G Model

BIOS 8996 114 BIOS : 8996 Model5G pp:114col:fig: : NIL

Biosensors and Bioelectronics ()

Contents lists available at ScienceDirect

Biosensors and Bioelectronics

journal homepage: www.elsevier.com/locate/bios

A review of recent progress in lens-free imaging and sensing

Mohendra Roy a,b, Dongmin Seo a, Sangwoo Oh a,c, Ji-Woon Yang a, Sungkyu Seo a,n
a
Department of Electronics and Information Engineering, Korea University, Sejong, Republic of Korea
b
Department of Physics, Rajiv Ghandi University, Rono Hills, Arunachal Pradesh, India
c
Korea Research Institute of Ships & Ocean Engineering, Daejeon, Republic of Korea

art ic l e i nf o a b s t r a c t

Article history: Recently, lens-free imaging has evolved as an alternative imaging technology. The key advantages of this
Received 7 June 2016 technology, including simplicity, compactness, low cost, and exibility of integration with other com-
Received in revised form ponents, have facilitated the realization of many innovative applications, especially, in the elds of the
27 July 2016
on-chip lens-free imaging and sensing. In this review, we discuss the development of lens-free imaging,
Accepted 31 July 2016
from theory to applications. This article includes the working principle of lens-free digital inline holo-
graphy (DIH) with coherent and semi coherent light, on-chip lens-free uorescence imaging and sensing,
Keywords: lens-free on-chip tomography, lens-free on-chip gigapixel nanoscopy, detection of nanoparticles using
Lens-free imaging and sensing on-chip microscopy, wide eld microscopy, and lens-free shadow image based point-of-care systems.
From theory to applications
Additionally, this review also discusses the lens-free uorescent imaging and its dependence on structure
Shadow images
and optical design, the advantage of using the compact lens-free driven equilibrium Fourier transform
Diffraction patterns
CMOS image sensor (DEFT) resolved imaging technique for on-chip tomography, the pixel super-resolved algorithm for
gigapixel imaging, and the lens-free technology for point-of-care applications. All these low-cost, com-
pact, and fast-processing lens-free imaging and sensing techniques may play a crucial role especially in
the elds of environmental, pharmaceutical, biological, and clinical applications of the resource-limited
settings.
& 2016 Elsevier B.V. All rights reserved.

1. Introduction of health care in resource-limited countries has been an ambitious

In recent years, the biomedical electronics industry has un-
dergone a remarkable evolution, bringing compact and low-cost
wearable devices to consumers. This evolution also triggered the
development of various advanced sensing and characterization
systems not just for academic research, but also for industries. The
advances in computers, such as high-performance graphics pro-
cessing units (GPUs) and central processing units (CPUs), fa-
cilitated the implementation of computational methods for fast
characterization and prediction, thus decreasing their complexity
and dependence on complex hardware components. This ad-
vancement in computational resources provided an opportunity to
develop advanced biomedical imaging techniques such as com-
putational tomography (CT) and X-ray imaging (Coskun and Oz-
can, 2014). In addition, considerable effort using these advanced
electronics and computational methods has gone into developing
portable point-of-care devices. The design and development of a
point-of-care diagnostic instrument that can improve the delivery
n
Correspondence to: Rm #205, Science & Technology Bldg. II, 2511 Sejong-ro,
Sejong 30019, Republic of Korea.
E-mail address: sseo@korea.ac.kr (S. Seo).
goal. Such an instrument will improve health care by reducing the
overall cost of testing and providing early diagnosis. The recent
development of miniaturization technology such as 20-nm feature
size fabrication and nano electromechanical systems (NEMS) has
fueled research in and development of compact and integrated
devices with very small form factors that are used on these kinds
of diagnosis devices. These developments in semiconductor en-
gineering lead to the compact and integrated optoelectronic pro-
ducts, including complementary metal oxide semiconductor
(CMOS) and charge-coupled device (CCD) sensors, for use in
imaging. These high-resolution sensors, along with the traditional
optical components such as lenses, can acquire high-resolution
microscopic images (Kim et al., 2012). However, the optical com-
ponents are large and expensive, which in turn makes the whole
system bulky and costly. Interestingly, due to the high pixel re-
solution, the image sensors can acquire diffraction patterns of
microsamples without the use of any lenses (Gurkan et al., 2011).
Recently, some research groups have tried to take advantage of
these high-resolution sensors to fabricate a compact and lensless
microscope (Kim et al., 2011). Signicant progress has been made
to achieve alternative imaging by integrating lens-free technology
with the advances in computational and image-processing algo-
rithms (Gorocs and Ozcan, 2013). The components used for this

http://dx.doi.org/10.1016/j.bios.2016.07.115
0956-5663/& 2016 Elsevier B.V. All rights reserved.

Please cite this article as: Roy, M., et al., Biosensors and Bioelectronics (2016), http://dx.doi.org/10.1016/j.bios.2016.07.115i
2 M. Roy et al. / Biosensors and Bioelectronics ()
Fig. 1. Schematic of a lens-free imaging system showing the simplicity of the
platform (Gorocs and Ozcan, 2013).
purpose, such as light-emitting diodes (LEDs) and CMOS image
sensors, are low cost, lightweight, and easy to integrate (see Fig. 1),
which makes the system simple. Several groups are working to
develop various applications that take advantage of the lens-free
imaging platform (Zhu et al., 2013a, 2013b). In this article, we
discuss the development of lens-free imaging technology and its
various applications.
2. Theory behind lens-free imaging and sensing
The basic principle of lens-free imaging can be explained using
digital inline holography (DIH), a two-step process in which the
amplitude and phase information of the wave front that originates
from the object are digitally recorded and then reconstructed
using computational algorithms (Gorocs et al., 2013). Digital inline
holography became popular during the 1990s (Picart and Leval,
2008). Since then, its importance in applications because of its
ability to image transparent objects and analyze images easily
through mathematical morphology has been demonstrated.
In DIH, the hologram of an object is formed by a spherical wave
of illumination of wavelength emitted from a point source about
the size of the wavelength. This spherical wave splits into a scat-
tered wave and an unscattered reference wave when it encounters
an object (Fig. 2), creating interference patterns (hologram). This
event can be expressed as Ascatr(r)A(r)Aref(r), where A(r) is the
amplitude of the incident wave, Ascatr(r) is the amplitude of the
scattered wave, and Aref(r) is the amplitude of the unscattered
reference wave. Thus, the contrast of the hologram is dened as
2 2
I ( r ) = A ref ( r ) + Ascatr ( r ) A ref ( r )
I ( r ) = A ref * ( r ) Ascatr ( r ) 2
* ( r )Ascatr ( r ) + A ref ( r )Ascatr
The rst term in Eq. (2), Aref * ( r ), is the
* ( r )Ascatr ( r ) + Aref ( r )Ascatr
holographic diffraction pattern generated by the superposition of
the reference wave directly from the source on the scattered wave
2
from the object, and Ascatr ( r ) is the classical diffraction pattern
generated by the interference of the scattered wave only (Xu et al.,
2002). Because a hologram results from the interference between
scattered and reference waves, the hologram of a transparent
object can be created with DIH.
In general, the phase and amplitude data of the hologram are
Please cite this article as: Roy, M., et al., Biosensors and Bioelectronics (2016), http://dx.doi.org/10.1016/j.bios.2016.07.115i
Fig. 2. Schematic of DIH. A coherent wave from a point source strikes an object and
forms a scattered wave, which, in turn, forms a diffraction pattern of the object
because of interference with the reference light.
recorded digitally by image sensors such as CMOSs or CCDs. These
data are used in the numerical reconstruction process with the
help of a computational algorithm. Reconstruction consists of
virtually illuminating the recorded holographic pattern with a
reference wave and evaluating the complex eld distribution. The
rst DIH reconstruction was reported by Goodman and Lawrence
in 1967 (Gorocs and Ozcan, 2013). However, the true revolution in
holography began with the use of modern digital imaging sensors
(i.e., CCDs and CMOSs) by Schnars and Jptner (1994). In general,
most current numerical reconstruction methods are based on the
famous KirchhoffHelmholtz transform, which can be described by
the following equation:
K (r ) = s dsI( )exp(2ir/ ) (3)
where ( x, y, L ) is the 3D coordinate, L is the distance of the X,Y
screen from the source, and I ( ) is the contrast of the hologram.
This integration yields the reconstructed information of the object.
Plotting K (r ) in an X,Y plane yields the 2D holographic re-
constructed image, and stacking the 2D hologram yields the 3D
information. The detailed steps of DIH are illustrated in Fig. 3.
Although the reconstruction provides high-precision surface
topography of objects, artifacts associated with the acquisition
process, such as twin images due to the zero-order term and
complex conjugates of the object wave, make the reconstruction
process a challenging one because the image sensor measures only
(1)
the intensity prole of these waves. However, these kinds of ar-
tifacts can be eliminated with phase retrieval algorithms, but they
(2) make the system complex and place more demand on computa-
tional resources. This is useful for applications in which high-
contrast topography of an object is required. Fortunately, common
biological applications such as gathering information on the con-
centration or population of cells, determining cell size, and dif-
ferentiating cell types do not require high-contrast topography.
Therefore, research on lens-free shadow imaging with partially
coherent planar wave illumination is ongoing. In this method, the
sample-to-sensor distance is smaller than the source-to-sample
distance. Fig. 4 presents a schematic of this approach. Because of
this distance relationship, the reference wave can be considered a
 M. Roy et al. / Biosensors and Bioelectronics ()
Fig. 3. Flowchart of the formation of images using DIH (Picart and Leval, 2008).
plane wave. This eliminates the aliasing effects associated with the
spherical wave and reduces the spatial and temporal coherence of
the illumination because the reduced difference between the
paths of the scattered and reference waves facilitated the use of a
common semicoherent light source such as an LED conjugated
with a pinhole. Because the sample-to-sensor distance is much
smaller than the source-to-sample distance, the cross interference
among the scattered waves is reduced signicantly, which, in turn,
leads to a reduction in the noise and spatial magnication of the
image. This reduced spatial magnication allows the system to use
the full active area of the sensor more efciently; however, this
reduces the spatial resolution because of the limited pixel size of
the detector (Bishara et al., 2010, 2011a; Mudanyali et al., 2010;
Seo et al., 2010; Isikman et al., 2011a; Greenbaum et al., 2012a,
2012b). The limited pixel size can be resolved by implementing the
super-resolution method in which several pictures taken by sub-
pixel shifting of the sample are stitched together using a
Fig. 4. Schematic of lens-free shadow imaging with partially coherent planar wave illumination. (a) Experimental setup used by Roy et al., (2014, 2015). (b) Generation of
diffraction pattern of a micro-object as described by Roy et al., (2014, 2015).
Please cite this article as: Roy, M., et al., Biosensors and Bioelectronics (2016), http://dx.doi.org/10.1016/j.bios.2016.07.115i
3
computational algorithm (Bishara et al., 2010). However, this
method is time-consuming and assembling the images puts great
demand on computational resources. Therefore, many research
groups are working on a cost-effective and compact system that
can directly process the shadow images. Recently, Roy et al. de-
monstrated the application of such a system for obtaining blood
cells counts, white blood cell differentiation, and determination of
the size of micro-objects by directly characterizing the shadow
parameters (Roy et al., 2014, 2015). An overall review of the lens-
free imaging was discussed by the Greenbaum et al. In this paper
they discussed the various categories of lens-free imaging, spe-
cially contact mode shadow imaging and diffraction based lens-
free imaging. A concise discussion of advantages, limitation of all
these techniques and their future challenges were carried out,
specially the limitation of spatial resolution, overlapping due to
sample density and etc. They also discussed the need of standar-
dization of remonstration methods and limitation of the on-chip
lens-free uorescence imaging, i.e. the limitation of the spatial and
temporal coherence of the uorescence light and the effect of the
thickness of the lters (Greenbaum et al., 2012b). However, due to
the growing application of lens-free imaging over time, a more
detail discussion on some of the specic applications of lens-free
imaging is desirable. In the following sections, we discuss the
development of lens-free imaging and its various applications.
3. Applications of lens-free imaging and sensing
3.1. Lens-free on-chip platforms
In this section, we discuss the development of lens-free ima-
ging in an on-chip imaging system. Although the discovery of lens-
free DIH was reported in the 1990s, the rst on-chip application
was in 2008 when it was used for automated detection and
characterization of microparticles on a microuidic chip that was
integrated on a lens-free imaging system (Ozcan and Demirci,
2008). The system, which consisted of a CCD image sensor with an
active area of 37.25 mm  25.70 mm, was very compact and had a
wider eld of view (FOV) than a conventional imaging system. The
working principle of this system is similar to that of lens-free
imaging with a partially coherent planar wave, as discussed in the
previous section. The goal was a compact optimized alternative
4 M. Roy et al. / Biosensors and Bioelectronics ()

Fig. 5. (a) Experimental apparatus (under blue light) and (b) schematic diagram of the holographic LUCAS (Lensless Ultra-wide-eld Cell monitoring Array platform based on
Shadow imaging) platform.

imaging system that is comparable to existing systems with re- 420 samples. Therefore,
spect to the information provided. To achieve this goal, various DevDevMIN
parameters such as the dependence of the system on the wave- Corr=1
DevMAX DevMIN
length of illumination, the effect of variation in object size, and the
variation in the error rate of the change in object concentration, where DevMAX and DevMIN are the maximum and minimum devia-
were optimized. This optimized system was used to automatically tion of f(x,y) calculated using the individual library images that
detect biological cells such as NIH-3T3m AML-12 hepatocytes and make up L(x,y).
CD4 T lymphocytes. The processed results obtained with this algorithm show that
Seo et al. further improved this on-chip imaging by introducing different signatures for different type of cells can be obtained, as
the ability to differentiate cell types with lens-free imaging (Seo shown in the Fig. 6, which shows that the diffraction patterns of red
et al., 2009). The system had a higher-resolution sensor with blood cells (RBC), a yeast cell, and polystyrene microbeads have
smaller pixels (2.2 mm), and the sample plane was illuminated signicant signal differences. However, cells with similar optical
with tunable monochromatic light, as shown in the Fig. 5. The properties but different work functions are very difcult to differ-
shadow images from this system were autocharacterized by a entiate with the lens-free imaging system, a problem recently ad-
custom algorithm developed based on the parameters of the dif- dressed by Wei et al. (2013). They implemented an optical mod-
fraction patterns of the samples, including digital signal-to-noise ulation method based on the plasmonic resonance effect produced
ratio (SNR), shadow radius (RRMS), and correlation index (Corr), by metallic nanoparticles (Schultz, 2003). Plasmonic nanoparticles
which are represented as follows: are being used as a contrast agent because of their unique optical
modulation property. Wei et al. used gold (Au) and silver (Ag) na-
3.1.1. Signal to noise ratio noparticles as contrast agents to label the CD4 and CD8 cells. This
produced distinguishable diffraction signatures in the lens-free
image, which could then be characterized using principle compo-
nent analysis (PCA) and multiclass support vector machine (SVM)
SNR = max( I ) b /b analysis on the multispectral data of the lens-free images. Their
system consisted of a CMOS image sensor with an active area of
where I is the intensity of the light on the sensor array and mb
24 mm2 and a monochromatic light source. Images of all samples
and sb are the mean and variance of the background noise region.
were taken with this setup in the wavelength range of 480900 nm
in 10-nm steps. Fig. 7(a) shows a schematic of the system.
3.1.2. Shadow radius (Rrms)
The lens-free system also has been used in the study of the
movement of microsubstances. Su et al. demonstrated the feasi-
bility of using the lens-free imaging system for analyzing the ac-
w 2 2
w 2 1/2 tivity of human sperm (Su et al., 2010a), a major parameter in the
x=1 (
R rms = (
x x ) f x, y = y0 ) /x = 1 f ( x, y = y0 ) evaluation of male fertility. Su et al. used lens-free technology to
develop a portable sperm activity monitoring system that cap-
where W is the maximum number of pixels in the region of tured shadow images of sperm at 20 frames per second. The
interest (ROI) and images were normalized to their mean intensity and digitally
w 2 w 2 summed to quantify the immotile sperm. To identify motion of
x= x=1 x (
f x, y = y0 ) /x = 1 f ( x, y = y0 ) motile sperm, a shadow image was subtracted from its subsequent
frame. A reconstruction algorithm then was used to reconstruct
Here, ( x, y ) is the index of the image pixel and f x, y = y0 is ( ) the subtracted images. The distance between the dark and bright
the detected intensity prole of a line represented by y = y0.
spots in the reconstructed image was the displacement of the
sperm, indicating movement (Fig. 8). The speed and trajectory of
3.1.3. Correlation index
the motile sperm can be determined from this relative distance. In
addition, Cui et al. (2008) demonstrated an on-chip optouidic
microscope based on lens-free imaging technology, which pro-
Dev = ( x, y) ROI f ( x , y) L ( x , y) mises the point of care analysis by facilitating the large scale
scanning of cell samples through the microuidics (Cui et al.,
where L( x, y ) is the mean library image formed by averaging 2008). Again Zhang et al. (2011) had demonstrated an on-chip

Please cite this article as: Roy, M., et al., Biosensors and Bioelectronics (2016), http://dx.doi.org/10.1016/j.bios.2016.07.115i
 M. Roy et al. / Biosensors and Bioelectronics () 5

Fig. 6. (a) and (b) illustrate the difference in performance between conventional LUCAS (with incoherent light) and holographic LUCAS (with coherent light). A high-
resolution CMOS sensor array (2.2-mm pixel) was used in each image. (c) and (d) show the cross-sectional intensity proles of the various micro-objects imaged in (a) and
(b), respectively. Due to increased spatial coherence, the holographic diffraction pattern of each micro-object exhibits more information on texture, with unique oscillating
features that contain phase information of the cell or microparticle. This phase information, usually lost during incoherent illumination, provides magnied images of the
holographic diffraction signatures of 10-mm beads, yeast cells (S. pombe), and RBCs. (e)(j) Microscope images of the same FOV, acquired with a 10-mm objective lens, are
shown next to the holographic LUCAS images of the three microparticles for comparison (Seo et al., 2009).

sperm motion characterization system. This enables the on-chip
monitoring of the sperm health in a wide eld of view, which
provides statistically more accurate result compare to low
throughput conventional devices (Zhang et al., 2011).
3.2. Lens-free uorescence microscopy
Fluorescent imaging is an important method for characterizing
cells at the molecular level. A uorescent imaging system with a
large FOV is desirable for the best characterization. Because a
conventional uorescent imaging system has a small FOV, it was
thought that a lens-free imaging system integrated with a uor-
escent imaging system would provide better cellular character-
ization. Recently, Coskun et al. (2010) demonstrated such a system.
They placed a prism on top of the sample plane to facilitate total
internal reection, which is the excitation light for the uorescent
samples. The emitted uorescent light is ltered by an absorption
lter and then recorded with a CCD or CMOS sensor. Fig. 9 shows
the schematic of this experimental setup. The advantage of this
system is that it can take uorescent and corresponding holo-
graphic images simultaneously with an FOV of 2.5 cm  3.5 cm.
The feasibility of the system was tested by imaging a hetero-
geneous sample of 10-mm uorescent particles and 2030-mm
microparticles and a uorescent sample by imaging a hetero-
geneous sample of uorescent and nonuorescent particles using
the vertical transmission imaging system depicted in Fig. 4. The
results in Fig. 11 show that the system can capture the uores-
cence and hologram images simultaneously in an 8-cm2 FOV,
making the system a unique portable uorescence detector.
The spatial resolution of this platform was improved to
 10 mm with the implementation of a compressive sensing algo-
rithm (Coskun et al., 2010). The system used for this purpose was
almost identical to that described above, except for the introduc-
tion of a ber optic faceplate in between the coverglass and the
absorption lter, as shown in Fig. 12. This lens-free system sig-
nicantly enlarges and distorts the uorescence image because of
diffraction and the introduction of the faceplate, but the distortion
is eliminated by using a compressive sensing algorithm, which is
described as follows.
The uorescent particle/cell distribution within the sample
volume is c = c1, c2, c3..... , cN , where N is the number of voxels,
and the physical grid size in c is d. In addition, c is sparse and only
the S coefcients of c are nonzero such that S { N. The intensity
distribution of light impinging on the detector array is determined
by c , and that above the detector plane is expressed as
N

non-uorescent particles. The 470-nm wavelength excitation light f ( x , y) = cii( x, y)

was introduced through the side walls of the prism. A detected
image is shown in Fig. 10(a). Fig. 10(b) is the same image with
improved resolution after reconstructing the image using the ac-
celerated LucyRichardson algorithm.
Coskun et al. (2010) also demonstrated the dual imaging of
i=1
where i( x, y ) is the intensity of a 2D wave immediately before the
detector. i( x, y ) can be measured for each object plane from the
uorescent particles. Without the use of a faceplate, i( x, y ) for a
given object plane becomes space invariant and is expressed as

Please cite this article as: Roy, M., et al., Biosensors and Bioelectronics (2016), http://dx.doi.org/10.1016/j.bios.2016.07.115i
6 M. Roy et al. / Biosensors and Bioelectronics ()

Fig. 7. (a) Schematic illustration of a multispectral lens-free in-line holographic system. (b) Full FOV of a lens-free hologram of Au-CD4 cells excited at 560 nm; the center
region of the white box is shown in (g). (c) and (f) Dark-eld scattering microscopy of CD4 and Au-CD4 cells, respectively, obtained by an objective lens. (d) and (g) are lens-
free super-resolved (SR) holograms and (e) and (h) are reconstructed amplitude images of the same regions of interest (ROIs) as the dark-eld scattering images in (c) and (f).
Scale bar 25 mm for c, d, e. Scale bar 10 mm for f, g, h (Wei et al., 2013).

( )
i( x, y ) = p x x i , y yi , where p( x, y ) is the incoherent point The intensity sampled by the detector is
spread function (PSF) of the system for a given object layer and ( ) (
Im = f ( x, y ) x x m, y ym dxdy , where x x m, y ym dxdy )
( )
x i , yi indicates the physical location of ci . Therefore, the general represents the sampling or measurement bases. Here, m 1:M
equation for lens-free imaging with or without the presence of a denotes the mth pixel of the detector array with center co-
faceplate can be written as ( )
ordinates of x m, ym , and ( x, y ) is the pixel function.
For better decoding, there should be good spatial correlation
N between m and i for all possible m 1:M and i1:N pairs. Good
f ( x , y) = cip( x xi , y yi ) spatial correlation provides the incoherence between the sampling
i=1 and representation bases, which in turn provides the probability of

Fig. 8. (a) Digitally subtracted lens-free hologram of three moving sperm is generated from two successive frames (500 ms apart). (b) Microscopic image, digitally re-
constructed from the lens-free differential hologram shown in (a), shows the positions of three sperm in the two successive frames, where white spots indicate the nal
positions and the black spots indicate the starting positions. S1, S2, and S3 are the displacement vectors of these sperm (Su et al., 2010a).

Please cite this article as: Roy, M., et al., Biosensors and Bioelectronics (2016), http://dx.doi.org/10.1016/j.bios.2016.07.115i
 M. Roy et al. / Biosensors and Bioelectronics ()
Fig. 9. Schematic diagram of the lens-free on-chip uorescent imaging platform
over a large FOV of 2.5 cm  3.5 cm. Fluorescent excitation is achieved via illumi-
nation through the side of a rhomboid prism (if more convenient, a different prism
geometry could be used). A simple LED or a Xenon lamp tuned by a mono-
chromator is used for excitation. Lens-free holographic imaging of the same FOV is
achieved with vertical incoherent illumination (another LED) through the at top of
the prism. Drawing is not to scale. Dimensions: prism height p, 17 mm; active area
of the imager, w1  w2, 25 mm  35 mm; depth of the solution reservoir, k,  10
100 mm; distance of the vertical source, h,  510 cm; distance of the uorescent
excitation source, f,  12 cm. Not shown here, an index-matching gel can be used
to avoid total internal reection and undesired scattering at the bottom facet of the
prism. In addition, to better control the vertical distance between the sample mi-
crochannel and the active region of the sensor, we removed the protective cover-
glass of the chip. The thin absorption lter above the CCD/CMOS acts as a protective
layer in this case, isolating the active region of the sensor chip from the micro-
channels (Coskun et al., 2010).
accurately reconstructing c from M measurements. The coherence
is obtained by calculating the correlation between the pixel
function ( x, y ) and the PSF p( x, y ). The smaller the correlation,
the better the compressive decoding is.
In conclusion, compressive sampling digitally counters the ef-
fect of diffraction induced by
N N
f ( x , y) = cii( x, y) and f ( x, y) = cip( x xi , y yi )
i=1 i=1
through decoding of the lens-free image pixels expressed in
( )
Im = f ( x, y ) x x m, y ym dxdy .
Fig. 13 presents the results from the compressive decoder method
and the conventional LucyRichardson deconvolution method for
comparison. The former technique reportedly improved the spatial
resolution of the lens-free imaging system to about 10 mm.
Despite the simplicity of lens-free imaging, it does have some
limitations, including a lack of resolution due to limited temporal
coherence of the incident light. In most lens-free imaging systems,
the light sources are either partially coherent light (in the case of
holography) or incoherent light (particularly in case of uorescent
objects), implemented mainly by LEDs, which are limited in tem-
poral coherence. In addition, in uorescent imaging, the light from
uorescent particles spatially overlaps. However, most traditional
reconstruction algorithms are single-wavelength procedures, which
limits the effective bandwidth of the light emitted by the uor-
escent particles. Sencan et al. (Sencan et al., 2014) addressed this
problem by introducing the spectral demultiplexing method in
which the polychromatic phase retrieval algorithm is used to de-
multiplex the superimposed spectral components of the holograms
generated with broadband LED illumination. For that, they used
super-resolved polychromatic intensity measurements. The multi-
color lens-free uorescent signatures are simultaneously demulti-
plexed using a sparse recovery algorithm with a PSF at a different
wavelength and the spectral response curve of the sensor. The
Please cite this article as: Roy, M., et al., Biosensors and Bioelectronics (2016), http://dx.doi.org/10.1016/j.bios.2016.07.115i
7
measured uorescent intensities from multiple isolated particles
are interpolated, aligned, and averaged. The intensity proles de-
pend on the structure and optical design of the sensor pixel, lter
response, and distance between the object and detector planes.
Sencan et al. demonstrated the demultiplexing method by
using it for on-chip lens-free uorescent imaging with various RGB
uorescent emitters; the results are shown in Fig. 14.
3.3. Lens-free tomography
Structural information on biological cells and microparticles
provides deep insight into cell morphology. Several approaches such
as confocal microscopy and optical coherence tomography are
available for obtaining 3D information on microparticles. However,
these techniques use bulky technology and are time-consuming. The
increase in biological research requires a structural analysis system
that is fast and compact. Recently, lens-free technology has been
studied as a potential alternative for 3D imaging (Fig. 15). Lens-free
technology makes the system simpler, compact, and lightweight. The
working principle of this technology for 3D imaging is similar to that
of conventional lens-free technology with the addition of the ability
for multiple-angle illumination. The reconstructed images from this
system provide 3D information on micro-objects. Recently, Su et al.
demonstrated multiangle lensless digital holography for depth-re-
solved imaging on a chip (Su et al., 2010b).
In the multiangle illumination method, the height of a micro-
object is determined by calculating the lateral shift of the holo-
gram of the object as function of the illumination angle. The ad-
vantage of the compact, lens-free, driven equilibrium Fourier
transform (DEFT) resolve imaging technique is that its FOV is
wider than that of the conventional DEFT resolve imaging tech-
nique. The compactness of this system facilitates imaging on
miniaturized platforms such as lab-on-a-chip. Recently, Bishara
et al. demonstrated the integration of lens-free tomography with a
microuidic channel (Bishara et al., 2011b). Isikman et al., who
presented a summary of a eld-portable benchtop lens-free op-
tical tomographic microscope, further explained this integration
(Isikman et al., 2011b). They also demonstrated sectional imaging
of a large volume (20 mm3) with a spatial resolution of o7 mm
(Isikman et al., 2011a), providing high-throughput images with a
long depth of eld. These systems provide better 3D information
on the microscopic world at minimal cost. The main advantage of
this kind lens-free tomography system is that the spatial resolu-
tion can be improved without compromising the FOV, which is a
trade-off encountered with lens-based microscopy systems.
3.4. Lens-free gigapixel nanoscopy
Nanoscale imaging using an optical imaging system has been a
challenge due to its diffraction limit, which research groups are trying
to overcome. In this section, we discuss some of the new approaches
developed to address this issue that are based on lens-free imaging.
The gigapixel (109 pixels) is the resolution of images obtained
with the lens-free technique, in which multiple images of the
holograms produced by the objects are captured by slightly
shifting the image plane. In general, the shift is less than the size of
the pixel. The images are then fused into to a single high-resolu-
tion image using a pixel super-resolved algorithm and an iterative
gradient algorithm (McLeod et al., 2013). The super-resolution
images are then reconstructed via back-propagation of the object
plane through an angular spectrum approach. Euan et al. pre-
sented a study on recent developments in lens-free imaging to-
ward the goal of high-resolution imaging (McLeod et al., 2013).
In the review by McLeod et al., they discussed methods to
improve the effective numerical aperture (NA) of lens-free imaging
to obtain nano-range resolution, including iterative error-
8 M. Roy et al. / Biosensors and Bioelectronics ()

Fig. 10. (a) Lens-free uorescent imaging of 10-mm uorescent beads over a FOV of 4 8 cm2 (excitation/emission: 495 nm/505 nm). (a) Raw lens-free image pumped
through an LED and acquired with an integration time of 1 s (b) Digitally deconvolved uorescent image of the same FOV as (a). (c)(e) From left to right: Magnied image of
the indicated area in (a), the result of the deconvolution process after 100 iterations, and after 600 iterations. The letters f, g, and h and dashed lines in (c) and (d) refer to
panels (f)(h), which illustrate the different cross sections of the raw and deconvolved uorescent images, demonstrating the  5  improvement in uorescent spot size.
Through the iterative deconvolution process, two particles that almost completely overlap in the raw lens-free image [see, for instance, (c) and (d)], are now separated, as
shown to the right of (c) and (d) and in (f)(h). The pixel size of the CCD in this experiment was 9 mm. The resolution of the deconvolved lens-free uorescent image could
have been further improved with the use of a smaller pixel (Coskun et al., 2010).

reduction algorithms, smaller pixel size of the image sensor, and
self-assembly. Self-assembly is an effective way to overcome the
limitation of optical microscope resolution and achieve nanor-
esolution imaging. In this technique, micro- or nanolenses are
fabricated by naturally adopting the conguration of individually
dispersed components over target nano-objects. This greatly in-
creases the scattering cross section of the target objects, which, in
turn, improves the amplitude of the scattered waves from the
objects. Euan et al. discussed the self-assembly process and the
integration of self-assembly technology with lens-free imaging
technology in detail (McLeod and Ozcan, 2014). This combination
has the advantages of the wide FOV of lens-free imaging and the
improved resolution of self-assembly technology.
Alon et al. investigated super-resolved on-chip color imaging in
which they implemented YUV color space averaging and Dijkstra's
shortest-path computational methods (Greenbaum et al., 2013).
Again Mudanyali et al. had demonstrated a high-throughput, on-
chip detection scheme that uses biocompatible wetting lms to
self-assemble aspheric liquid nanolenses around individual na-
noparticles to enhance the contrast between the scattered and
background light (Mudanyali et al., 2013). In this work they used
the pixel super resolution on-chip imaging technique to capture
the nano lens enhanced light signals of the nanoparticles. The
nanolenses were fabricated by using dispersed components over
the individual nanoparticles as stated above. They demonstrated
the feasibility of this system by detecting the viral particles of
H1N1. This indicates the feasibility of the lens-free system for
detecting nano particles. Sobieranski et al. (2015) had also de-
monstrated the pixel super resolution lens-free imaging technol-
ogy to detect sperms and platelets that are in the dimension of a
few microns (Sobieranski et al., 2015). This work also showed the
feasibility of the on-chip gigapixel nanoscopy available for the
characterization of sub-micron cells or particles.
3.5. Lens-free imaging and sensing systems for point-of-care
diagnostics
The lens-free imaging system is compact, inexpensive and has
high throughput, all of which are suitable characteristics for a point-
of-care system. In this section, we discuss recent developments in

Please cite this article as: Roy, M., et al., Biosensors and Bioelectronics (2016), http://dx.doi.org/10.1016/j.bios.2016.07.115i
 M. Roy et al. / Biosensors and Bioelectronics () 9

Fig. 11. Lens-free holography and on-chip uorescent imaging is demonstrated. Left column: raw lens-free uorescent images. Middle column: results of digital decon-
volution. Right column: results of lens-free holographic imaging of the same FOV, showing the shadow signatures of all the particles, both uorescent (F) and nonuorescent
(NF). The images in the other two columns show only the uorescent signatures (Coskun et al., 2010).

lens-free imaging technology with respect to a point-of-care sys-
tem. Umut et al. presented a study on a lens-free imaging system as
a point-of-care system (Gurkan et al., 2011), and Zhu et al. published
a critical review on optical imaging technology for point-of-care
diagnostics, in which they reviewed the development of the lens-
free imaging system as a point-of-care system (Zhu et al., 2013a).
Portability, low power consumption, functionality in heat and
humidity, and continuous monitoring are the major challenges in
realizing point-of-care devices. Some have built point-of-care
systems with the goal of incorporating all these features by in-
tegrating microuidics technology with the lens-free imaging
platform. Moon et al. diagnosed the CD4 T-lymphocyte count for
HIV with such a point-of-care system (Moon et al., 2009). In ad-
dition, Biener et al. combined reection and transmission imaging
by integrating lens-free imaging, i.e., transmission imaging, and an
optical imaging system, i.e., reection microscopy (see Fig. 16).
This setup provides spatial and phase information of samples si-
multaneously (Biener et al., 2011). On the other hand, studies on
retrieving phase and spatial information using computational
methods are underway. Mudanyali et al. demonstrated the
reconstruction of lensless incoherent images from a compact and
lightweight microscope (Mudanyali et al., 2010).
Many groups are using the self-assembly approach to eliminate the
optical limitation of the lens-free imaging system. In the self-assembly
process, the individual dispersed components naturally adopt the
desired conguration and form micro- and nanolenses that enhance
the signal with a spatial resolution of o100 nm (McLeod and Ozcan,
2014).
The cell viability test is a major component in the diagnosis of
many cell-borne diseases. Measurement of cell viability in a point-of-
care system has been a major challenge due to its complexity and
reagent-dependent methods. Recently, Jin et al. successfully de-
monstrated a reagent-free method for characterizing cell viability
using a lens-free imaging platform (Jin et al., 2012). They monitored
the continuous viability of human alveolar epithelial A549 cells
without the use of a labeling reagent. By integrating video micro-
scopy with lens-free imaging, this work improved upon that of Ke-
savan et al., who utilized this system to continuously monitor cell
proliferation inside a standard incubator (Kesavan et al., 2014).
However, video processing is not intended for point-of-care systems.

Please cite this article as: Roy, M., et al., Biosensors and Bioelectronics (2016), http://dx.doi.org/10.1016/j.bios.2016.07.115i
10 M. Roy et al. / Biosensors and Bioelectronics ()

Fig. 12. Schematic diagram of the lens-free on-chip uorescent imaging platform with unit magnication such that the imaging FOV is equal to the entire active area of the
sensor array (i.e.,48 cm2). Total internal reection occurs at the glassair interface at the bottom facet of the coverglass. To avoid detection of scattered pump photons, a
plastic absorption lter is introduced after the faceplate. Typical dimensions (see caption for Fig. 9 for a description of the parameters): w1  w2 25 mm  35 mm;
p 1.7 cm; k  10100 mm; f 12 cm. (b) Microscope image of the faceplate. The numerical aperture of each ber is  0.3 (Coskun et al., 2010).

The most common point-of-care diagnostic test is the complete operate them and analyze their results. Because a point-of-care
blood cell count. In general, a complete blood cell count is performed system needs to be quick and easy, Roy et al. developed a custom
using a sophisticated hematology analyzer or a conventional optical telemedicine system that can perform a complete blood cell count of
microscope, both of which are bulky and need a trained user to human whole blood. They also used the device to calculate the

Fig. 13. Top row: Raw lens-free uorescent images of different pairs of 10-mm-diameter particles. As the particles get closer to each other, their signatures in the raw lens-
free image become indistinguishable to the naked eye. The inset images (bottom right corner) are transmission microscope images of the same particles and were used to
calculate the center-to-center distance (g) for comparison purposes. Middle row: Results of compressive decoding of the lens-free images in the top row. gCS is the center-to-
center distance of the resolved uorescent particles, where CS means compressive sampling. Bottom row: results of deconvolution of the lens-free images in the top row
using the LucyRichardson algorithm. gLR is the center-to-center distance of the resolved uorescent particles, where LR means LucyRichardson (Coskun et al., 2010).

Please cite this article as: Roy, M., et al., Biosensors and Bioelectronics (2016), http://dx.doi.org/10.1016/j.bios.2016.07.115i
 M. Roy et al. / Biosensors and Bioelectronics () 11

Fig. 14. Lens-free simultaneous multicolor uorescent imaging on a chip. (a1), (b1), (c5), and (d5) Raw lens-free uorescence signatures of microbeads cropped from a large
FOV of 2.42 cm2. These raw images were decoded to retrieve color and spatial distribution information. (a3), (b3), (c6), (c7), (c8), (d6), (d7), and (d8) Decoding results are
veried with [insets in (a4) and (b4)] bright-eld and [insets in (c3), (c4), (d3), and (d4)] uorescent microscope images of the same objects. (a2), (b2), (c1), and (d1)
Debayered lens-free images illustrate the limitations of the color sensor for lens-free uorescent imaging without our demultiplexing method. The arrows highlight two
regions where our method effectively resolved the color and the spatial overlap of lens-free uorescent signatures.

Please cite this article as: Roy, M., et al., Biosensors and Bioelectronics (2016), http://dx.doi.org/10.1016/j.bios.2016.07.115i
12 M. Roy et al. / Biosensors and Bioelectronics ()

Fig. 15. Schematic diagram illustrating the principles of multiangle lens-free holographic imaging. For each illumination angle, a spatially incoherent source, such as an LED,
is ltered by a large aperture (  0.050.1-mm diameter), which is  6 cm from the object plane. Unlike conventional inline holography, the sample plane is much closer to
the detector plane, with a vertical distance of  1 mm, so that the entire active area of the sensor becomes the imaging FOV. (a) The shadow of each cell shifts laterally on the
sensor plane as a function of the illumination angle of the incoherent source, encoding its axial position. (b) The shadows of the cells acquired at different illumination angles
are matched to their sources by forming imaginary rays between each shadow and the corresponding source. The shift of the centroid position as a function of illumination
angle was used to determine the depth of the cells on a chip. This technology demonstrated the accuracy of the depth of a cell to be  300400 nm (Su et al., 2010b).

Fig. 16. Combination of transmission lens-free imaging and optical imaging via reection microscopy.

subpopulation of white blood cells in the samples, i.e., the con-
centration of lymphocytes, monocytes, and neutrophils, on the basis
of the characterization of the custom-developed diffraction para-
meter, as shown in the Fig. 17 (Roy et al., 2014). The work of Zhu et al.
similarly characterized blood cells using a smartphone-powered
point-of care system (Zhu et al., 2013b).
Automated determination of the size of cells and other biolo-
gical substances is a major criterion for the diagnosis of many
diseases. However, until now, sophisticated instruments such as
the Coulter counter and conventional optical microscopes were
used for measuring cell size. Integrating these functions with a
point-of-care system has been a major challenge. Recently, Roy
et al. demonstrated a low-cost lens-free telemedicine system that
can automatically detect and characterize the size of cells on the
basis of their shadow parameter, the latter which they custom
developed, as shown in Fig. 18 (Roy et al., 2015). The compactness
and the ability to perform reagent-free detection makes this sys-
tem a suitable candidate for a point-of-care diagnostic instrument.
4. Conclusions and perspectives
In this review, we summarized the theory, basic design, and
application of the lens-free imaging system. With respect to theory

Please cite this article as: Roy, M., et al., Biosensors and Bioelectronics (2016), http://dx.doi.org/10.1016/j.bios.2016.07.115i
 M. Roy et al. / Biosensors and Bioelectronics () 13

Fig. 17. Illustration of the custom-developed diffraction parameters (shadow parameters). (a) and (b) Shadow images of 10- and 20-mm beads. (c) and (d) Intensity proles
from (a) and (b), respectively. (e) 3D plot of the diffraction parameters of heterogeneous samples of 10- and 20-mm beads.

and design, the lens-free imaging system can be divided into three DIH on phase and amplitude data and on the advanced compu-
categories: DIH, lens-free imaging with partially coherent light, tational method have been partly eliminated by the introduction
and on-chip lens-free imaging. The complexity and dependence of of lens-free imaging with partially coherent light. These simplied

Fig. 18. Detailed evaluation of the mathematical relationship between peak-to-peak distance (PPD) and the actual size of a microparticle. (a) Optical micrograph of a
heterogeneous sample of beads. (b) Digitally magnied lens-free shadow image of the sample in (a). (c) Intensity prole of the 21-mm-diameter bead. (d) Intensity prole of
the 10-mm-diameter bead. (e) PPD vs. actual size of 222 beads. (f) Statistical size distribution of 222 beads calculated from the PPD.

Please cite this article as: Roy, M., et al., Biosensors and Bioelectronics (2016), http://dx.doi.org/10.1016/j.bios.2016.07.115i
14 M. Roy et al. / Biosensors and Bioelectronics ()
systems are used for quantication and size determination. A
range of applications has been developed from these techniques.
The integration of on-chip uorescent imaging with traditional
optical microscopy provides better insight into samples by facil-
itating cross inspection in a large FOV. Spatial resolution can be
improved by implementing a compressive sensing algorithm, and
the effective bandwidth of the emitted light of the uorescent
particles can be improved by implementing the demultiplexing
method based on the polychromatic phase retrieval algorithm. On-
chip tomography can provide depth-resolved imaging of micro-
particles with a wide FOV. This compact system is suitable for
integration with existing on-chip analysis tools such as lab-on-a-
chip. A super-resolved lens-free image of microparticles can be
obtained by capturing multiple images of holograms of the objects
by subpixel shifting of the image plane. These images are then
fused into a single high-resolution image using a pixel super-re-
solved algorithm and an iterative gradient algorithm. This super-
resolution image can be reconstructed using back-propagation of
the object plane and an angular spectrum approach. The com-
pactness and low cost of the lens-free system make it suitable for
point-of-care diagnostic instruments. Integration of the automated
shadow image feature extraction and characterization algorithm
with the lens-free system creates an alternative system that per-
forms a complete blood cell count and size characterization. The
lens-free system integrated with sophisticated microuidics can
be used for measuring continuous cell viability. All these low-cost,
compact, and fast-processing lens-free imaging and sensing
techniques may play a crucial role especially in the elds of en-
vironmental, pharmaceutical, biological, and clinical applications
of the resource-limited settings.
Acknowledgment
This research was supported by the Basic Science Research
Program (Grant#:2013-010832, Grant#:2014R1A6A1030732)
through the National Research Foundation of Korea (NRF) funded
by the Ministry of Science, ICT & Future Planning and the Ministry
of Education, Korea. This work was also supported by the Korea
Research Institute of Ships and Ocean Engineering (KRISO) En-
dowment-Grant (PES2440) and a part of the project Development
of Management Technology for HNS Accident, funded by the
Ministry of Oceans and Fisheries, Korea.
References
Bishara, W., Su, T.-W., Coskun, A.F., Ozcan, A., 2010. Opt. Express 18 (11),
1118111191.
Bishara, W., Sikora, U., Mudanyali, O., Su, T.-W., Yaglidere, O., Luckhart, S., Ozcan, A.,
2011a. Lab Chip 11 (7), 1276.
Please cite this article as: Roy, M., et al., Biosensors and Bioelectronics (2016), http://dx.doi.org/10.1016/j.bios.2016.07.115i
Bishara, W., Isikman, S.O., Ozcan, A., 2011b. Ann. Biomed. Eng. 40 (2), 251262.
Biener, G., Greenbaum, A., Isikman, S.O., Lee, K., Tseng, D., Ozcan, A., 2011. Lab Chip
11 (16), 27382743.
Coskun, A.F., Ozcan, A., 2014. Curr. Opin. Biotechnol. 25, 816.
Coskun, A.F., Su, T.-W., Ozcan, A., 2010. Lab Chip 10 (7), 824827.
Cui, X., Lee, L., Heng, X., Zhong, W., Sternberg, P., Psaltis, D., Yang, C., 2008. Proc.
Natl. Acad. Sci. USA 105, 1067010675.
Gorocs, Z., Ozcan, A., 2013. IEEE Rev. Biomed. Eng. 6, 2946.
Greenbaum, A., Sikora, U., Ozcan, A., 2012a. Lab Chip 12 (7), 1242.
Greenbaum, A., Feizi, A., Akbari, N., Ozcan, A., 2013. Opt. Express 21 (10),
1246912483.
Gurkan, U.A., Moon, S., Geckil, H., Xu, F., Wang, S., Lu, T.J., Demirci, U., 2011. Bio-
technol. J. 6 (2), 138149.
Greenbaum, A., Luo, W., Su, T.-W., Grcs, Z., Xue, L., Isikman, S., Coskun, A., Mu-
danyali, O., Ozcan, A., 2012b. Nat. Methods 9, 889895.
Isikman, S.O., Bishara, W., Mavandadi, S., Yu, F.W., Feng, S., Lau, R., Ozcan, A., 2011a.
Proc. Natl. Acad. Sci. USA 108 (18), 72967301.
Isikman, S.O., Bishara, W., Sikora, U., Yaglidere, O., Yeah, J., Ozcan, A., 2011b. Lab
Chip 11 (13), 22222230.
Jin, G., Yoo, I.-H., Pack, S.P., Yang, J.-W., Ha, U.-H., Paek, S.-H., Seo, S., 2012. Biosens.
Bioelectron. 38 (1), 126131.
Kesavan, S.V., Momey, F., Cioni, O., David-Watine, B., Dubrulle, N., Shorte, S., Allier,
C., 2014. Sci. Rep. 4. http://dx.doi.org/10.1038/srep05942.
Kim, S.B., Bae, H., Cha, J.M., Moon, S.J., Dokmeci, M.R., Cropek, D.M., Kha-
demhosseini, A., 2011. Lab Chip 11 (10), 18011807.
Kim, S.B., Bae, H., Koo, K., Dokmeci, M.R., Ozcan, A., Khademhosseini, A., 2012. J. Lab.
Autom. 17 (1), 4349.
McLeod, E., Luo, W., Mudanyali, O., Greenbaum, A., Ozcan, A., 2013. Lab Chip 13 (11),
20282035.
McLeod, E., Ozcan, A., 2014. Nano Today 9, 560573.
Moon, S., Keles, H.O., Ozcan, A., Khademhosseini, A., Haeggstrom, E., Kuritzkes, D.,
Demirci, U., 2009. Biosens. Bioelectron. 24 (11), 32083214.
Mudanyali, O., Tseng, D., Oh, C., Isikman, S.O., Sencan, I., Bishara, W., Ozcan, A., 2010.
Lab Chip 10 (11), 14171428.
Mudanyali, O., McLeod, E., Luo, W., Greenbaum, A., Coskun, A., Hennequin, Y., Allier,
C., Ozcan, A., 2013. Nat. Photonics 7, 247254.
Ozcan, A., Demirci, U., 2008. Lab Chip 8 (1), 98106.
Picart, P., Leval, J., 2008. J. Opt. Soc. Am. A 25 (7), 17441761.
Roy, M., Jin, G., Seo, D., Nam, M.H., Seo, S., 2014. Sens. Actuators B: Chem. 201,
321328.
Roy, M., Seo, D., Oh, C., Nam, M., Jun, Y., Seo, S., 2015. Biosens. Bioelectron. 67,
715723.
Schnars, U., Jptner, W., 1994. Appl. Opt. 33, 179181.
Schultz, D.A., 2003. Curr. Opin. Biotechnol. 14, 1322.
Sencan, I., Coskun, A.F., Sikora, U., Ozcan, A., 2014. Sci. Rep. 4. http://dx.doi.org/
10.1038/srep03760.
Seo, S., Su, T.-W., Tseng, D.K., Erlinger, A., Ozcan, A., 2009. Lab Chip 9 (6), 777787.
Seo, S., Isikman, S.O., Sencan, I., Mudanyali, O., Su, T.-W., Bishara, W., Ozcan, A., 2010.
Anal. Chem. 82 (11), 46214627.
Sobieranski, A., Inci, F., Tekin, H., Yuksekkaya, M., Comunello, E., Cobra, D., Wan-
genheim, A., Demirci, U., 2015. Light Sci. Appl. 4, e346.
Su, T.-W., Erlinger, A., Tseng, D., Ozcan, A., 2010a. Anal. Chem. 82 (19), 83078312.
Su, T.-W., Isikman, S.O., Bishara, W., Tseng, D., Erlinger, A., Ozcan, A., 2010b. Opt.
Express 18 (9), 96909711.
Wei, Q., McLeod, E., Qi, H., Wan, Z., Sun, R., Ozcan, A., 2013. Sci. Rep. 3, 1699. http:
//dx.doi.org/10.1038/srep01699.
Xu, W., Jericho, M.H., Meinertzhagen, I.A., Kreuzer, H.J., 2002. Appl. Opt. 41 (25),
53675375.
Zhang, X., Khimji, I., Gurkan, U., Safaee, H., Catalano, P., Keles, H., Kayaalp, E., De-
mirci, U., 2011. Lab Chip 11, 25352540.
Zhu, H., Isikman, S.O., Mudanyali, O., Greenbaum, A., Ozcan, A., 2013a. Lab Chip 13
(1), 5167.
Zhu, H., Sencan, I., Wong, J., Dimitrov, S., Tseng, D., Nagashima, K., Ozcan, A., 2013b.
Lab Chip 13 (7), 12821288.