Enhanced Epifluorescence Microscopy by Uniform
and Intensity Optimized Illumination
S�ebastien Herbert,1 Ricardo Henriques1,2*
ALTHOUGH epifluorescence microscopy has been a key tool
in biological research for the last few decades, it is still ham-
pered by limitations imposed by the optical setups used. The
conventional nonhomogeneous light excitation produced
by most fluorescent microscopes modulates the observed fluo-
rescence and cripples its intensity quantification across the
sample field-of-view (FOV). We read with interest the recent
work of Coumans et al. featured in this issue of Cytometry
Part A (page 324). Here, a dual microlens array is employed to
flatten and optimize the intensity of the otherwise generally
nonuniform
illumination
in
epifluorescence
microscopes
equipped with mercury arc lamp light sources. This work
provides a key progress in the realm of existing methods to
solve the inconsistent illumination problem in microscopy
and demonstrates its potential application through the
improvement of cell detection such as in the case of circulating
tumor cells.
HOW NONHOMOGENEOUS ILLUMINATION
AFFECTS IMAGES?
By default, fluorescence emission is proportional to the
illumination intensity, except when fluorophore excitation is
saturated or photobleaching occurs. Quantification of the
fluorescence is perturbed when an uneven sample illumination
is applied. This is the case for most standard fluorescence
microscopy methods ranging from widefield epifluorescence,
confocal (1), to total-internal reflection fluorescence (TIRF)
microscopy (2). Without proper care, this problem will lead to
an incorrect image interpretation and analysis. Fluctuations in
the illumination profile will yield anomalous intensity read-
ings or even hide structures-of-interest in areas where the
illumination drops considerably, mostly at the edges of the
FOV. For example, take the case of different intensities
observed for objects with similar fluorophore concentrations
but at different regions of the irregularly illuminated FOV
(3,4) or the incorrect threshold-based segmentation of objects
whose intensities are altered by this unevenness.
Techniques based on the correlation between multiple
emission wavelengths, such as F€orster resonance energy
transfer (FRET), can suffer particularly from anomalous quan-
tifications caused by the irregular illumination. In FRET, the
interaction between fluorescently labeled molecules able to
undergo resonance energy transfer is calculated through the
analysis of their relative intensities (e.g., donor fluorophore
quenching and acceptor emission increase). Here, the ratio of
emission between both classes of fluorophores—FRET-donor
and FRET-acceptor—is corrupted in function of the irregular
illumination, especially when the illumination profile is differ-
ent for both channels.
CHOOSING BETWEEN KOEHLER OR SEMICRITICAL
ALIGNMENT
In epifluorescence microscopy, a compromise must be
made between either using Koehler illumination (the standard
method in microscopy) to minimize inhomogeneities in light
irradiation of the FOV (5) or a semicritical alignment of the
illumination where higher excitation intensities are achieved
but at the cost of a less homogeneous irradiation (Coumans
et al.). On the sample observation side, the microscopist
is faced with one of these two choices: the low-intensity in
1Institut Pasteur, Groupe Imagerie et Mod�elisation, CNRS URA 2582,
Paris 75015, France
2Unidade de Biofisica e Express~ao Genetica, Instituto de Medicina
Molecular, Faculdade de Medicina, Universidade de Lisboa, Lisboa,
Portugal
Received 13 October 2011; Accepted 8 November 2011
Grant sponsor: Institut Pasteur, The Fondation pour la Recherche
M�edicale, The Centre National de la Recherche Scientifique, and
Fundac¸~ao para a Ci^encia e a Tecnologia.
*Correspondence to: Ricardo Henriques, Institut Pasteur, Groupe
Imagerie et Mod�elisation, CNRS URA 2582, 25 rue du Docteur Roux,
Paris 75015, France.
E-mail: ricardoh@pasteur.fr
Published online 5 March 2012 in Wiley Online Library
(wileyonlinelibrary.com)
DOI: 10.1002/cyto.a.21177
© 2012 International Society for Advancement of Cytometry
Commentary
Cytometry Part A • 81A: 278�280, 2012

----!@#$NewPage!@#$----
Koehler illumination, in which longer exposure times will be
needed to achieve a usable signal-to-noise ratio (SNR), a
factor that will also impose restrictions on the maximum
frame-rate for time-lapse imaging, or the semicritical align-
ment, which will lead to an unwanted lower illumination at
the edges of the FOV increasing the overall anomalous varia-
tions in the fluorescence emission.
A POSTACQUISITION CORRECTION IS NOT IDEAL
Postacquisition image analysis methods can be applied to
flatten the irregular illumination present in both the Koehler
and semicritical alignment setups. The most common proto-
col involves a background subtraction followed by the division
of the image intensity values by a field mask estimated from
the nonuniform illumination pattern (3). However, this proce-
dure also increases inhomogeneously the noise and back-
ground profile of the image. The quality of this process is
highly dependent on the SNR observed and the gain linearity
of the light detector used, generally a charge-coupled device
(CCD) camera.
FINDING A SOLUTION FOR MERCURY ARC
LAMP-BASED EPIFLUORESCENCE MICROSCOPES
Coumans et al. propose a solution for epifluorescence
microscopes
that
use
mercury
arc
lamp
illumination.
Fluorescence excitation in these kinds of microscopes can be
especially tricky, as they tend to present a nonuniform illumi-
nation pattern broader in one of the lateral (XY-) axis com-
pared with the other. To reduce the illumination variability,
these types of setups have an FOV considerably smaller than
the illumination spot. This implies that an area greater than
the imaged region is irradiated, leading to a considerable waste
of light power in the FOV and inevitably to an excessive
unneeded and unwanted photobleaching and photodamage of
the sample. In this regard, we note that some epifluorescence
microscope setups provide a square aperture that prevents
illumination outside of the CCD’s FOV.
The proposed solution by Coumans et al. is able to yield
both a high intensity and highly uniform illumination. Here, a
dual microlens array is set on an optimized collimator and
inserted into the excitation path of the microscope, making
it
possible
to
segment
and
reorient
the
light
beam.
Consequently, it can be used to reshape the beam into a flat
rectangular pattern that matches the FOV shape of the CCD
detector. Furthermore, light rays otherwise hitting the exterior
region of the FOV are cleverly redirected thus minimizing light
leakage and unwanted photobleaching of nonimaged regions.
Readdressed rays will additively contribute to the illumination,
maximizing the intensity for the observed region while main-
taining a flat field.
Through the homogenization of the FOV illumination,
this work opens the door to an enhanced assessment and
quantification of fluorescence across images with confidence
that these values are not altered by the variation of the
excitation over the sample. Image analysis can be applied
globally allowing for a better quality of analytical studies,
implying an improvement to procedures such as thresholding,
segmentation, object counting, morphology, and intensity
quantification.
Moreover, both the stable illumination and reduction of
bleaching outside of the imaged area allow for a better stitch-
ing or tiling quality of neighboring images by reducing indivi-
dual-image edge artifacts. This permits their combination to
obtain a single larger image of concatenated FOVs, a feature
frequently used to observe samples larger than an individual
camera-acquired image.
Altogether, this method is a new key tool for researchers
and should greatly facilitate the gathering of fluorescence
microscopy information of large populations of cells, tissues,
or organisms whose examination implies the observation,
processing, and correlation of numerous epifluorescence
images.
CONCLUSIONS AND OUTLOOK
Over the last decades, we have seen an increased aware-
ness into the obstacles imposed by nonuniform illumination
and a growth of tools to challenge them. Several other works
have described methods for illumination homogenization in
the FOV area through engineering of the optical setup (6–10).
Yet, the illumination problem is not restricted to two-dimen-
sions but also to three-dimensional space and/or time (1).
Achieving a steady illumination in time over the experimental
image acquisition is mostly dependent on the stability of the
light source used. Although postacquisition image processing
steps can reduce time-dependent illumination anomalies, the
process alters the true quantitative nature of the data. Keeping
a uniform illumination profile at increasing image depths also
becomes complex as light scatters through the multiple struc-
tures of the sample. Currently, tissue-clearing protocols that
use reagents such as methyl salicylate or benzyl alcohol–benzyl
benzoate increase sample transparency, thus diminishing this
problem (11). However, these procedures are not compatible
with live cell imaging, causing sample shrinkage and can dis-
solve labeled fluorescence probes (resulting in weak detection
signals).
With the growth of adaptive optics methods, such as the
use of spatial light modulators (SLMs), we now enter a new
period where it is possible to dynamically adapt the sample
illumination to the irregular shape of the sample itself or
to heterogeneously structures of interest before imaging (12).
These methods should lead to novel ways to improve
sample illumination and visualization in three-dimensional
space plus time but at the expense of a complex computa-
tional estimation of the optimal light distribution over the
sample.
Overall the work of Coumans et al. provides a simple, sta-
tic, and stable solution for some of the most common epifluo-
rescence microscopes currently used.
ACKNOWLEDGMENTS
The authors thank J. Oreopoulos, G. Martins, and C.
Zimmer for advice and critical reading of the manuscript.
COMMENTARY
Cytometry Part A • 81A: 278�280, 2012
279

----!@#$NewPage!@#$----
LITERATURE CITED
1.
Murray JM, Appleton PL, Swedlow JR, Waters JC. Evaluating performance in three-
dimensional fluorescence microscopy. J Microsc 2007;228:390–405.
2.
Fiolka R, Belyaev Y, Ewers H, Stemmer A. Even illumination in total internal reflec-
tion fluorescence microscopy using laser light. Microsc Res Tech 2008;71:45–50.
3.
Wolf DE, Samarasekera C, Swedlow JR. Quantitative analysis of digital microscope
images. Methods Cell Biol 2007;81:365–396.
4.
Waters JC. Accuracy and precision in quantitative fluorescence microscopy. J Cell
Biol 2009;185:1135.
5.
Salmon ED, Canman JC. Proper alignment and adjustment of the light microscope.
Curr Protoc Cell Biol 2001; Chapter 4: Unit 4.1. http://dx.doi.org/10.1002/04711
43030.cb0401s00.
6.
Hoffnagle JA, Jefferson CM. Design and performance of a refractive optical system
that converts a Gaussian to a flattop beam. Appl Opt 2000;39:5488–5499.
7.
Wippermann F, Zeitner UD, Dannberg P, Bra€uer A, Sinzinger S. Beam homogenizers
based on chirped microlens arrays. Opt Express 2007;15:6218–6231.
8.
Nemoto K, Fujii T, Goto N, Takino H, Kobayashi T, Shibata N, Yamamura K, Mori Y.
Laser beam intensity profile transformation with a fabricated mirror. Appl Opt
1997;36:551–557.
9.
Chang SP, Kuo JM, Lee YP, Lu CM, Ling KJ. Transformation of Gaussian to
coherent uniform beams by inverse-Gaussian transmittive filters. Appl Opt 1998;37:
747–752.
10. Sales TRM. Structured microlens arrays for beam shaping. Opt Eng 2003;42:3084.
11. Zucker RM, Hunter ES, Rogers JM. Confocal laser scanning microscopy of morphol-
ogy and apoptosis in organogenesis-stage mouse embryos. Methods Mol Biol
2000;135:191–202.
12. Caarls W, Rieger B, De Vries AH, Arndt-Jovin DJ, Jovin TM. Minimizing light
exposure with the programmable array microscope. J Microsc 2011;241:101–
110.
COMMENTARY
280
Enhanced Epifluorescence Microscopy