Article

Optical sectioning microscopy with planar or structured illumination

Department of Biomedical Engineering, Boston University, Boston, Massachusetts, USA.
Nature Methods (Impact Factor: 32.07). 09/2011; 8(10):811-9. DOI: 10.1038/nmeth.1709
Source: PubMed
ABSTRACT
A key requirement for performing three-dimensional (3D) imaging using optical microscopes is that they be capable of optical sectioning by distinguishing in-focus signal from out-of-focus background. Common techniques for fluorescence optical sectioning are confocal laser scanning microscopy and two-photon microscopy. But there is increasing interest in alternative optical sectioning techniques, particularly for applications involving high speeds, large fields of view or long-term imaging. In this Review, I examine two such techniques, based on planar illumination or structured illumination. The goal is to describe the advantages and disadvantages of these techniques.

Full-text preview

Available from: nature.com
Anyone who has used a standard fluorescence micro-
scope has faced a familiar problem: when an object goes
out of focus, its image becomes blurred but does not dis-
appear. This simple problem presents a major headache
in microscopy, particularly when imaging thick samples
in which only a small portion of the sample is in focus
and the rest of the sample can generate so much blurred
background that the in-focus contribution becomes
indiscernible. How do we solve this problem? In imag-
ing parlance, a standard fluorescence microscope does
not provide optical sectioning. To properly reject out-of-
focus background, two strategies readily come to mind:
the most straightforward strategy is to not generate out-
of-focus background in the first place. If there is no back-
ground to detect, it will not appear in the image. A more
involved strategy is to reject out-of-focus background,
either before or after the detection process, thus remov-
ing it from the image. Both strategies have been used to
develop a variety of microscopy techniques, each with
advantages and disadvantages. An example of the first
strategy is two-photon excitation fluorescence micros-
copy
1
, where fluorescence signal is produced only in a
small volume confined at the focus owing to a nonlinear
interaction between light and matter, that is, out-of-focus
background is inherently not generated. An example of
the second strategy is confocal laser scanning micros-
copy (CLSM)
2
. In this case, the interaction between light
and matter is linear. Fluorescence is generated above
and below the focus, but it is subsequently rejected by
a pinhole mask. Both CLSM and two-photon excitation
fluorescence have been reviewed elsewhere
3–5
, and serve
here as reference standards. For consistency of notation,
I will refer to these as one-photon and two-photon LSM
(1P-LSM and 2P-LSM), respectively.
In this Review, I will concentrate on other techniques
that have been gaining new or renewed attention over
the last few years and are only now appearing as com-
mercial products (in some cases). The two techniques I
will consider are based on planar illumination and struc-
tured illumination. I will examine these techniques and
describe their features. As we will see, different methods
are prescribed for different applications.
Plane illumination microscopy
The idea of plane illumination microscopy (PIM) is to
illuminate the sample side-on with a thin laminar sheet
of light, thus exciting only a two-dimensional (2D) sec-
tion of the sample. The emitted fluorescence is imaged
from above or below the sample, along an optical axis
perpendicular to the illumination plane. In an ideal case,
no fluorescence is generated above or below the plane
of illumination, meaning that PIM provides intrinsic
optical sectioning based on the first strategy described
above. The origins of this idea go back to 1903, when
planar illumination was used to study colloidal solutions
by scattered light imaging
6
. More recently, fluorescence
PIM was demonstrated in bioimaging applications
7–15
,
accumulating many names and acronyms.
Optical sectioning microscopy with planar
or structured illumination
Jerome Mertz
A key requirement for performing three-dimensional (3D) imaging using optical microscopes
is that they be capable of optical sectioning by distinguishing in-focus signal from out-
of-focus background. Common techniques for fluorescence optical sectioning are confocal
laser scanning microscopy and two-photon microscopy. But there is increasing interest in
alternative optical sectioning techniques, particularly for applications involving high speeds,
large fields of view or long-term imaging. In this Review, I examine two such techniques,
based on planar illumination or structured illumination. The goal is to describe the
advantages and disadvantages of these techniques.
Department of Biomedical Engineering, Boston University, Boston, Massachusetts, USA. Correspondence should be addressed to J.M.
(jmertz@bu.edu).
PUBLISHED ONLINE 29 SEPTEMBER 2011; DOI:10.1038/NMETH.1707
NATURE METHODS
|
VOL.8 NO.10
|
OCTOBER 2011
|
811
REVIEW
© 2011 Nature America, Inc. All rights reserved.© 2011 Nature America, Inc. All rights reserved.
Page 1
signal is given by K
1
= Dt
s
1P
I for one-
photon excitation and K
2
= Dt
s
2P
I
2
for
two-photon excitation, where I is the time-
averaged local intensity incident on the mol-
ecule while illuminated, Dt is the duration
the molecule is illuminated and
s
1P(2P)
is
the overall fluorescence cross-section of the
molecule (for two-photon excitation, this
includes any advantage factors related to the
pulsed nature of the excitation beam
1
). The
results are summarized in Table 1, where I
denoted the illumination plane thickness
(or axial resolution) as Dz and assumed that
the attenuation of the laser power owing to
scattering or absorption is negligible.
These results warrant a closer look. For
PIM, the local intensity scales as 1/N
y
when
using widefield illumination and is indepen-
dent of N
x
and N
y
when using line scanning. This is a feature of
side-on illumination. With head-on illumination, the local intensity
scales as 1/(N
x
N
y
) with widefield and 1/N
x
with line scanning. Given
that N
x
can be as large as a few orders of magnitude, such a scaling
law highlights a distinct intensity advantage of side-on over head-
on illumination. How this advantage influences the actual signal
depends on the particular version of PIM in question.
When using PIM with one-photon excitation (1P-PIM), the signal
K
1
increases by a factor R
1
= N
x
(Dx/Dz) relative to 1P-LSM. This
factor is the same regardless of whether one uses widefield or line-
scanning illumination. Because in the usual case of Gaussian optics
the span over which the illumination remains approximately flat is
itself governed by Dz (a thicker plane is flatter), R
1
1/NA
ill
(neglect-
ing prefactors of order unity), where NA
ill
is the numerical aper-
ture of the illumination optics. As approximate as this estimate is,
it points to the fact that the longer the extent of the illumination
plane along its axis (that is, the smaller the value of NA
ill
), the greater
the benefits of 1P-PIM. For example, illumination planes extend-
ing a millimeter or so can lead to signal gains typically one to two
orders in magnitude. The benefits of 1P-PIM are thus particularly
pronounced in microscopy applications with large FOVs
10
.
Although in principle pixel size (Dx) and plane thickness (Dz)
can be chosen independently in PIM, a fair comparison in terms of
resolution prescribes that the depth of field of the point-scanning
system should be roughly set equal to the axial resolution (or plane
thickness) of the light-sheet system. That is, Dx and Dz should be
chosen so that Dx/Dz NA
det
, where NA
det
is the numerical aperture
of the detection optics, leading to the prescription (NA
det
)
2
NA
ill
.
Whereas the above results suggest that 1P-PIM widefield and line-
scanning configurations are formally equivalent, one should also
weigh technical considerations such as better intensity uniformity
with line scanning versus the convenience of no moving parts with
widefield.
I turn now to PIM with two-photon excitation (2P-PIM). Both
widefield
16
and line-scanning
17,18
configurations have been report-
ed, but from Table 1 we observe that the scaling law advantage
derived from side-on illumination is mostly lost with widefield
illumination, whereas it is maintained with line-scanning illumina-
tion. That is, although widefield and line-scanning configurations
provide similar signals with 1P-PIM, the same cannot be said for
2P-PIM, where line scanning is preferable by far
17
. In the case of
Illumination strategies. There are two main classes of PIM (Fig. 1).
Both are based on widefield imaging with a digital camera but differ
in their illumination strategies. In a widefield illumination strategy,
the plane of illuminating light is static and spread over the entire
field of view (FOV)
9
. In a line-scanning illumination strategy, the
light plane is created by laterally sweeping a small diameter light
beam across the FOV, forming a full plane over the course of an
exposure
12
. Before examining the advantages and disadvantages
of these strategies, let us first revisit the more familiar point scan-
ning illumination strategy used in LSM, where laser light is directed
head-on into the sample and focused to a single point that is laterally
scanned in two dimensions. This LSM point-scanning strategy is
well known and will be useful for comparison. Signal detection in
LSM is performed with a single element detector, typically a photo-
multiplier tube (PMT) or avalanche photodiode (APD). As such,
pixel size in an LSM image, that is, the distance between consecutive
scan points in an image, can be chosen arbitrarily. In practice, pixel
size, when projected onto the sample plane, is usually chosen to be
roughly the same as the lateral spot size of the laser focus (to within a
Nyquist factor of 2). Let us call this projected pixel size Dx (assumed
square). Let us also assume that we wish to acquire an image com-
prising N
x
× N
y
such pixels (x being along the illumination axis), in
a total time T.
We start by using purely geometric arguments to compare some
efficiency advantages of PIM (widefield or line-scanning) with stan-
dard LSM (point-scanning), allowing the possibility of either one-
photon or two-photon excitation. Let us consider the total number
of signal photons produced by any given molecule over the course
of an image acquisition time, provided a fixed laser power P. This
ab
y
x
z
y
x
z
Δz
Δx Δx
Δz
N
y
Δ
xN
y
Δx
Figure 1 | PIM configurations. (a,b) PIM with widefield (a) and line-scanning (b) illumination.
Illumination light is shown in blue. Resulting fluorescence is collected with an objective. The detection
PSF (on axis only) is shown in green. The grid pattern in both panels corresponds to the N
x
× N
y
camera
pixels projected into the illumination plane.
Table 1 | Signal parameters
Local
intensity
Illumination
time Dt
K
1
(one-photon
excitation)
K
2
(two-photon
excitation)
PIM,
widefield
P
N
y
ΔxΔz
T
T
N
y
ΔxΔz
P
σ
1P
T
N
y
Δx
2
Δz
2
P
2
σ
2P
2
PIM,
line-scanning
P
Δz
2
T Δz
N
y
Δx
T
N
y
ΔxΔz
P
σ
1P
T
N
y
ΔxΔz
3
P
2
σ
2P
LSM,
point-scanning
P
Δx
2
T
N
x
N
y
T
N
x
N
y
Δx
2
P
σ
1P
T
N
x
N
y
Δx
4
P
2
σ
2P
812
|
VOL.8 NO.10
|
OCTOBER 2011
|
NATURE METHODS
REVIEW
© 2011 Nature America, Inc. All rights reserved.© 2011 Nature America, Inc. All rights reserved.
Page 2

You are reading a preview. Would you like to access the full-text?

Page 9
  • Source
    • "In addition, due to the specific shape of the optical transfer function, it is plagued by an important out-of-focus signal coming from the lowfrequency sample structures which deteriorates significantly the image contrast. Optical sectioning techniques, such as confocal microscopy, light sheet microscopy [2], and others [3,4] , ameliorate the image contrast but give little resolution improvement over brightfield . In contrast, structured illumination microscopy (SIM) improves both the image contrast and the transverse and axial resolutions [5] , but it requires careful control of the threedimensional excitation pattern which is not always possible in thick samples. "
    [Show abstract] [Hide abstract] ABSTRACT: We consider a fluorescence microscope in which several three-dimensional images of a sample are recorded for different speckle illuminations. We show, on synthetic data, that by summing the positive deconvolution of each speckle image, one obtains a sample reconstruction with axial and transverse resolutions that compare favorably to that of an ideal confocal microscope.
    Full-text · Article · Jun 2016 · Journal of the Optical Society of America A
  • Source
    • "Calcium imaging enables the study of activity in large populations of neurons with high sensitivity and high frame rates [1]. Camera-based imaging in wide-field (WF) mode can be fast, but imaging in thick specimen might require optical sectioning to suppress spurious out-offocus signals [2]. Since the brain is an inherently three-dimensional structure, such sectioning is essential for resolving neuronal signals and deciphering its function. "
    [Show abstract] [Hide abstract] ABSTRACT: Imaging intracellular calcium concentration via reporters that change their fluorescence properties upon binding of calcium, referred to as calcium imaging, has revolutionized our way to probe neuronal activity non-invasively. To reach neurons densely located deep in the tissue, optical sectioning at high rate of acquisition is necessary but difficult to achieve in a cost effective manner. Here we implement an accessible solution relying on HiLo microscopy to provide robust optical sectioning with a high frame rate in vivo. We show that large calcium signals can be recorded from dense neuronal populations at high acquisition rates. We quantify the optical sectioning capabilities and demonstrate the benefits of HiLo microscopy compared to wide-field microscopy for calcium imaging and 3D reconstruction. We apply HiLo microscopy to functional calcium imaging at 100 frames per second deep in biological tissues. This approach enables us to discriminate neuronal activity of motor neurons from different depths in the spinal cord of zebrafish embryos. We observe distinct time courses of calcium signals in somata and axons. We show that our method enables to remove large fluctuations of the background fluorescence. All together our setup can be implemented to provide efficient optical sectioning in vivo at low cost on a wide range of existing microscopes.
    Full-text · Article · Dec 2015 · PLoS ONE
  • Source
    • "Imaging thick samples under fluorescence microscopes is challenging because the emission light from the out-of-focus region of samples still leaks into the detection sensor, causing blurred images. In general, two strategies are used to solve the problem in general: do not generate out-of-focus background and reject the out-of-focus background [1] . Using the first strategy, lightsheet microscopy [2,3] employs a thin laminar sheet of light to illuminate layers of samples so that only the illuminated area is in the focus of the imaging optics, which is usually aligned along an optical axis perpendicular to the light sheet. "
    [Show abstract] [Hide abstract] ABSTRACT: In the interest of improving the temporal resolution for light-sheet microscopy, we designed a fast frame scanning camera system that incorporated a galvanometer scanning mirror into the imaging path of a home-built light-sheet microscope. This system transformed a temporal image sequence to a spatial one so that multiple images could be acquired during one exposure period. The improvement factor of the frame rate was dependent on the number of sub-images that could be tiled on the sensor without overlapping each other and was therefore a trade-off with the image size. As a demonstration, we achieved 960 frames/s (fps) on a CCD camera that was originally capable of recording images at only 30 fps (full frame). This allowed us to observe millisecond or sub-millisecond events with ordinary CCD cameras.
    Full-text · Article · Oct 2015 · Applied Optics
Show more