Confocal imaging for 3-D digital microscopy
Kjell Carlsson and Nils Aslund
Optical serial sectioning based on the depth-discriminating ability of confocal laser scanning can be combined
with digital image processing to realize fast and easy-to-use 3-D microscopy. A great advantage as compared
with traditional methods, e.g., using a microtome, is that the specimen is left undamaged. An account is given
of an instrument designed for this purpose and of feasibility studies that have been carried out to assess the
usefulness of the method in fluorescence microscopy.
1.
Introduction
An ordinary light microscope is not well suited for
studying the 3-D structure of a specimen. It presents
the operator with a 2-D image consisting of a superposition of in-focus and out-of-focus regions of the specimen. Stereomicroscopes are useful in some applications requiring only low magnifications, but are not
useful in high-resolution microscopy. In spite of these
problems, researchers have for a long time used micro-
scopes for studying very complicated depth structures.
One method of doing this is to photograph the specimen at a number of different focus settings. By manually tracing only the focused parts of each photograph, it is possible to get information on the 3-D
structure of the specimen. This may be difficult, however, if the specimen does not contain very clear borderlines. The problem can be avoided by using a
computer to process digitized photographs in such a
way that the out-of-focus information is filtered out.1 -5
While the feasibility of computer processing has been
demonstrated in practical applications, a number of
difficulties are encountered; e.g., the optical transfer
function of the microscope must be known for different
amounts of defocusing, and faint in-focus regions may
be difficult to record in the presence of strong out-offocus regions.
Another method to realize 3-D microscopy is to actually cut the specimen into a large number of thin sections, which are then studied under a microscope one
at a time. While circumventing a number of problems
associated with the previous methods, this method
creates a number of new problems. The cutting process is time-consuming and may deform the specimen.
It is impossible to study living specimens, and the
reconstruction of the 3-D structure from the individual
sections may require considerable computer processing to align the recorded sections.
II.
Confocal Scanning Microscopy
A promising method for 3-D studies of specimens is
confocal microscopy. The principle of confocal imag-
ing, and the properties of the recorded images have
been described in a patent, 6 a number of articles,7 -' 2
and a textbook.13 The basic principle is that the specimen is illuminated one point at a time, and the detector only registers light emanating from the illuminated
point (Fig. 1). An image can be scanned either by
moving the specimen or the light beam. A number of
confocal scanning microscopes have been described in
the literature.1 4 -18 In one design a large number of
points are illuminated simultaneously, making it pos19 20
sible to build a direct-view confocal microscope
which allows the operator to view a depth section in
real time through the eyepieces without the need for
electronic recording.
The point-illumination and point-detection scheme
results in imaging properties that are different from
those of an ordinary microscope. One such difference
is that the resolution limit is improved. Another is
that a very pronounced depth discrimination is obtained (compare Fig. 1). This means that different
depth layers can be studied much more clearly, since
virtually no out-of-focus information is superimposed
on the image.2 1-2 7 The depth discrimination can also
be used to study surface structures.2 8
Confocal scanning offers a unique possibility for 3-D
The authors are with Royal Institute of Technology, Physics IV, S100 44 Stockholm, Sweden.
Received 6 November 1986.
0003-6935/87/163232-07$02.00/0.
© 1987 Optical Society of America.
3232
APPLIEDOPTICS / Vol. 26, No. 16 / 15 August 1987
microscopy, since a recording of the entire structure of
a specimen can be made by scanning a number of
confocal images, refocusing the microscope between
successive images.
The result is a stack of images
representing the 3-D structure of the specimen. This
Specimen
Beamsplitter
Aperture
I
I
objective
Laser
light
Fig. 1. Simplified ray path of a confocal microscope with incident
light illumination. A focused laser beam provides intense illumination of a small specimen volume located in the focal plane of the
microscope, A. Reflected or fluorescent light from A is transmitted
through the detector aperture, which effectively blocks light from
out-of-focus planes, e.g., B. By scanning either the laser beam or the
specimen, an image can be recorded that represents a thin section
located at A. Repeated scanning, using different focus settings on
the microscope, results in a stack of images representing the 3-D
Fig. 3.
Confocal microscope scanning system.
structure of the specimen.
recording, optical serial sectioning, can be done quickly, and there is no need for processing the pictures to
men from different viewing angles.2 4 - 26 Moreover, the
possibility to perform photometry in three dimensions
align them with respect to each other. The method
also makes it possible to record thin sections of living
is opened.
scanning with digital recording and image processing.
The images resulting from optical serial sectioning can
aimed at digital 3-D microscopy has been built.1 8 24 25
The instrument has been used in a number of feasibility studies, some of which will be reported in this
article. The instrument (Figs. 2 and 3) uses the beam
scanning technique, i.e., a laser beam from an argon
specimens. 2 3
New perspectives are opened by combining confocal
also be dealt with in this way, resulting in a digitally
recorded data volume. It is then possible to make
projections of this volume (Sec. V) to display the speci-
I11. Equipment
At our institute
a confocal scanning microscope
Optional
I
IEEE 488
i SBX488
I
(GPIB)
(GPIB)I
>
I
GaIv.
scanner
Fig. 2. Schematic diagram of the confocal microscope scanning system used at The Royal Institute of Technology, Stockholm. The
instrument is controlled by a microprocessor (Intel 80186) which coordinates scanning in three dimensions, data collection, and communication with a host computer (using GPIB or Ethernet interfaces). A laser beam is scanned over the specimen by means of two mirrors, and the reflected or fluorescent light from the specimen is detected by a photomultiplier tube (PM), A/D converted, and stored in an image memory. In
this way digital images with up to 1024 X 1024 pixels and 1024 gray levels can be registered. Scanning in the third dimension is done by a computer-controlled stepping motor connected to the focus control of the microscope. Using this instrument a stack of digital images can be
scanned and transferred to a host computer automatically.
15 August 1987 / Vol. 26, No. 16 / APPLIEDOPTICS
3233
laser performs scanning in two perpendicular directions while the specimen remains stationary. Incident
light illumination is used, and the reflected or fluorescent light from the specimen is detected by a photomultiplier tube. A Zeiss Universal microscope is used
in the scanner, the only modification being that a
stepping motor has been connected to the fine focusing
Computer processing of the digital images will be de-
mechanism, allowing a 2-mm computer-controlled focusing range. In fact, the scanner can be regarded as a
microscope accessory that can be used instead of a
well as the spacing between successive sections.
camera unit.
Prior to scanning, the operator can view the specimen using conventional microscopictechniques to find
an interesting region. Scanning with the laser beam is
performed by two mirrors located above the eyepiece
of the photographic tube. The fast line scan mirror is
driven by a galvanometric scanner (General Scanning,
Watertown, MA) and the slow frame scan mirror is
driven by a stepping motor. The scanned specimen
area corresponds to 13 X 13 mm in the image plane of
the microscope objective. This area can be scanned
with a maximum resolution of 1024 X 1024 pixels (pic-
ture elements), each pixel containing 10 bits of data.
The useful number of bits depends, of course, on the
signal-to-noise ratio of the detector signal, which varies widely between different types of specimen. Usually only the eight most significant bits are used.
To illuminate the specimen, an Ar-ion laser (Coherent Innova 70-4)is used. The maximum output power
of this laser, 4 W, is much higher than the power
needed for scanning (a few milliwatts). The reason for
using such a powerful laser is that it offers a wider
range of wavelengths (458-514 nm). In many applications, however, a small air-cooled argon laser or a HeCd laser would be sufficient. For reflected light scanning, the beam splitter (compare Fig. 1) is of the ordinary 50/50 type, splitting all wavelengths equally.
When scanning fluorescent specimens a dichroic beam
splitter, reflecting the excitation light while transmitting the fluorescent light, is used. To facilitate the
change between different beam splitters, these are
scribed in Sec. V.
IV.
Resolution
The optical resolution both along and perpendicular
to the optical axis is of fundamental importance when
selecting the pixel spacing in each depth section, as
As in
ordinary microscopy, the resolution is determined by
the wavelength, the numerical aperture of the objective, and the coherence properties of the light. Although a laser is usually used for illuminating the
specimen in confocal microscopy, imaging may be incoherent. This is the case, e.g., in fluorescence micros-
copy, since the coherence of the laser light is not preserved in the fluorescent light. The focal plane
resolution (two-point resolution) for an ordinary microscope and incoherent imaging is given by29
_ 0.61X
f
N.A.
(1)
where N.A. is the numerical aperture.
In confocal
microscopy, the resolution is improved by a factor of
1.32for incoherent imaging,12 and, therefore, the resolution limit is given by
0.46\
Rf= N.A.
(2)
A more important improvement, as far as 3-D microscopy is concerned, is that a confocal microscope has a
very pronounced depth discrimination. A suitable
way of studying this is to calculate the fall-off in the
total light intensity in the image of a point object as the
microscope is defocused. The light fall-off as a function of a normalized distance u from the focal plane is
plotted in Ref. 9, giving a 50% fall-off at u
4.4 (u is
defined in Ref. 29, p. 437). For small numerical apertures u can be approximated by
= 2r(N.A.)2 z
An
(3)
mounted on a wheel. Another wheel, located in front
where n is the refractive index of the immersion medium, and z is the distance along the optical axis. Combining these results it is possible to estimate the halfwidth (full width half-maximum) along the optical axis
resolution
as
of the photomultiplier tube, holds a number of detector apertures of different sizes. While the aperture
size should ideally be as small as possible to get high
(see Sec. IV), a certain minimum area is
necessary to get enough light for the detector. Since
the optimum trade-off between resolution and signal
quality depends on the type of specimen, and the number of pixels in the digital image, we feel that a system
allowing a quick change of aperture size is advantageous.
To control scanning and data collection a 16-bit
microprocessor, Intel 80186,is used. Connected to the
microprocessor is an image memory (Matrox MIP1024) that can hold 1 Mbyte of data. During scanning,
which takes
5s for a 256 X 256 image, the image
gradually appears on a TV screen. When the scanning
of an image is complete, data are transferred to a host
computer (Perkin-Elmer 3210) via an IEEE 488 data
bus (Fig. 2). The transfer speed of the bus is 25 kbyte/s.
3234
APPLIEDOPTICS / Vol. 26, No. 16 / 15 August 1987
Rd =
1.4nX2
)
(N.A.)
(4)
In an ordinary microscope there is no light fall-off
whatsoever when the specimen is defocused. Using
Eqs. (2) and (4) it is possible to calculate the maximum
resolution obtainable in confocal microscopy, using
commercially available objectives (N.A.
1.4) and
visible light (X> 400 nm). The result, Rf = 0.13 and Rd
= 0.43 m, shows that a depth discrimination comparable to the focal plane resolution cannot be expected.
Furthermore, while the focal-plane resolution is inversely proportional to the numerical aperture, the
depth discrimination is inversely proportional to the
square of the numerical aperture. Therefore, the
depth discrimination drops considerably faster than
the focal-plane resolution when the numerical aperture becomes smaller, making it very important to use
objectives with a large numerical aperture.
ZZ
A serious
is that the working distance is very small, often below
100 Am. This proves to be a limiting factor when using
confocal microscopy for 3-D studies of thick specimens. A numerical aperture of -1.0, giving a working
distance of 300-400 ,um, is often a good compromise
I
11-1"
-
-
7)"
l >
I
is 0.8 m for X =
The resolution limits given are not exact, since approximations have been made in the calculations.13
For example, the assumption that the numerical aperture is small is often not valid. Practical observations
of the focal-plane resolution have shown, nevertheless,
good agreement even for large numerical apertures.'
at-==
1S!,f
between depth discrimination and working distance.
In this case the depth discrimination
400 nm.
Z ZJ
I
disadvantage with large aperture objectives, however,
(a)
/II~~~
6
Equations (2) and (4) also assume that the detector
aperture is infinitely small. According to theory, the
focal-plane resolution is rather insensitive to aperture
size. Even if the aperture is removed altogether, the
two-point resolving power would only drop to a point
where it is identical to that of a conventional microscope,7 i.e., Eq. (1). The real situation, however, is
often complicated by light scatter in the specimen,
which severely limits the image contrast if the detector
aperture is too large. Therefore, the aperture should
not be larger than the pixel spacing.
Depth discrimination, on the other hand, is sensitive
to the size of the detector aperture. With a very large
aperture, the scanning microscope behaves like an ordinary microscope in this respect, i.e., light from all
depth layers can reach the detector. We have used
aperture sizes ranging from one to one-quarter the
diameter of the Airy disk with good results. An aperture size roughly equal to the Airy disk is sometimes
the smallest that is practically useful in fluorescence
microscopy, if a good signal quality is to be obtained.
When studying the depth discrimination, using the
interface between a fluorescein solution and a cover
glass, it was not possible to detect any degradation
when the larger apertures were used. For a numerical
aperture of 1.3 and X = 458 nm, the 10 to 90% transition
in intensity took place over a vertical distance of 1.7
gm. We have not used apertures larger than the Airy
disk, since this would reduce the lateral resolution
(aperture size > pixel spacing). One could argue that,
by simply using a higher laser power, it would always
be possible to use a very small aperture. This is usually not the case in fluorescence microscopy, since most
specimens are prone to fading if illuminated too
strongly.
In our microscope, using N.A.
=
1.3 and X = 458 nm,
the half width of the lateral spread function in the
scanned images is -0.2 ,um (corresponding to a two-
point resolution of -0.25 gm), when using an aperture
size of half of the Airy disk. To fully utilize this
resolution the pixel spacing must be smaller than the
resolution limit. But scanning the entire field of view
of the microscope with such a small pixel spacing can
(b)
Fig. 4. (a) Values of volume elements along projection lines (here
parallel to one of the coordinate axes) are added to generate a
projected image. (b) For many objects the amount of information
that needs to be stored can be considerably reduced, e.g., if the object
consists of a shell. This can be done by retaining only the interesting
volume elements in vector form. Figures 5-7 have been handled in
this way.
yield prohibitively large amounts of data, especially
when many depth sections are recorded. By scanning
a smaller area, the optical resolution can, of course, be
fully utilized. In a typical case we scan 200 depth
sections, each having 256
X
256 pixels.
The total
amount of data, 13 Mbyte, is rather large but still
manageable.
V.
Computer Hardware and Software
Computer technology offers the possibility to store
and access 3-D information in a fast and efficient way.
The stack of digital images constitutes a 3-D matrix,
where the values of the elements are related to the
amount of light recorded from the corresponding volume element in the specimen. The sections of this
stack can be displayed and evaluated one by one. Al-
ternatively, it is possible to produce projections
through the data volume24 -2 6 30 (Fig. 4). The 3-D matrix stored in the computer may be processed in a
number of ways. For example, it is possible to generalize the well-established techniques for dealing with
images in two dimensions.
Concepts like resampling,
filtering, edge-detection, segmentation, etc. all have
their counterparts in three dimensions.
Another approach is to perform data compression
based on vector representation 3 The underlying as15 August1987 / Vol. 26, No. 16 / APPLIEDOPTICS
3235
sumption of this approach is that the information content of the matrix is sparse. If this is the case, and the
uninteresting elements can be filtered out, e.g., by
thresholding, only the interesting elements need be
retained in vector form. In many practical cases this
proves to be a very economical way of storing the
information compared with storing the entire matrix.
Each specific vector describes the position and the
measured light value for a volume element. Once the
information has been transformed to this representation, the well-known set of procedures to deal with
vectors can be applied.
To prepare for the vector representation it is often
advantageous to enhance the border surfaces, which
can be done by computing the approximate 3-D gradi-
ent for each point in the original data. The transformed volume is subsequently replaced by a representation based on vectors.
Another class of operations to perform on the matrices has to do with the reconstructions of the true light
values from the recorded ones. This is a prerequisite
for establishing a 3-D microphotometry.
A need for
such reconstructions comes from the fact that the
depth discrimination
is not ideal; there is always a
small amount of cross talk between adjacent sections.
Another need comes from the fact that the light on its
way to and from a specific element is attenuated
by
absorption in the parts of the specimen that it passes
through. The amount of this absorption depends on
the shape, density, etc., of the specimen, i.e., on factors
that are unknown at the outset of the reconstruction.
The problem is complicated by the fact that absorption is a nonlinear phenomenon. Much basic development work still remains to be done in this field, and all
the studies reported in this article have been made
without employing any reconstructions.
To speed up the vector calculations, e.g., when making projections, we use an array processor (Floating
Point Systems, FPS-100). For display and evaluation
purposes we use a workstation (Koala Bear Systems,
Ltd.) with an internal memory to store a set of images
locally (typically 24 images of the size 256 X 256 X 8
bits). Flipping between images in this memory can be
performed instantaneously, an important feature both
when doing subjective comparisons between different
images and when simulating a rotation using a stack of
projection images.
VI.
Results
To assess the usefulness of confocal scanning and
computer reconstructions in 3-D microscopy, a number of feasibility studies have been carried out at our
institute. The specimens studied represent a wide
field of applications including medicine, zoology,botany, dentistry, and environmental protection among
others. For confocal microscopy to be successfully
applied, the specimen must be reasonably transparent
to allow light to penetrate to regions below the surface
of the specimen. If this is not the case, only microtome
sectioning can solve the problem.
3236
APPLIEDOPTICS / Vol. 26, No. 16 / 15 August 1987
We have selected three different applications for
presentation, two physiological applications, and one
from botany. In all cases fluorescence microscopy
with epi-illumination was used. The first application
relates to the study of neurons, which often display a
very complicated 3-D structure. In a joint project
between our institute and the Karolinska Institute in
Stockholm, Sweden, we have undertaken to make a
study of the structure of neurons. The neurons studied include motoneurons, lateral interneurons, edge
cells, and dorsal cells in the spinal cord of lamprey
(Ichthyomyzon unicuspis). The neurons were stained
by intracellular injection of the fluorescent dye Lucifer
Yellow,fixed in formaldehyde, dehydrated in ethanol,
and cleared in methyl salicylate.3 2 3, 3 Confocal microscopy has proved useful in studying specimens of this
kind (Fig. 5). By displaying a number of projections
through the recorded data volume in rapid succession
on a TV screen, a good understanding of the general 3D structure can be obtained. A more detailed study of
the entire tree of neuronal processes, which often extends for several millimeters and in which the individual processes may have a thickness of <1 im, is difficult because of the enormous amount of data that must
be handled. Even when using vector representation, it
is impractical with our present computer equipment.
However, a detailed study of part of the dendritic tree,
e.g., tracing individual processes and calculating their
entire length, seems feasible. A difficulty that often
arises in the study of neurons is that the specimen
thickness, often several hundred microns, makes it
impossible to scan the deepest part of the specimen
using an objective with a large numerical aperture.
Although projections have been successfully used
for displaying, e.g., neurons, there are a number of
cases where the original data consisting of a number of
sections are more useful. If this is the case, display
and evaluation can be performed more quickly since
much computer processing can be avoided. One application where the original sections have proved useful is the study of plant seeds. In cooperation between
our institute and the University of Gothenburg, Sweden, a new species of Neuwiedia (an orchid) has been
studied (Fig. 6). The traditional way to study ovules
of plants is to embed the ovule in paraffin, section it
with a microtome, and stain the sections to make their
structure visible in transmitted light microscopy. Because the ovules are fragile, thinner sections than 8um
are difficult to obtain. By using confocal microscopy,
serial sections with a thickness of -1 /im can easily be
recorded from whole ovules. Furthermore, it turned
out that the autofluorescence from the ovule wall, and
the nuclei located in the ovule, was so strong that
staining the specimen was not necessary. As a result,
ovules could be studied much more quickly and accurately by confocal microscopy than by using traditional methods. Since data are in digital form, it was also
easy to perform quantitative measurements of 3-D
positions of nuclei, the size of the ovule, etc.
The last application reported is a study of lung tissue, which was started only recently, but has shown
(a)
E
. .9 *- no'~~~~~~~~~~~~~~~~~~~..
;.MW-//fmgB.
.i.
75...
......;......
..z....;_......
I. 'l
11..
Fig. 6. Digital image (256 X 256 pixels) of an ovule from a plant (a
species of Neuwiedia). This picture, showing one depth section -2
gm thick, was registered by confocal scanning. The specimen was
illuminated by an argon laser (X = 458 nm), and the autofluorescence
from the specimen was detected. To bring out the details more
clearly on the television screen, the gray scale has been inverted, i.e.,
dark regions correspond to a high light intensity. A specimen area
Objective: Zeiss
of 320 X 320 Am is shown in the photograph.
planapo 40/1.0.
(b)
Fig. 5. Lateral interneuron from the spinal cord of a lamprey
(Ichthyomyzon unicuspis). To make the neuron visible, intracellular injection with a fluorescent dye (Lucifer Yellow) was made.
After confocal scanning, using a laser wavelength of 458 nm and
detecting the fluorescent light, projections were calculated from a
data volume consisting of 110 digital images, each containing 256 X
256 pixels. The pixel size, as well as the spacing between the images,
was 1.3 sum. In (a) the specimen is viewed from the same direction as
when looking through the microscope. In (b) the specimen is viewed
from the side, i.e., in a direction parallel to the plane of the object
glass. The specimen area shown in the photographs is 320 X 320 Am.
Objective: Zeiss planapo 40/1.0. (Note: Any moir6 fringes visible
in the photographs are caused by the printing process; they were not
present in the originals.)
promising results. The study is made in cooperation
with The Memorial Hospital, Brown University,
Rhode Island, U.S.A. Again it seems that data in the
form of scanned optical sections of lung tissue can be
used directly for evaluation. As in the previous case,
the traditional way to study the specimen has been to
physically section it with a microtome. By fluorescent
Fig. 7. Digital image (256 X 256 pixels) of lung tissue from rat. By
confocal scanning, a thin section of the fluorescently labeled tissue
has been registered, clearly showing the alveoli. A sequence of such
depth sections can be used to reconstruct the 3-D structure of the
tissue.
Compared with physical sectioning of the specimen using a
microtome, this method is faster and does not damage the specimen.
Furthermore,
no processing to align the recorded images with re-
spect to each other is needed. The specimen area shown in this
photograph is 320 X 320 Am. Objective:
Zeiss planapo 40/1.0.
15 August 1987 / Vol. 26, No. 16 / APPLIEDOPTICS
3237
staining of the tissue before it is embedded in epoxy, it
has proved possible to use confocal microscopy to register optical sections with high contrast (Fig. 7). From
a stack of such sections interesting parameters concerning the 3-D structure of lung tissue can be measured.
Vil.
Conclusions
Feasibility studies have shown the potential of confocal microscope scanning as a tool in 3-D microscopy.
Since the recorded volume data are in digital form,
they are directly accessible for computer processing
and evaluation. This opens up a whole new field of
possibilities for display and evaluation. However, the
amount of data to be processed is often very large,
requiring powerful computer hardware and efficient
software. We have found that fluorescent specimens
are often wellsuited for 3-D scanning. Objectives with
large numerical apertures should be used, but the
small working distance of very large aperture objec-
tives (>1.0) can sometimes create problems.
We wish to thank the following persons for very
rewarding cooperation and for providing us with excellent specimens: Sten Grillner and Peter Wallen, Karolinska Institute, Stockholm, Margit Fredrikson, University of Gothenburg, and Eben Oldmixon, The
Memorial Hospital, Brown University, Rhode Island.
We would also like to thank our colleagues at Physics
IV, The Royal Institute of Technology, and at the
Picture Processing Laboratory, Linkoping University.
This research was supported by the National Swedish Board for Technical Development.
References
1. K. R. Castleman, Digital Image Processing (Prentice-Hall, Englewood Cliffs, NJ, 1979), p. 351.
2. D. A. Agard and J. W. Sedat, "Three-Dimensional Architecture
of a Polytene Nucleus," Nature 302, 676 (1983).
3. D. A. Agard, "Optical Sectioning Microscopy: Cellular Architecture in Three Dimensions," Ann. Rev. Biophys. Bioeng. 13,
191 (1984).
4. A. Erhardt, G. Zinser, D. Komitowski, and J. Bille, "Reconstructing 3-D Light Microscopic Images By Digital Image
Processing," Appl. Opt. 24, 194 (1985).
5. N. Streibl, "Three-Dimensional Imaging By a Microscope," J.
Opt. Soc. Am. A 2, 121 (1985).
6. M. Minsky, "Microscopy Apparatus," U.S. Patent 3,013,467
(December 19, 1961).
7. C. J. R. Sheppard and A. Choudhury, "Image Formation in the
Scanning Microscope," Opt. Acta 24, 1051 (1977).
8. C. J. R. Sheppard and T. Wilson, "Image Formation in Scanning
Microscopes with Partially Coherent Source and Detector,"
Opt. Acta 25, 315 (1978).
9. C. J. R. Sheppard and T. Wilson, "Depth of Field in the Scanning Microscope," Opt. Lett. 3, 115 (1978).
10. T. Wilson, "Imaging Properties and Applications of Scanning
Optical Microscopes," Appl. Phys. 22, 119 (1980).
11. D. K. Hamilton, T. Wilson, and C. J. R. Sheppard, "Experimen-
tal Observations of the Depth-Discrimination Properties of
Scanning Microscopes," Opt. Lett. 6, 625 (1981).
12. I. J. Cox, C. J. R. Sheppard, and T. Wilson, "Super-Resolution
by Confocal Fluorescent Microscopy," Optik 60, 391 (1982).
3238
APPLIEDOPTICS / Vol. 26, No. 16 / 15 August 1987
13. T. Wilson and C. Sheppard, Theory and Practice of Scanning
Optical Microscopy (Academic,London, 1984).
14. P. Davidovits and M. D. Egger, "Scanning Laser Microscope for
Biological Investigations," Appl. Opt. 10, 1615 (1971).
15. A.F. Slomba, D.E. Wasserman, G. I. Kaufman, and J. F. Nester,
"A Laser Flying Spot Scanner for Use in Automated Fluorescence Antibody Instrumentation,"
6, 230 (1972).
J. Assoc. Adv. Med. Instrum.
16. G. J. Brakenhoff, P. Blom, and P. Barends, "Confocal Scanning
Light Microscopy with High Aperture Immersion Lenses," J.
Microsc. 117, 219 (1979).
17. H. J. B. Marsman, R. Stricker, R. W. Wijnaendts van Resandt,
G. J. Brakenhoff, and P. Blom, "Mechanical Scan System for
Microscopic Applications," Rev. Sci. Instrum. 54, 1047 (1983).
18. N. Aslund, K. Carlsson, A. Liljeborg, and L. Majlf, "PHOIBOS,
a Microscope Scanner Designed for Micro-Fluorometric Appli-
cations, Using Laser Induced Fluorescence," in Proceedings,
Third Scandinavian Conference on Image Analysis (Studentlitteratur, Lund, Sweden, 1983), p. 338.
19. M. Petran, M. Hadravsky, M. D. Egger, and R. Galambos, "Tan-
dem-Scanning Reflected-Light Microscope," J. Opt. Soc. Am.
58, 661 (1968).
20. C. J. R. Sheppard and T. Wilson, "The Theory of the DirectView Confocal Microscope," J. Microsc. 124, 107 (1981).
21. A. Boyde, M. Petran, and M. Hadravsky, "Tandem Scanning
Reflected Light Microscopy of Internal Features in Whole Bone
and Tooth Samples," J. Microsc. 132, 1 (1983).
22. I. J. Cox, "Scanning Optical Fluorescence Microscopy," J. Microsc. 133, 149 (1984).
23. R. W. Wijnaendts van Resandt, H. J. B. Marsman, R. Kaplan, J.
Davoust, E. H. K. Stelzer, and R. Stricker, "Optical Fluorescence Microscopy in Three Dimensions: Microtomoscopy," J.
Microsc. 138, 29 (1985).
24. K. Carlsson, P. E. Danielsson, R. Lenz, A. Liljeborg, L. Majl6f,
and N. Aslund, "Three-Dimensional Microscopy Using a Confocal Laser Scanning Microscope," Opt. Lett. 10, 53 (1985).
25. K. Carlsson, A. Liljeborg, and N. Aslund, "Three-Dimensional
Scanning of Microscopic Specimens Using a Laser Microscope
Scanner," in Proceedings, Fourth Scandinavian Conference on
Image Analysis (Tapir, Trondheim, Norway, 1985),p. 905.
26. G. J. Brakenhoff, H. T. M. van der Voort, E. A. van Spronsen, W.
A.M. Linnemans, and N. Nanninga, "Three-Dimensional Chromatin Distribution in Neuroblastoma Nuclei Shown by Confocal Scanning Laser Microscopy," Nature 317, 748 (1985).
27. E. H. K. Stelzer and R. W. Wijnaendts van Resandt, "Applications of Fluorescence Microscopy in Three Dimensions/Microtomoscopy," Proc. Soc. Photo-Opt. Instrum. Eng. 602,63 (1985).
28. D. K. Hamilton and T. Wilson, "Three-Dimensional
Surface
Measurement Using the Confocal Scanning Microscope," Appl.
Phys. B 27, 211 (1982).
29. M. Born and E. Wolf, Principles of Optics (Pergamon, New
York, 1983).
30. R. Lenz, "Reconstruction, Processing, and Display of 3-D Images," Ph.D. Dissertation No. 151, Linkbping University, Sweden
(1986).
31. P. 0. Forsgren, "Using Vector Representation to Display 3-D
Data from a Confocal Laser Scanning Microscope," in Proceedings, Fifth Scandinavian Conference on Image Analysis; p. 367
(1987).
32. W. W. Stewart, "Functional Connections Between Cells as Revealed by Dye-Coupling with a Highly Fluorescent Naphthalimide Tracer," Cell 14, 741 (1978).
33. P. Wallen, S. Grillner, J. L. Feldman, and S.Bergelt, "Dorsal and
Ventral Myotome Motoneurons and Their Input During Fictive
Locomotion in Lamprey," J. Neurosci. 5, 654 (1985).
S