fluctuations associated with
the cardiac cycle. Subtraction techniques isolate those
signals associated specifically with neural activity, and
tend to eliminate vessel artifacts. Attempts to image
dynamic processes require more sophisticated characterization of optical changes associated with the cardiac
cycle. In addition, a number of in vitro studies have
identified fast light scattering changes associated with
neural swelling during activation that parallel electrical
events (Salzberg et al., 1985; Tasaki and Byrne, 1992).
In order to image such changes in vivo, it is necessary
to achieve high sensitivity, high time resolution, and to
0165-0270/99/$ - see front matter © 1999 Elsevier Science B.V. All rights reserved.
PII: S 0 1 6 5 - 0 2 7 0 ( 9 9 ) 0 0 0 8 8 - 6
136
D.M. Rector et al. / Journal of Neuroscience Methods 91 (1999) 135–145
adequately account for the dynamics of other ongoing
physiological processes that affect the overall optical
signal these movement and spectral changes associated
with the cardiac cycle.
Contact of the image probe with the tissue surface
can serve as a mechanism for minimizing movement
artifacts associated with the cardiac and respiratory
cycles. Such contact stabilizes the tissue relative to the
probe surface, and reduces pulsations typically seen in
these preparations. For chronic studies, the bone is
subsequently sealed around the probe using bone wax
and dental acrylic allowing restoration of CSF pressure
and further stabilizing the brain tissue through hydrostatic forces. Another mechanism for minimizing movement artifact involves an opening in the skull which is
then sealed with a glass window and filled with oil.
Such methods minimize brain movement through the
establishment of CSF pressure related hydrostatic
forces; however, movements in neighbouring brain regions (or a nearby vessel) may introduce movements
through hydraulic effects. Additionally, use of an oil
filled chamber precludes dark field techniques, and
specular reflectance becomes a major issue in the scattered light signals.
Previous in vivo optical techniques using coherent
fiber optic image conduit have a focal plane limited to
the tissue surface in contact with the probe (Rector and
Harper, 1991; Rector et al., 1991, 1993, 1997b). Such
procedures generally form good images of the tissue
surface. Also, since illumination surrounds the imaged
area, dark-field methods eliminated specular reflectance
from the tissue surface, and provide scattering information from deeper tissue (Rector et al., 1997b). Unfortunately, light from deeper structures is out of focus and
the image is blurred. Because cells of interest are frequently located several hundred microns below the
surface, it is desirable to focus below the surface to
accurately measure deep structures. Optical techniques
with deep focus capabilities would be especially useful
for brain structures which are otherwise difficult to
access.
We describe an imaging system with the ability to
detect light scattering events at least 900 mm below the
tissue surface. The device can record from a large
neural area and acquire long temporal image sequences.
Tissue stabilization is achieved through contact of the
tissue surface with a gradient index (GRIN) lens. The
design allows changing focal depth without disturbing
the tissue/probe interface. The device is compact in size,
allowing stereotaxic mounting and a minimal skull
opening. The device has the potential for miniaturization to be chronically implanted in freely behaving
animals. Related prototype devices we developed can
also generate confocal and spectral images.
We tested the depth-of-field of the image probe by
imaging a 200 mm black suture through successive
tissue depths, and measured light modulation by the
suture as a function of depth. We characterized scattered light changes within the exposed dorsal medulla
of acute rat preparations using three wavelengths to
visualize dynamic patterns correlated with perfusion
during the cardiac cycle. In addition to characteristic
patterns of perfusion, we observed light scattering
changes consistent with neurophysiological activation.
In order to image optical signatures of neural activation, nerves within the vagal bundle were stimulated to
produce electrical and optical evoked potentials in the
tissue under the probe.
2. Materials and methods
A diagram of the probe and camera is illustrated in
Fig. 1. A gradient index lens (GRIN, Fig. 1a) (Gradient
Index, Rochester, NY) is affixed to a lens holder (Fig.
1b) with positioning screws to center the gradient index
lens relative to an objective lens (Fig. 1c). The objective
lens moves within a telescoping tube (Fig. 1d) and are
Fig. 1. Three-dimensional rendering of the endoscope shows the
major components and internal views. The existing device is constructed from telescoping aluminum tubing. Backscattered light from
the tissue enters the gradient index lens (a) mounted in a centering
C-mount adapter (b). The image from the tissue is formed in front of
a microscope objective (c) mounted in a housing which can be moved
through telescoping tubing (d) to adjust the depth of focus through
the gradient index lens. A linear translation stage (e) supports the
microscope objective and CCD assembly and is positioned by an
oil-filled hydraulic piston (f). The microscope objective projects onto
a CCD camera (g) mounted in a 0.5-inch brass tube (h), nine wires
lead to a connector (i) for camera control and signal output to a
12-bit digitizer. Illumination is provided by flexible plastic fiber optics
(j) mounted around the perimeter of the gradient index lens as
illustrated in the enlarged view in the upper left hand corner. Illumination fibers were split into two bundles for separate and alternating
illumination with 560 and 660 nm light. The telescoping tubing for
the microscope objective and brass holder for the camera have been
cut away to show internal parts.
D.M. Rector et al. / Journal of Neuroscience Methods 91 (1999) 135–145
137
mounted on a linear translation stage (Fig. 1e) driven
by a hydraulic piston (Fig. 1f) for focusing the image at
different depths beyond the surface of the GRIN lens.
A miniature CCD video camera (Fig. 1g) (TC211,
Texas Instruments, Dallas, TX) is mounted in a brass
tube (Fig. 1h) behind the objective lens, also with
centering screws. Details for camera construction are
given elsewhere (Rector et al., 1997b). Camera output
leads are connected via a nine-pin connector (Fig. 1i),
to a computer-based digitizer which simultaneously collected physiological signals (1 kHz each) and produced
an interleaved file format. Physiological channels included the electrocardiogram (ECG), tracheal pressure,
blood pressure, and end tidal CO2. Flexible optic fibers
(Fig. 1j) affixed around the perimeter of the gradient
index lens provide illumination from several different
light sources including a 780 nm laser diode array, and
660 and 560 nm light-emitting diodes (LED), halogen,
flash and arc lamps.
2.1. Gradient index lens
The GRIN lens consisted of a 100 mm long glass
cylinder, 3 mm in diameter (also available in other
lengths and diameters), formed with a radially varying
index of refraction (Moore, 1993) (Fig. 2A). The index
of refraction is a continuous parabolic gradient which is
higher in the center and lower at the edge. The result of
this index variation is that a ray incident on the surface
follows a sinusoidal path through the rod every half
period (P/2), a distance which is determined by the
gradient properties during manufacture (Fig. 2B) which
is determined by the gradient properties during manufacture Fig. 2B. The cylinder has very similar properties
to a relay lens system, which conjugate image planes
within the cylinder. In this application, the GRIN lens
effectively extends the image plane of detection optics
through the length of the cylinder. This property allows
the GRIN lens to remain in a fixed position while
moving the detection optics to change the plane of
focus (Fig. 2C). GRIN lenses are manufactured in
integral P/2 lengths (P/2, P, 3P/2, 2P, etc.), thus light
entering the GRIN lens at a particular angle exits the
cylinder at the same angle for whole period lengths (P,
2P), resulting a non-inverted image, or at opposite
angles for half period lengths (P/2, 3P/2) resulting an
inverted image.
2.2. Depth-of-field measurement
To test the depth-of-field of the device, we placed a
200 mm black suture across a homogeneous sample of
rat cortical tissue, and imaged the suture through various tissue thicknesses. As many as thirty tissue slices,
increasing in thickness by 60 mm increments, were
placed between the GRIN lens and the suture. At each
Fig. 2. The gradient index (GRIN) lens is a cylinderical glass rod with
a radially varying index of refraction (A). Light entering the lens
bends and forms an image within the cylinder every half period (P/2),
a property determined by the manufacturing process (B). A block
diagram of the major components for the image probe (C) shows the
GRIN lens and the camera assembly with the microscope objective
and CCD camera. Moving the camera assembly closer to the GRIN
lens moves the effective focal plane of the microscope objective
beyond the GRIN lens without disturbing the probe/tissue interface.
tissue thickness, we focused the objective lens and
camera assembly onto the suture center and collected
an average image of 100 frames. We do not expect
cortex slices prepared in this way to have exactly the
same optical properties as brain tissue in vivo, primarily
due to the lack of hemoglobin which would absorb a
significant portion of the shorter wavelengths. Thus,
these data may overestimate the depth to which scattered signals can be detected.
The suture image in the first average was not obscured by tissue, and appeared the sharpest. We assessed blurring of the suture image through increasing
tissue thickness by plotting the image intensity profile
as a function of depth, and characterized the family of
intensity profiles by convolving the first intensity profile
(no obscuring tissue), with a Gaussian curve to model
the consequence of optical blurring by tissue. Using
IDL (Interactive Data Language, IDL, Research Systems, Boulder, CO), we fit the model to profiles at each
depth by optimizing the amplitude and width parameters of the Gaussian, and plotted each parameter as a
function of tissue thickness. These data were also compared to data collected with an earlier imaging device
138
D.M. Rector et al. / Journal of Neuroscience Methods 91 (1999) 135–145
which utilized coherent image conduit, and did not
have focusing ability (Rector et al., 1997b). To characterize how the shadow of the suture affected the image
at various depths, the suture was kept at a constant 360
mm depth while changing the objective/camera focus.
2.3. Cardiac cycle experiments
To explore the application of the system for functional neural imaging, we undertook a series of experiments to image physiologically correlated optical
changes within the rat dorsal medulla. Rats were anesthetized with urethane (1.3 g/kg i.p.), and one femoral
vein and artery were cannulated with 10 cm long
catheters. Rats were paralyzed with veceronium bromide in some experiments, artificially ventilated, and a
pneumothorax was performed to minimize ventilationrelated movement. Tracheal pressure, arterial blood
pressure, and the electrocardiogram (ECG) were written onto polygraph paper and digitized during the
experiment. Dorsal brainstem exposure was achieved by
flexing the head ventrally 60°, and removing the
dural sheet between the skull and the first vertebrae.
The imaging device was placed above the cardiorespiratory areas of the nucleus tractus solitarious (NTS),
and illuminated with 780 nm light from a laser diode
array, a high intensity 660 nm LED, or a high intensity
560 nm LED. Back-scattered light was imaged with the
CCD camera, digitized at 100 frames per second
(f.p.s.), along with the physiological record. Although
acquisition rates of 1000 fps or higher are possible with
the current system, we chose an optimal rate to maximize light capture and minimize the size of the data set.
Image sequences from the recording were triggered
from the cardiac R-wave and average sequences divided
by the first frame in the sequence to normalize varying
illumination intensity across the image. Images were
pseudocolored such that cool colors (blue to purple)
represent an increase in reflectance, warm colors (yellow to red) represent a decrease in reflectance, and
green represents no significant change from the first
frame of the sequence (a =0.05). Cross-correlations
were also calculated for each pixel against the blood
pressure signal. Resulting images are displayed with the
average ECG traces. Image sequences were generated
during alternating 660 and 560 nm illumination (50
f.p.s. per color), or single wavelength 780 nm illumination (100 f.p.s.).
2.4. Stimulation experiments
To assess scattered light changes associated with
stimulation of tissue under the image probe, we dissected the vagal bundle within the neck of the animal,
and placed stimulating hooks on nerves that project to
the regions under the probe. Single-shock, 40 mA stim-
uli were delivered to the nerves (0.2 ms in duration, 1–2
s random intervals). Images were acquired continuously
at 100 Hz with a field potential made from a macrowire
placed under the image probe. Average image sequences consisted of 200–400 trials triggered by a 100
ms pre-trigger input. To illustrate changes across time,
the average of ten baseline images (100 ms) were averaged, and divided into the remaining images in the
sequence. Images were pseudocolored and displayed in
the same manner as cardiac-triggered data.
3. Results
3.1. Depth of field measurements
Fig. 3 illustrates average suture images, and intensity
profiles across tissue thicknesses for 660 nm (Fig. 3A)
and 780 nm (Fig. 3B) illumination. Intensity profiles
(Fig. 3C) show a change in shape as the suture image
becomes blurred through increasing tissue thickness.
Fig. 4 illustrates two parameters of the model used to
characterize the suture blurring. The amplitude and
width parameters estimated from fits to the Gaussian
used to model the blurring, are plotted as a function of
tissue depth. The suture becomes undetectable as determined by the amplitude parameter B0.1 and steepness
of the width parameter curve, after 720 mm using 660
nm illumination and after 900 mm using 780 nm illumination. Fig. 4 also compares data from an earlier
version of the probe using image conduit (Rector et al.,
1997b) analyzed in the same manner. Using 660 nm
reflectance mode illumination and a macro lens, the
suture is obscured after 300 mm tissue thickness. Surround illumination (darkfield) and image conduit increased detectability to 600 mm.
Fig. 5 reveals the extent of the shadow which the
black wire produces at a fixed tissue depth while changing the focal plane through 1620 mm. The width
parameter of the Gaussian fit shows a minimal width
(sharpest edge) at 360 mm focus depth, corresponding
to the actual suture depth (Fig. 5C).
3.2. Cardiac- and stimulus-triggered e6ents
Time-triggered averages of 560 nm illuminated images with the cardiac cycle show patterns that resemble
hemodynamic changes, along with other components.
Fig. 6A illustrates a single depth measurement across
the cardiac cycle using alternating 560 and 660 nm
wavelengths, and shows a dramatic difference in image
pattern. A sketch of the anatomy under the probe
shows several structures, including that NTS and hypoglossal nucleus, which control aspects of the cardiac
rhythm.
D.M. Rector et al. / Journal of Neuroscience Methods 91 (1999) 135–145
Specific regions within the image were plotted to
reveal three basic relationships between 560 and 660 nm
scattering changes. Fig. 7A shows a region over a vessel
where 560 and 660 nm changes are out of phase, and
660 nm changes precede the 560 nm changes by 40 ms.
Fig. 7B illustrates a region were the two colors are 180°
out of phase, and Fig. 7C, D show positions where they
are almost in phase. Differing rise times are also apparent for the 560 nm image regions between areas immediately over a vessel, and those areas away from vessels.
Correlations between blood pressure and image intensity on a pixel by pixel basis reveal specific regions
which are correlated with various phases of the blood
pressure signal (Fig. 6C). In particular, 560 nm correlation images show well-defined vessels with zero lag, and
139
660 nm correlation images show a patch of activity
which is spatially distinct from 560 nm images, and has
a − 40 ms lag. Fig. 8 shows correlation plots from the
same regions defined in Fig. 7.
Image sequences using 780 nm illumination through
several tissue depths (Fig. 9A) show vessel perfusion.
This wavelength represents a point of maximum difference between the absorption spectra of oxy- and deoxyhemoglobin, and vessel outlines are apparent in the
surface images. Images acquired at deeper levels are
more diffuse, and have a longer time to peak intensity
(Fig. 9A, B). A blood pressure correlation image at zero
lag (Fig. 9C) shows the vessel patterns seen in the image
sequences. The same data set included stimulation of
the aortic nerve which is illustrated as an image se-
Fig. 3. A black suture, 200 mm in diameter was placed on a block of rat cortical tissue. Images were gathered though successive tissue thicknesses
placed over the suture. For each depth measurement, the focus of the microscope lens was adjusted to bring the plane of focus onto the suture.
Panel A shows a series of images using 660 nm illumination, and images in panel B were acquired with 780 nm illumination. For each image, an
average line representing the suture intensity profile was plotted as is demonstrated in panel C for the 780 nm series. Vertical bars represent the
SEM for each data point.
140
D.M. Rector et al. / Journal of Neuroscience Methods 91 (1999) 135–145
In comparison to earlier studies using 660 nm light
and image conduit (Rector et al., 1997b), the GRIN
lens increased detection of an object within brain tissue
from 600 mm to 720 mm. The 780 nm wavelength
increased that depth to 900 mm since neural tissue is
more transparent to longer wavelengths. Dark-field illumination improved our ability to image below the
surface through decreased specular reflectance
(Holthoff et al., 1994; Holthoff and Witte, 1996; Rector
et al., 1997b); however, our ability to obtain a sharp
image is limited by a high degree of tissue scattering.
Interference from shadows of shallow objects in Fig. 5
shows that an object at 360 mm can still be clearly seen
current device with higher numerical aperture optics as
at focus depths well beyond 1700 mm. For sharper
focusing, we are developing adaptations of the current
device with higher numerical aperture optics as well as
capabilities for confocal and multi-photon microscopy.
Fig. 4. Intensity profiles (representing the black suture cross-section
at each tissue thicknesses) were modeled by convolving the first
profile with a Gaussian curve, fitting amplitude and width parameters. The top panel plots Gaussian amplitude as a function of tissue
thickness for four illumination types. The lower panel plots the
Gaussian width. The ‘2’ data were collected from tissue imaged with
surround illumination (dark-field) using 780 nm light through the
gradient index (GRIN) lens. The ‘’ data were collected using 660
nm light. The ‘’ data were collected with surround illumination
using 660 nm light and coherent image fiber optics (F.O.), and the
‘× ’ data were collected from tissue illuminated directly (bright-field)
using 660 nm light and a macro lens (M.L.). Vertical lines represent
the SEM.
quence (Fig. 9D) and a plot of average intensity change
across time (Fig. 9E). Stimulus sequences show rapid
optical changes that are spatially distinct from vascular
patterns, and track the electrical evoked response, as
measured by a macrowire placed under the probe.
4. Discussion
These results demonstrate the use of GRIN technology to achieve focusing capability below tissue surfaces
in contact with the probe. A black suture remained
detectable through layers of brain tissue down to 900
mm, although the image was blurred due to increased
light scattering of deeper tissue. The chief advantage of
the GRIN lens is an ability to shift the relative focal
plane, without moving the probe and disturbing the
tissue–probe interface.
Fig. 5. Five representative images of a 200 mm diameter black suture
placed at a fixed depth of 360 mm with increasing focus settings show
the extent of the shadow cast by the suture (A). Intensity profiles (B)
illustrate blurring evidenced by decreasing slope at the suture edge.
An analysis of the width parameter in the Gaussian fitting routine (C)
shows the width parameter is lowest (sharpest) around the actual
depth (360 mm).
D.M. Rector et al. / Journal of Neuroscience Methods 91 (1999) 135–145
Fig. 6. Panel A contains images collected with alternating 560 nm (G) and 660 nm (R) light. The 560 nm images show the perfusion of two vessels, and the 660 nm images exhibit a different
temporal and spatial pattern. Panel B illustrates the anatomical features of the tissue under the image probe (GRIN). Major structures include the nucleus of the solitary tract (NTS), dorsal motor
nucleus (DMV), hypoglossal nucleus (XII), area postrema (AP), and the tract of the solitary nucleus (TS). Panel C is a plot of the blood pressure and EKG waveforms collected and averaged
along with the images in panel A. The last panel (D) shows images representing the correlations of blood pressure with 560 nm images (BPvsG) and with 660 nm images (BPvsR) on a pixel by
pixel basis. Lag times from − 180 to +180 ms are represented. The color scale indicates the magnitude for each colour image. For correlations, green represents no correlation, white represents
a correlation of 1.0, and black represents a correlation of − 1.0. For the time averaged images, the colour scale represents relative scattered light intensity from the first frame in the sequence.
Green represents no change, white represents a − 0.2% change (less back scattered light or increased absorbance), and black represents + 0.2% change (more back scattered light or less
absorbance).
141
142
D.M. Rector et al. / Journal of Neuroscience Methods 91 (1999) 135–145
Fig. 7. Representative regions within the images of Fig. 6, Panel B
show different relationships between the 560 nm and 660 nm illuminated images. Regions of interest are illustrated at the bottom of the
figure. For region A, representing signals obtained immediately over
a vessel, 560 nm images are out of phase with the 660 nm images, and
have a low amplitude. Region B is located over the brightest 660 nm
region. Sixty hundred and sixty nm signals have a large amplitude
and are out of phase with 560 nm signals. Region C is also immediately over a vessel, and though the 560 nm images behave similarly to
region A, 660 nm images are closer in phase in this region. Region D
also shows a close phase relationship between 560 and 660 nm
images; however, there is no evidence of a vessel in this region.
Regions A and C, which are immediately over a vessel, also demonstrate a rapid rise time in the 560 nm images, where as regions B and
D show slow rise times, possibly corresponding with slow perfusion
of the capillary bed. Abscissa for each plot represents absolute
percent change from initial conditions.
Cardiac-triggered sequences show a number of notable features. Correlation images show a dramatic dif-
Fig. 8. Correlations between the 560 nm (G) and 660 nm (R) images
with blood pressure illustrate different relationships between the
spectral components of the light scattering changes across the cardiac
cycle. Panel A is a correlation plot of region A (regions defined in
Fig. 8) with blood pressure for 560 and 660 nm images, showing the
two colors are out of phase, with the 660 nm images leading the 560
nm images by 40 ms. In Panel B (Region B), the spectral components
are 180° out of phase, and in Panel C (Region C) and Panel D
(Region D), the phase relationships are close, however, 560 nm leads
the 660 nm images by 20 ms in Panel C and 660 nm leads the 560 nm
images by 20 ms in panel D. Abscissa for each plot represents the
Pearson correlation coefficient ‘r’ for each lag time.
ference in spatial structure between 560 nm and 660
nm illumination. Light at 780 nm and 560 nm show
vessel perfusion in superficial layers and at least two
kinds of perfusion were observed. A direct measurement of blood oxygen level (560 nm absorbance) in
Fig. 9. Panel A illustrates a sequence of dorsal medulla images (780 nm illumination) which were time averaged with the cardiac cycle (left to
right), at five depths from 200 to 1800 mm (top to bottom). Perfusion becomes more diffuse, and develops later in time in the deeper image
sequences as seen in the image sequences and plots of intensity across time (Panel B). The shallow image sequence (200 mm) begins early in the
cycle with an early peak which is well defined at 1000 mm depth, where as deeper image sequences have an initial dip and show a smaller,
delayed peak. Panel C shows a blood pressure correlation image with zero lag, also outlining the region selected for the sequence in Panel A.
Panel D shows a sequence of images averaged after a single shock stimuli to the aortic nerve. A plot of average intensity change in Panel E
shows that the time course of the early optical response follows the evoked electrical response. The colour scale indicates the magnitude for
each colour image. For correlations, green represents no correlation, white represents a correlation of 1.0, and black represents a correlation of
− 1.0. For the cardiac triggered averaged images in Panel A, the colour scale represents relative scattered light intensity from the first frame in
the sequence. Green represents no change, white represents a − 0.1% change (less back scattered light or increased absorbance), and black
represents +0.1% change (more back scattered light or less absorbance). The vertical axis in Panel B also has maximum and minimum values
of − 0.1% and +0.1% respectively. For the stimulus triggered averaged images in Panel D, white represents −0.2% change, and black
represents +0.2% change.
D.M. Rector et al. / Journal of Neuroscience Methods 91 (1999) 135–145
143
Fig. 9.
144
D.M. Rector et al. / Journal of Neuroscience Methods 91 (1999) 135–145
the vessel was correlated directly with blood pressure
(Fig. 6A, D). Secondly, perfusion within the capillary
bed had a delayed latency with respect to the blood
pressure, and was evident across a more diffuse and
deeper tissue area (Fig. 9A, B). Two vessels are visible
within the imaged area (Fig. 6B) at the top and left
edges of the image. Pixels representing these vessels
exhibit a high correlation with zero lag using 560 nm
illumination. Blood pressure correlations with 660 nm
light show optical changes that lead the blood pressure
by 40 ms (Fig. 7A, 8A). Other areas show a negative
correlation with 560 nm light and positive correlation
with 660 nm light (Fig. 7B, 8B). Fig. 7D and 8D show
a region of positive correlation with 660 nm light which
has no apparent structure within the 560 nm images,
and may represent neural groups which are active in
phase with the cardiac cycle. Evidence of vessel and
capillary bed perfusion differences also appear, since
the 560 nm plots of each area shows a sharp rising
slope for areas immediately over a vessel, and a slow,
gradual slope for areas not immediately over a vessel.
Time-triggered averages of stimulus sequences with
780 nm illumination show rapid optical changes that
parallel the evoked electrical response. The spatial pattern is distinct from vessels imaged in the EKG triggered averages from the same data set, and correspnds
to anatomical regions to which the stimulated nerve
projects. There is some evidence that the fastest component of the optical response is a negative deflection that
may precede the arrival of the population spike; an
observation made by Salzberg et al., 1985. We speculate
that a compression wave set up at the stimulus point
will travel through the axon at the speed of sound,
arriving before the propagating action potential volley.
A number of studies have investigated the relationship between light scattering changes and neural activity (Cohen, 1973; Grinvald et al., 1988). Many physical
processes are mobilized concurrently with cell activity.
Predominant mechanisms include: direct cellular structural changes associated with ion and water influx,
cellular swelling and membrane unfolding occurring
with a sub-millisecond time-course (Salzberg et al.,
1985; Tasaki and Byrne, 1992; Rector et al., 1997a);
rapid vascular coupling observed in response to stimulation (Lindauer et al., 1993; Villringer and Dirnagl,
1995; Ruben et al., 1997); and a slower (1 – 10 s) hemodynamic change in oxygen level and blood volume
associated with energy demand (Malonek and Grinvald, 1996, 1997; Malonek et al., 1997). There is evidence that glial cells or large dendritic fiber beds
contribute to light scattering changes in vitro (Andrew
and MacVicar, 1994), but it is unclear how significant
these changes are in vivo. Physical mechanisms which
contribute to light scattering changes may be present
across all illumination wavelengths. However, the rela-
tive contribution of hemoglobin absorbance changes is
significantly higher for 560 nm light than for 660 or 780
nm light. Similarly, the relative contribution of light
scattering changes from cellular processes such as
swelling and membrane unfolding appears greater with
660 and 780 nm light.
The ability to focus deep below the surface allows
more sensitive and less invasive approaches for recording cellular activation. Images of a test object placed
within neural tissue show improved imaging capability
over previous techniques for such imaging. Since different cellular and tissue processes modulate the amount
of back-scattered light with varying efficiencies as a
function of wavelength, we can characterize the contributing processes based on illumination at different
selected wavelengths. Time-triggered averaged image
sequences associated with physiological activity as well
as electrical stimulation shoe spatio–temporal patterns
corresponding to fast neurally related changes, as well
as slower hemodynamic processes. Such signals might
be used to map activation patterns and study dynamic
network behaviour within large neural populations.
Acknowledgements
This study was funded by the US Department of
Energy, Technology Transfer Initiative; D.R. is a Director’s Fellow at Los Alamos National Laboratory;
R.R. is a scientist with the Du Pont De Nemours
Company and supported by 1-R01-HL-54194..
References
Andrew RD, MacVicar BA. Imaging cell volume changes and neuronal excitation in the hippocampal slice. Neuroscience
1994;62(2):371– 83.
Cohen LB. Changes in neuron structure during action potential
propagation
and
synaptic
transmission.
Physiol
Rev
1973;53(2):373– 413.
Grinvald A, Frostig RD, Lieke E, Hildesheim R. Optical imaging of
neuronal activity. Phsyiol Rev 1988;68(4):1285– 368.
Holthoff K, Dodt HU, Witte OW. Changes in intrinsic optical signal
of rat neocortical slices following afferent stimulation. Neurosci
Lett 1994;180(2):227– 30.
Holthoff K, Witte OW. Intrinsic optical signals in rat neocortical
slices measured with near-infrared dark-field microscopy reveal
changes in extracellular space. J Neurosci 1996;16(8):2740–9.
Lindauer U, Villringer A, Dirnagl U. Characterization of CBF response to somatosensory stimulation: model and influence on
anesthetics. Am J Physiol 1993;264:H1223– 8.
Malonek D, Dirnagl U, Lindauer U, Yamada K, Kanno I, Grinvald
A. Vascular imprints of neuronal activity: relationships between
the dynamics of cortical blood flow, oxygenation, and volume
changes following sensory stimulation. Proc Natl Acad Sci USA
1997;94(26):14826– 31.
Malonek D, Grinvald A. Interactions between electrical activity and
cortical microcirculation revealed by imaging spectroscopy: implications
for
functional
brain
mapping.
Science
1996;272(5261):551– 4.
Malonek D, Grinvald A. Vascular regulation at sub millimeter range.
Sources of intrinsic signals for high resolution optical imaging.
Adv Exp Med Biol 1997;413:215 – 20.
D.M. Rector et al. / Journal of Neuroscience Methods 91 (1999) 135–145
Moore DT, editor. Selected Papers on Gradient-Index Optics.
Bellingham, Washington: SPIE Press, 1993.
Rector DM, Harper RM. Imaging of hippocampal neural activity in
freely behaving animals. Behav Brain Res 1991;42:143–9.
Rector DM, Poe GR, Harper RM. Fiber optic imaging of subcortical
neural tissue in freely behaving animals. In: Dirnagl U, et al.,
editors. Optical Imaging of Brain Function and Metabolism. New
York: Plenum Press, 1993:81–6.
Rector DM, Poe GR, Kristensen MP, Harper RM. Light scattering
changes follow evoked potentials from hippocampal schaeffer
collateral stimulation. J Neurophysiol 1997a;78:1707–13.
Rector DM, Poe GR, Redgrave P, Harper RM. A miniature CCD
video camera for high-sensitivity light measurements in freely
behaving animals. J Neurosci Methods 1997b;78:85–91.
.
145
Ruben J, Wenzel R, Obrig H, Villringer K, Bernarding J, Hirth C,
Heekeren H, Dirnagl U, Villringer A. Haemoglobin oxygenation
changes during visual stimulation in the occipital cortex. Adv Exp
Med Biol 1997;428:181 – 7.
Salzberg BM, Obaid AL, Gainer H. Large and rapid changes in light
scattering accompany secretion by nerve terminals in the mammalian neurohypophysis. J Gen Physiol 1985;86:395 – 411.
Tasaki I, Byrne PM. Rapid structural changes in nerve fibers evoked
by electrical current pulses. Biochem Biophys Res Commun
1992;188(2):559– 64.
Villringer A, Dirnagl U. Coupling of brain activity and cerebral
blood flow: basis of functional neuroimaging. Cerebrovasc Brain
Metab Rev 1995;7(3):240 – 76.