Scoring of collagen organization in healthy and diseased human dermis by multiphoton microscopy. J. Biophoton. 3, 34-43

L.E.N.S. and Department of Physics, University of Florence, Sesto Fiorentino, Italy.
Journal of Biophotonics (Impact Factor: 4.45). 09/2009; 3(1-2):34-43. DOI: 10.1002/jbio.200910062
Source: PubMed
We have used nonlinear imaging to evaluate collagen organization in connective tissue ex-vivo samples. Image analysis methods were tested on healthy dermis, normal scars, and keloids. The evaluation of the second harmonic to autofluorescence aging index of dermis (SAAID) has allowed a first characterization of tissues by scoring the collagen/elastin content. Further analyses on collagen morphology in healthy dermis and keloids were performed by image-pattern analysis of SHG images. The gray-level co-occurrence matrix (GLCM) analysis method has allowed classification of different tissues based on the evaluation of geometrical arrangement of collagen fibrillar bundles, whereas a pattern analysis of the FFT images has allowed the discrimination of different tissues based on the anisotropy of collagen fibers distribution. This multiple scoring method represents a promising tool to be extended to other collagen disorders, as well as to be used in in-vivo skin-imaging applications.


Available from: Dimitrios Kapsokalyvas, Oct 05, 2014
Scoring of collagen organization in healthy and
diseased human dermis by multiphoton microscopy
Riccardo Cicchi
; 1
, Dimitrios Kapsokalyvas
, Vincenzo De Giorgi
, Vincenza Maio
Annelies Van Wiechen
, Daniela Massi
, Torello Lotti
, and Francesco S. Pavone
; 1; 5
L.E.N.S. and Department of Physics, University of Florence, Via Nello Carrara 1, 50019, Sesto Fiorentino, Italy
Department of Dermatology, University of Florence, Via della Pergola 58, 50121, Florence, Italy
Department of Human Pathology and Oncology, University of Florence, Viale G.B. Morgagni 85, 50134, Florence, Italy
Kavli Institute of Nanoscience, Delft University of Technology, Lorentzweg 1, 2628 CJ, Delft, The Netherlands
C.N.I.S.M., Unity of Florence, Sesto Fiorentino, Italy
Received 28 July 2009, revised 7 August 2009, accepted 10 August 2009
Published online 22 September 2009
Key words: multiphoton microscopy, second-harmonic generation microscopy, dermal diseases
PACS: 42.65.Ky,
2010 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
2010 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Journal of
We have used nonlinear imaging to evaluate collagen
organization in connective tissue ex-vivo samples. Image
analysis methods were tested on healthy dermis, normal
scars, and keloids. The evaluation of the second harmo-
nic to autofluorescence aging index of dermis (SAAID)
has allowed a first characterization of tissues by scoring
the collagen/elastin content. Further analyses on col-
lagen morphology in healthy dermis and keloids were
performed by image-pattern analysis of SHG images.
The gray-level co-occurrence matrix (GLCM) analysis
method has allowed classification of different tissues
based on the evaluation of geometrical arrangement of
collagen fibrillar bundles, whereas a pattern analysis of
the FFT images has allowed the discrimination of differ-
ent tissues based on the anisotropy of collagen fibers
distribution. This multiple scoring method represents a
promising tool to be extended to other collagen disor-
ders, as well as to be used in in-vivo skin-imaging appli-
Slice of a sample of keloid labeled with H&E staining
and imaged using nonlinear microscopy. Field of view:
1 0.5 mm
Corresponding author: e-mail:
J. Biophoton. 3, No. 12, 3443 (2010) / DOI 10.1002/jbio.200910062
Page 1
1. Introduction
Nonlinear microscopy is a high-resolution laser scan-
ning technique enabling deep optical tissue imaging
[1, 2]. Cells and extracellular matrix intrinsically con-
tain a variety of fluorescent molecules (NADH, tryp-
tophan, keratins, melanin, elastin, cholecalciferol
and others), so that biological tissues can be imaged
by nonlinear microscopy without any exogenously
added fluorophore or probe [3, 4].
Among nonlinear imaging techniques, two-
photon excited fluorescence (TPEF) microscopy re-
presents a powerful tool for skin imaging. It has al-
ready been widely used on ex-vivo tissue samples to
perform optical biopsy by means of a morphological
characterization [59]. In recent years, its in-vivo
use on both animal models and human subjects [9
12] has also been exploited. Another nonlinear im-
aging technique providing morphological information
on biological tissues is second-harmonic generation
(SHG) microscopy. SHG has already been largely
used to image noncentrosymmetric molecules inside
cells [13, 14] and tissues [1517] and to measure
their organization. Collagen fibers produce a high
SHG signal [18] and can be imaged inside skin der-
mis with SHG microscopy [4, 18]. Recently, SHG
was also used for investigating collagen-fiber orien-
tation and their structural changes in fibrotic col-
lagen [19], human dermis [2023], keloid [24], cor-
nea [25, 26] and in the tumor microenvironment
[2730]. The combination of TPEF and SHG is par-
ticularly useful when imaging dermis tissue because
the two main components of dermis (collagen and
elastin) can be imag ed wi th SHG and TPEF micro-
scopy, respectively. Combined TPEF-SHG micro-
scopy was applied to skin physiology and pathology,
and specifically to the study of normal skin [59,
17], cutaneous photoaging [31, 32], psoriasis [9], and
selected skin tumors, including basal cell carcinoma
(BCC) [28, 29, 33, 34], and malignant melanoma
(MM) [9, 11, 12].
Collagen is a protein found in organisms to pro-
vide strength. There are at least sixteen types of col-
lagen of which the types I, II, and III constitute at
least 8090% of the body’s collagen [35]. In this re-
search we have studied collagen type-I in skin der-
mis. The fundamental structure of a collagen fibril is
called tropocollagen. This is a long (300 nm), thin
(1.5 nm diameter) protein consisting of three coiled
subunits [36]. Tropocollagen molecules bundle them-
selves into collagen fibrils by a side by side arrange-
ment of the tropocollagen. The diameter of the fi-
brils ranges from 50 to 200 nm. The fibrils arrange
together to form collagen fibrous bundles with a di-
ameter varying between 0.5 and 3 mm [37, 38].
In healthy dermis (HD) the collagen-fiber bun-
dles are arranged in a parallel way to the epithelial
surface and are interconnected by fine fibrillar
strands of collagen [39]. Cutaneous keloids (K) and
hypertrophic scars (HS) are examples of abnormal
wound healing and are characterized by excessive
dermal fibrosis. They are formed during wound heal-
ing after skin damage. In K a proliferating activity of
fibroblasts takes place, producing a large amount of
collagen, which is arranged in large, thick fibers that
are chaotically connected in loose sheets that appear
randomly oriented to the epithelial surface. Collagen
synthesis in K is approximately 20 times greater than
in normal unscarred skin. K develop in genetically
susceptible individuals and, unlike normal scars
(NS), do not regress. K and HS scars are difficult to
be managed and the treatment (surgery vs. intra-
lesional corticosteroid injection vs. pressure therapy,
radio- or cryotherapy, and other topical treatments)
must be carefully tailored for each patient. K are a
common disease among the darker-pigmented races.
Until today the origin of this diseas e was unknown,
and an effective cure for K has not yet been devel-
The aim of this paper is to test and compare dif-
ferent image-pattern analysis methods for collagen
scoring to be used in-vivo in applications involving
imaging of collagen remodeling. The methods pre-
sented here could be helpful not only to understand
the process of scar formation both in-vitro and in-
vivo but also to be used in in-vivo collagen follow-up
after skin treatments, including laser-based therapies.
Three different scoring methods were tested on ex-
vivo skin samples of healthy dermis (HD), normal
scar (NS), and keloid (K). In parti cular, by using
combined SHG and TPEF microscopy, we have in-
vestigated morphological alterations of skin connec-
tive tissue by scoring the collagen/elastin content.
The use of second harmonic to autofluorescence
aging index of dermis (SAAID) parameter, intro-
duced for the first time to assess skin photoaging
[31], was extended to map the collagen-elastic tissue
ratio of scarring-related conditions of dermis. Dif fer-
ent distributions of elastin were found in HD and
NS, respectively. In NS, elastin fibers were found
tightly packed together in region well-separated
from collagen, instead of intermixed with collagen fi-
bers, as in HD. A complete lack of elastin was found
in K. Further analysis on SHG images of collagen
was performed in HD and K by using two different
image-pattern analysis methods. The two methods
were carried out on two different repres entation
spaces. The first method is based on the calculation
of correlation, homogeneity, and energy of the gray-
level co-occurrence matrix (GLCM) of the images.
With this method we were able to characterize tis-
sues by estimating the typical length within which
collagen maintains its organization. The second
method was carried out in the Fourier domain by
analyzing the aspect ratio of the bidimensional inten-
sity distribution of the FFT images. This method al-
J. Biophoton. 3, No. 12 (2010) 35
2010 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Page 2
lowed characterization of the SHG images of tissues
based on their anisotropy.
2. Materials and methods
2.1 Experimental setup
The experimental setup used is a custom-made up-
right microscope (see Figure 1). It can perform com-
bined TPEF and SHG microscopy. All the micro-
scope components were mounted onto a custom
vertical honeycomb steel breadboard 60 90 cm,
070BH0450 (Melles-Griot, Rochester, NY, US),
fixed through two square brackets onto an anti-vi-
brating optical table 100 200 cm (TMC, Peabody,
MA, US). A mode-locked Ti : sapphir e laser CHA-
MELEON (Coherent Inc, Santa Clara, CA, US)
provides the excitation light (120 fs width pulses, 90-
MHz repetition rate, 705980 nm wavelength range).
The laser light beam path includes a collimating tele-
scope (L1,L2), a half-wavelength broadband wave-
plate (l/2), coupled with a calcite polarizing beam
splitter (PBS) for laser power adjustment, and a cus-
tom-made electronic shutter (S), before passing to
the scanning head. The scanning head comprises two
galvanometer mirrors (G1,G2) VM500 (GSI Lumo-
nics, Munich, Germany), rotate d about orthogonal
pivots and coupled by a spherical mirrors pair
(Sm1,Sm2, f ¼ 100 mm). A scanning lens (L3,
f ¼ 50 mm) and microscope tube lens (TL,
f ¼ 200 mm) expand the beam to a dimension of
around 1 cm in order to fill the objective back-aper-
ture, before being focused onto the specimen by the
objective lens XLUM 20, N.A. 0.9, W.D. 2 mm
(Olympus Co., Japan). The objective lens is mounted
on a custom-made mechanical support, fixed to a
manual actuator (25 mm range, 0.05 mm resolution)
that allows its positioning along the optical axis. A
piezoelectric stage (PZT) PIFOC P-721 (Physik In-
strumente GmbH, Karlsruhe, Germ any) allows fine
axial displacements of the objective up to 100 mm
with 1-nm resolution and hence the optical section-
ing of the sample. A blocking filter (SBF) E-700-
SP2P (Chroma Technology Corporation, Rocking-
ham, VT, US) prevent back-reflected laser light to
be detected.
The detection system, constituted by a collecting
lens (L4), a dichroic (D2) with 457 nm wavelength
(DC457LP Semrock, Rochester, NY, US), and two
photomultiplier tubes (PMTs) H7713 (Hamamatsu
Photonics, Hamamatsu City, Japan) working in the
proportional regime, allows simultaneous acquisition
of TPEF and SHG intensity images. A narrow
(20 nm FWHM) bandpass optical filter centered at
half the excitation wavelength is placed (HQ420BP,
Chroma Technology Corporation, Rockingham, VT,
US) in front of the SHG detector to cut out un-
wanted fluore scence light from the SHG channel.
2.2 Patients
This study included 3 samples of healthy dermis
(HD), 2 of NS, and 3 of K, surgically removed from
5 patients who agreed to participate in the study.
Normal healthy skin was obtained from residual spe-
cimens discarded after plastic surgery of the breast
or abdomen. Normal scars and keloids were excised
due to cosmetical reasons. Upon excision, all skin
specimens were bi-halved, half of the specimen for-
malin fixed and paraffin embedded for conventional
histopathological examination and the other portion
evaluated and imaged using TPEF and SHG. Im-
aging was performed after obtaining a written consent
from all patients. Each lesion was excised with at
least 23 mm healthy skin margins.
2.3 Sample preparation
The samples for nonlinear imaging were prepared
by freezing them and by cutting them in a compact
routine cryostat. Slices ranging from 50 to 100 mm
thickness were obtained. Before imaging, samples
were unf rozen and closed with a microscope cover-
slide together with some PBS droplets in order to
maintain the natural tissue osmolarity.
Figure 1 Schematic of the experimental setup for non-
linear imaging.
R. Cicchi et al.: Scoring of collagen organization36
Journal of
2010 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Page 3
2.4 Image acquisition and analysis
The excitation was accomplished using a wavelength
of 740 nm for TPEF and 840 nm for SHG and a
mean laser power at the sample between 10 and
40 mW, depending on the depth of recording. These
power levels are sufficiently low to avoid photo-
bleaching and photodegradation effects, as demon-
strated also in in-vivo multiphoton skin imaging [32].
SHG or TPEF detection was accomplished by
switching the excitation wavelength between 740 nm
and 840 nm. TPEF-SHG images were acquired with
512 512 pixels or 1024 1024 pixels spatial resolu-
tion, from 60 mm to 200 mm lateral dimension, using
a pixel dwell time of 5 ms. The scanning time was ap-
proximately 1.3 s and 5 s for a 512- and for a 1024-
pixels image, respectively.
For SAAID scoring images were processed with
LabView 7 (National Instruments, Austin, TX, US).
The calculated score is the geometrical mean of the
value calculated from 10 different 50 mm side
squared ROI for each sample. For GLCM analysis
images were processed using Matlab (Mathworks,
Natick, MA, US). GLCM matrix were calculated
from SHG images according to [40]. A running
(from 1 to 200) neighbor index allowed calculation
of a GLCM matrix per neighbor index per SHG im-
age. Correlation, homogeneity, and energy were cal-
culated and presented versus neighbor index. For
FFT statistical analysis images were process ed using
a Matlab routine. FFT images and image merging
were done using ImageJ (NIH, Bethesda, Maryland,
US). A statistical t-test was done on SAAID and
on FFT results to highlight statistical differences at
the 0.05 level using Microcal Origin Pro8 software
(OriginLab Corporation, Northampton, MA, US).
GLCM and FFT analysis were applied to all the
samples excluding NS because the small dimension
of NS samples allowed averaging only on an extre-
mely limited number of slices with respect to HD
and K.
3. Results and discussion
3.1 Combined two-photon excited
fluorescence (TPEF) and second-harmonic
generation (SHG) imaging
In all the examined samples morpholog ical features
of human skin can be resolved using combined
TPEF and SHG microscopy. In Figure 2 examples of
TPEF (first column), SHG (second column), and
merged (third column) images of HD, NS, and K are
presented. We have verified that the signal repre-
Figure 2 (online color at:
Transversal optically sectioned ex-vivo skin sample of
healthy dermis (HD), imaged using TPEF (a), SHG (b),
and the merge between the two images (c). Scale bars:
15 mm. Transversal optically sectioned ex-vivo sample of
normal scar (NS), imaged using TPEF (d), SHG (e), and
the merge between the two images (f). Scale bars: 20 mm.
Area with fibroblastic proliferation inside a transversal op-
tically sectioned ex-vivo skin sample of keloid (K), imaged
using TPEF (g), SHG (h), and merging the two images (i).
Scale bars: 20 mm. Area with keloidal fibers inside a trans-
versal optically sectioned ex-vivo skin sample of keloid
(K), imaged using TPEF (j), SHG (k), and merging the
two images (l). Scale bars: 20 mm. SAAID scores, calcu-
lated in five different zones for each of the four regions
shown in the images, are plotted in a bar graph (m). All
the SAAID values were found to be statistically different
at the 0.05 level after a two-sample statistical t-test.
J. Biophoton. 3, No. 12 (2010) 37
2010 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Page 4
sented in Figures 2b, e, h, and k is only SHG, by
performing spectral and lifetime analysis of the col-
lected signal (data not shown). In the acquired
images HD (Figures 2ac), K (Figures 2gl) and NS
(Figures 2df) are resolved. We have characterized
the morphology of these three different tissues by
measuring the SAAID score. The SAAID value is
a measure of the ratio between collagen and elastic
tissue. It can be used to evaluate intrinsic and ex-
trinsic skin aging, as well as to give a measure of
the fibrotic status of the dermis. It is defined as fol-
þ I
The SAAID score was found to differ between
HD, K, and NS. Within a sample of K we were also
able to discriminate different regions having differ-
ent histopathologic features, as such the fibroblastic/
myofibroblastic cellular proliferation (Figures 2gi)
and the so-called “amiantoid collagen fibers” (Fig-
ures 2jl), typical of K, which were found in the dee-
pest layers of the lesion. For each region analyzed,
we calculated the SAAID score by averaging five
different images acquired inside the same region.
Here, we found that NS has a negative SAAID val-
ue, due to the high TPEF signal coming from elastin.
HD has a small pos itive value, whereas fibrobl astic/
myofibroblastic cellular proliferation and keloidal
fibers exhibited a SAAID value close to the unit,
because of their high collagen concentration and
because of their reduced TPEF, with respect to HD.
The obtained values of SAAID score for the four
regions are plotted in a bar graph in Figure 2m. The
SAAID calculated values for each ROI were found
to be statistically different at the 0.05 level after a
two-sample statistical t-test.
3.2 Image-pattern analysis with gray-level
co-occurrence matrix (GLCM)
on SHG images
The GLCM space of an image is a good method to
analyze texture patterns. It gives information on the
spatial relationships between pixel brightness values
in an image. It compares the gray level of a pixel in
the original image to the gray level of its neighbors.
Based on the amount of difference in gray level be-
tween the two it creates a new matrix with this infor-
mation. A detailed explanation on how this matrix
in created from the original image can be found in
[39]. On the GLCM different methods of analysis
can be used. These methods concern measurements
on the homogeneity, energy, mean value, contrast
and correlation of the GLCM [39]. They are com-
monly classified into co ntrast methods, orderliness
methods, and statistical methods. Among them, one
method for each group was selected. Measurements
on correlation, homogeneity and energy of the
GLCM wer e carried out in this experiment.
Correlation is a measur e of dependence of two
different pixel values. If two pixels are well corre-
lated their intensity values are consistent with a
well-defined relationship. In calculating the correla-
tion of the GLCM, information on the correlation of
a combination of pixels is provided. The GLCM cor-
relation calculates the similarity of a pixel value in
combination with its neighboring pixel to all the
other pairs of neighboring pixels in the image. Com-
paring the GLCM correlation signal s at different off-
sets can provide useful information on our image. If
a certain offset returns a high co rrelation value, this
means the image has a returning structure at a cer-
tain offset that corresponds to a certain distance be-
tween pixels. Furthermore, it can be used in a more
general approach to provide information on the dis-
tance between pixels within an image in which the
image can be considered to be correlated. This can
provide information on the sudden change or regu-
larity of a linear structure, such as breakage of a fi-
brillar structure. Correlation (R), expectation values
and m
, and standard de viations s
and s
are calcu-
lated for a GLCM matrix with a normalized intensity
i, j
as follows:
R ¼
i; j
ði m
Þðj m
Þ p
i ¼ 1
i p
i; j
; m
i ¼ 1
j p
i; j
N 1
i ¼ 1
i; j
; s
N 1
i ¼ 1
i; j
Further parameters, derivable from the GLCM
analysis, and useful in evaluating image pattern mor-
phology are represented by homogeneity and en-
ergy. Homogeneity is the weighted sum of the
GLCM pixel values. The weights are nonlinearly de-
creasing in value as the distance from the GLCM
matrix diagonal increases. Keeping in mind that the
distance from the diagonal represents the difference
in intensity between a pixel in the image and its
neighbor (at a fixed neighbor index), homogeneit y
gives information on the similarity of a pixel value in
combination with its neighboring pixel to all the
other pairs of neighboring pixels in the image. En-
ergy is the root squared sum of the GLCM pixel
values. Considering that this parameter gives higher
weight to the hot spots of the GLCM matrix, en-
ergy is a measure of orderliness of the image.
Homogeneity H and energy E are calculated for a
R. Cicchi et al.: Scoring of collagen organization38
Journal of
2010 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Page 5
GLCM matrix with a normalized intensity p
i, j
H ¼
i; j
1 þði jÞ
E ¼
i; j
According to literature values collagen fibers have a
diameter ranging between 0.53 mm. In dermis these
fibers are arranged mainly parallel to the epithelial
surface. This means a repeating structure with a pix-
el distance ranging between 0.53 mm should be ex-
pected to be able to be derived from the GLCM. K
contain large thick fibers that are randomly oriented
to the epithelial surface. This means the GLCM cor-
relation signal is expected to drop on a longer scale.
The values of decay length for HD and K were ob-
tained by fitting the correlation data with a single
exponential decay function. The decay length values
obtained are: 3.7 0.1 mm for HD, and 6.8 0.1 mm
for K. The typical distance found is not exactly the
fiber bundle diameter, even if it is probably related
in some way to it. This fact is confirmed by the me-
asurement on K, which has exhibited a larger value.
Larger fiber bundles with respect to those of HD in
fact are typical of K. What we can assert is that the
keloid has a larger decay length for correlation, as
expected. Regarding the other two parameters,
homogeneity (Figure 3b) and energy (Figure 3c), K
shows larger values with respect to HD as expected.
In fact, K is expected to be a more homogenous tis-
sue with respect to HD and the graph in Figure 3b
confirms this expectation. Energy (also known as
uniformity) is also higher in K with respect to HD,
confirming the fact that K is a more uniform tissue
with respect to HD. Homogeneity and energy give
similar results because they are not completely inde-
pendent each other, even if they are related with
contrast and orderliness of the image, respectively.
In general, GLCM analysis was demonstrated to be
a powerful method to characterize the organization
of fibrillar collagen in SHG images.
3.3 Image-pattern analysis with FFT
on SHG images
The aim of the Fourier transform is to distinguish
between the frequency components of an image. The
intensities in our images are compared and a repre-
sentation of their periodicity within the image is giv-
en. This makes the Fourier transform of an image a
very good method of assigning a degree of organiza-
tion to this image. FFT analysis is pa rticularly helpful
in characterizing images of fibrillar collagen by their
degree of symmetry. In fact, for an image containing
Figure 3 Results of the GLCM analysis plotted versus the
neighbor index expressed in microns instead of number of
pixels after proper rescaling. Correlation (R) of HD (black
squares) and K (gray circles) versus neighbor index with
superimposed the exponential decay fits of data points for
HD (black line) and for K (gray line) (a). The two decay
lengths are reported in an inset. Homogeneity (H) of HD
(black squares) and K (gray circles) versus neighbor index
(b). Energy (E) of HD (black squares) and K (gray cir-
cles) versus neighbor index (c).
J. Biophoton. 3, No. 12 (2010) 39
2010 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Page 6
a set of aligned fibers, we expect a corresponding
FFT image with higher values along the direction
orthogonal to the direction of the fibers and its inten-
sity plot is expected to have an elliptic behavior. By
contrast, for an image with randomly oriented fibers,
the intensity plot of the corresponding FFT image
should show a circular behavior. The anisotropy of
the image can be evaluated by measuring the ratio of
the two axes of the ellipsis. In a previous work on cor-
nea [26], the authors performed an elliptic fit on the
thresholded FFT images. Here, we have used a
slightly different method to measure the aspect ratio.
In order to evaluate the length of the two axes of the
intensity distribution, even if they are not aligned to
the x,y coordinates we have calculated the covariance
matrix between rows and columns of the FFT image.
Then, by calculating the eigenvectors of the covar-
iance matrix, we kept as a measure of the aspect ra-
tio, the square root ratio of the two eigenvalues, cor-
responding to the two calculated eigenvecto rs. In
fact, the two eigenvalues of the covariance matrix re-
present the variance of the FFT intensity distribution
along the two main axes of the ellipsis. Their square
root ratio AR, ranging from 0 to 1, can give a meas-
ure of the anisotropy of the sample. The sample is
more anisotropic as the AR is close to 0, whereas it is
more isotropic when the AR is close to 1.
Because of the parallel arrangement of collagen
fibers within HD we expect to see different fre-
quency components with similar directions. This
means that the Fourier image of dermis should be
more elliptic than K. The results obt ained have con-
firmed what was expected (Figure 4): the FFT image
of HD has shown a more elliptic profile (Figure 4c)
with respect to the FFT image of K (Figure 4d),
which was found to have a more circular behavior.
The aspect ratio (AR) value, averaged on all the ex-
amined samples has shown a larger value for K
(0.59 0.09) with respect to HD (0.41 0.08) as
was expected. The small overlap of the error bars in
Figure 4e is due to the fact that we have used stand-
ard deviation of data, instead of standard error, as
error to take into account of biological variability.
The two values were found to be statistically diffe r-
ent at the 0.05 level after a two-sample statistical t-
test. This result confirms the fact that HD is a more
organized tiss ue with respect to K. In particular,
even if fibrillar collagen is organized in fiber bundles
in both HD and K, this fiber bundles differs between
HD and K in dimension (as we have observed by
GLCM analysis) and in their mutual organization. In
particular, collagen-fiber bundles were found to be
organized and to maintain a certain alignment over
a larger scale in HD with respect to K, which were
found to lose the organization between fiber bundles
on a shorter scale. In fact, HD maintain its anisotro-
py over a larger scale with respect to K, confirming
it as a more organized tissue.
4. Conclusion
We have teste d the capability of different image-pat-
tern analysis approaches to ch aracterize skin tissue
with some particular dermal disorder. In particular,
we have examined ex-vivo skin samples with HD, K,
and NS by means of a collagen/elastin content scor-
ing through measurement of the SAAID score. A
first characterization of connective tissue was done
using this method. SAAID scoring has resulted a
powerful method to investigate dermis at a large
scale (hundreds of mm). In fact, the collagen/elastin
ratio fluctuations have to be averaged on a large
Figure 4 (online color at:
Transversal optically sectioned ex-vivo skin sample, con-
taining healthy skin dermis (a) and keloid (b), imaged
using second-harmonic generation microscopy. Field of
view: 60 60 mm. Scale bars: 6 mm. FFT intensity images
obtained after 2D-DFT of the image of dermis (c) and ke-
loid (d). The FFT images in (c) and (d) correspond to the
2D-DFT of the images shown in (a) and (b), respectively.
The result of the aspect ratio for healthy dermis and ke-
loid, averaged on all the examined sample and on a differ-
ent window size for 2D-DFT calculation, are plotted in a
bar graph (e). The two values for HD and K were found
to be statistically different at the 0.05 level after a two-
sample statistical t-test.
R. Cicchi et al.: Scoring of collagen organization40
Journal of
2010 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Page 7
scale to get physiological information. Further analy-
sis on SHG images of collagen at a smaller scale was
performed on samples with HD and K. SHG images
were analyzed using the GLCM image-pattern ana-
lysis method and using statistical analysis in the
Fourier domain. These methods have highlighted
complementary differences in the organization of
collagen fiber. In particular, GLCM analysis was
found to be more powerful in investigating the intra-
fiber-bundle organiz ation (less than 10 mm scale),
whereas FFT analysis has given information on the
mutual organization of collagen-fiber bundles at a
larger scale (some tens of mm scale). We have seen
that, even if both HD and K have fibrillar collagen
similarly organized in fiber bundles, the dimension
of a fiber bundle differs between HD and K. We
were able to relate in some way the decay length of
correlation, calculated with GLCM analysis, to the
dimension of a fiber bundle. Moreover, by examin-
ing the organization of collagen-fiber bundles
through FFT analysis, we have found a larger aniso-
tropy for HD with respect to K. This result suggests
that HD fiber bundles are more organized and their
alignment is maintained over a longer scale with re-
spect to K. On the other hand, K has shown a more
isotropic behaviour of the FFT image, confirming
that collagen-fiber bundles are less organized than in
HD and they are randomly oriented inside the tis-
sue. Each of these methods presented in this work
represents a promising analysis/scori ng tool for con-
nective tissue. The analysis method is particularly
powerful if a cro ss-check of different approaches is
performed, because it gives information at different
scale levels. The combination of different image-ana-
lysis approaches presented here may represent a
powerful tool to investigate collagen organization
and remodeling as well as to monitor the effect of
topical therapies for the treatment of pathological
scars. Specifically, these techniques could be un-
iquely suited to better understand the regulation of
tissue repair in man as well as monit oring the clinical
efficacy of new pharmacological tools in pathological
wound healing.
Acknowledgements We thank the European Laboratory
for Non-linear Spectroscopy (Integrate Infrastructure In-
itiative “LASERLAB EUROPE” JRA-Optobio), Italian
Space Agency (ASI, MOMA project), and Fondazione
Ente Cassa di Risparmio di Firenze for financial support
to this project. We thank R. Ballerini and A. Hajeb from
the mechanical workshop of LENS, and M. Giuntini,
M. De Pas and A. Montuori from the electronic workshop
of LENS for their help in building the mechanics and
electronics of our experimental setup.
Riccardo Cicchi received his
degree in physics in 2003 at
the University of Florence,
working on skeletal muscle
myosin mechanics. In 2007
he got his Ph.D. in physics
at the University of Florence
with a thesis titled “Non-lin-
ear imaging of human skin”.
His research involved multi-
photon microscopy imaging of human skin tissue. Cur-
rently, he is a Post-Doc at the University of Florence.
His research involves multiphoton microscopy and life-
time imaging of diseased tissues.
Dimitrios Kapsokalyvas re-
ceived his degree in physics
in 2002, and the M.Sc. de-
gree in Optoelectronics in
2004, from the University of
Crete, Greece. He is cur-
rently conducting his Ph.D.
thesis, in LENS, in the Uni-
versity of Florence, on the
subject of ‘In vivo imaging
of skin’. His scientific goal is to develop new techniques
for imaging different aspects of human skin.
Vincenzo De Giorgi is working at the Department of
Dermatology, University of Florence.
Vincenza Maio received her
degree in medicine from
University of Pisa in 1999,
and residency in Anatomic
Pathology, University of
Florence in 2004. She is a
currently research Assistant
at the Department of Hu-
man Pathology and Oncol-
ogy, University of Florence
(20052009). Research project “To consolidate and im-
prove the quantity and quality of services in relation to
the introduction of specific technologies”.
Annelies van Wiechen suc-
cessfully graduated from the
European School III in
Brussels and received her
European Baccalaureate di-
ploma. She started her
study in Physics at Delft
University of Technology in
2005 and as part of her Ba-
chelor’s degree successfully
completed a research pro-
ject at LENS in 2009.
J. Biophoton. 3, No. 12 (2010) 41
2010 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Page 8
[1] W. Denk, J. H. Strickler, and W. W. Webb, Science
248, 7376 (1990).
[2] W. R. Zipfel, R. M. Williams, and W. W. Webb, Nat-
ure Biotechnol. 21, 13691377 (2003).
[3] A. Zoumi, A. Yeh, and B. J. Tromberg, P. Natl. Acad.
Sci. USA 99, 1101411019 (2002).
[4] W. R. Zipfel, R. M. Williams, R. Christie, A. Y. Niki-
tin, B. T. Hyman, and W. W. Webb, P. Natl. Acad. Sci.
USA 100, 70757080 (2003).
[5] B. R. Masters, P. T. C. So, and E. Gratton, Laser
Med. Sci. 13, 196203 (1998).
[6] P. T. C. So, H. Kim, and I. E. Kochevar, Opt. Exp. 3,
339351 (1998).
[7] J. C. Malone, A. F. Hood, T. Conley, J. Nu¨ rnberger,
L. A. Baldridge, J. L. Clendenon, K. W. Dunn, and
C. L. Phillips, J. Cutan. Pathol. 29, 453458 (2002).
[8] B. R. Masters and P. T. C. So, Opt. Exp. 8,29
[9] K. Ko¨ nig and I. Riemann, J. Biomed. Opt. 8(3), 432
439 (2003).
[10] M. J. Koehler, S. Hahn, A. Preller, P. Elsner, M. Zie-
mer, A. Bauer, K. Ko¨ nig, R. Buckle, J. W. Fluhr, and
M. Kaatz, Exp. Dermatol. 17, 519523 (2008).
[11] E. Dimitrow, I. Riemann, A. Ehlers, M. J. Koehler,
J. Norgauer, P. Elsner, K. Ko¨ nig, and M. Kaatz, Exp.
Dermatol. 18, 509515 (2009).
[12] E. Dimitrow, M. Ziemer, M. J. Koehler, J. Norgauer,
K. Ko¨ nig, P. Elsner, and M. Kaatz, J. Invest Dermatol.
129, 17521758 (2009).
[13] L. Moreaux, O. Sandre, S. Charpak, M. Blanchard-
Desce, and J. Mertz, Biophys. J. 80(3), 15681574
[14] P. J. Campagnola and L. M. Loew, Nature Biotechnol.
21, 13561360 (2003).
[15] P. J. Campagnola, A. C. Millard, M. Terasaki, P. E.
Hoppe, C. J. Malone, and W. A. Mohler, Biophys. J.
81, 493508 (2002).
[16] S. Roth and I. Freund, J. Chem. Phys. 70, 16371643
[17] K. Ko¨ nig, K. Schenke-Layland, I. Riemann, and U. A.
Stock, Biomaterials 26, 495500 (2005).
[18] R. M. Williams, W. R. Zipfel, and W. W. Webb, Bio-
phys. J. 88, 13771386 (2005).
[19] M. Strupler, A.-M. Pena, M. Hernest, P.-L. Tharaux, J.-
L. Martin, E. Beaurepaire, and M.-C. Schanne-Klein,
Second harmonic imaging and scoring of collagen in fi-
brotic tissues, Opt. Exp. 15, 40544065 (2007)
[20] P. Stoller, K. M. Reiser, P. M. Celliers, and A. M. Ru-
benchik, Biophys. J. 82, 33303342 (2002).
[21] T. Yasui, Y. Tohno, and T. Araki, J. Biomed. Opt.
9(2), 259264 (2004).
[22] Y. Sun, W. L. Chen, S. J. Lin, S. H. Jee, Y. F. Chen,
L. C. Lin, P. T. C. So, and C. Y. Dong, Biophys. J. 91,
26202625 (2006).
[23] P. J. Su, J. B. Hong, T. H. Li, R. J. Wu, C. K. Chou,
S. J. Chen, C. Hu, S. J. Lin, and C. Y. Dong, Opt. Exp.
17, 1116111171 (2009).
[24] V. Da Costa, R. Wei, R. Lim, C. H. Sun, J. J. Brown,
and B. J. F. Wong, Arch. Facial Plast. Surg. 10
[25] M. Han, G. Giese, and J. F. Bille, Opt. Exp. 13, 5791
5797 (2005).
[26] P. Matteini, F. Ratto, F. Rossi, R. Cicchi, C. Stringari,
D. Kapsokalyvas, F. S. Pavone, and R. Pini, Opt. Exp.
17, 48684878 (2009).
Daniela Massi received her
degree in medicine from the
University of Florence in
1993, and completed her re-
sidency in Anatomic Pathol-
ogy and Ph.D. in Oncology
in 1998 and 2003, respec-
tively. She was a visiting fel-
low at the Institute for Der-
matopathology, Thomas
Jefferson University, Philadelphia (USA) in 19951996.
She joined the Department of Human Pathology and
Oncology of the University of Florence in 2003 as an
Associate Professor. Current research interests include
inflammatory and neoplastic skin pathology.
Torello Lotti received his
MD degree in 1978 from the
University of Florence Med-
ical School, Florence, Italy.
Postgraduate training was at
the University of Florence
as a dermatology resident
and as a fellow followed by
a fellowship at the Institute
of Dermatology, London,
UK. Currently he is a Full Professor of Dermatology at
the Department of Dermatology and Venereology and
Director of the Postgraduate School of Dermatology
and Venereology. Since 2009 he has been the chair of
the “International Society of Dermatology”.
Francesco Saverio Pavone
obtained his degree in Phy-
sics in 1989 at the Univer-
sity of Florence. In 1990 he
became a Research Officer
at the European Laboratory
for Non-Linear Spectro-
scopy (University of Flor-
ence). In 1993 he obtained
a Ph.D. in optics at the Na-
tional Institute for Optics (Florence, Italy). Currently,
he is directing a research group working in the field of
biophotonics on single-molecule biophysics, microscopy
imaging-spectroscopy techniques, biomedical imaging,
and laser manipulation of bio-samples.
R. Cicchi et al.: Scoring of collagen organization42
Journal of
2010 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Page 9
[27] E. Brown, T. McKee, E. DiTomaso, A. Pluen, B.
Seed, Y. Boucher, and R. K. Jain, Nature Med. 9,
796800 (2003).
[28] S. J. Lin, S. H. Jee, C. J. Kuo, R. J. Wu, W. C. Lin, J. S.
Chen, Y. H. Liao, C. J. Hsu, T. F. Tsai, Y. F. Chen,
and C. Y. Dong, Opt. Lett. 31, 27562758 (2006).
[29] P. P. Provenzano, K. W. Eliceiri, J. M. Campbell, D. R.
Inman, J. G. White, and P. J. Keely, BMC Med. 4(1),
38 (2006).
[30] R. Cicchi, S. Sestini, V. De Giorgi, D. Massi, T. Lotti,
and F. S. Pavone, J. Biophoton. 1, 6273 (2008).
[31] S. J. Lin, R. J. Wu, H. Y. Tan, W. Lo, W. C. Lin, T. H.
Young, C. J. Hsu, J. S. Chen, S. H. Jee, and C. Y.
Dong, Opt. Lett. 17, 22752277 (2005).
[32] M. J. Koehler, K. Ko¨ nig, P. Elsner, R. Bu¨ ckle, and M.
Kaatz, Opt. Lett. 19, 28792881 (2006).
[33] R. Cicchi, D. Massi, S. Sestini, P. Carli, V. De Giorgi,
T. Lotti, and F. S. Pavone, Opt. Exp. 15, 1013510148
[34] J. Paoli, M. Smedh, A. M. Wennberg, and M. B. Eric-
son, J. Invest. Dermatol. 128, 12481255 (2008).
[35] H. Lodish, A. Berk, S. L. Zipursky, P. Matsudaira,
D. Baltimore and J. Darnell, Molecular Cell Biology
4th edn. (W. H. Freeman & Co., New York 2000),
pp. 979985.
[36] M. J. Buehler, Proc. Nat. Acad. Sci. USA 103, 12285
12290 (2006).
[37] T. Igarashi, K. Nishino, and S. K. Nayar, The Appear-
ance of Human Skin: A Survey, Foundat. Trends
Comput. Graph. Vis. 3, 195 (2007).
[38] I. Rocha-Mendoza, D. R. Yankelevich, M. Wang,
K. M. Reiser, C. W. Frank, and A. Knoesen, Biophys.
J. 93, 44334444 (2007).
[39] R. S. English and P. D. Shenefelt, Dermatol. Surg. 25,
631638 (1999).
[40] M. Hall-Beyer, The GLCM Tutorial Homepage, 2.10
>>> communicates important findings with
a high degree of novelty and need for
express publication, as well as other
results of immediate interest to the
solid state physics and materials science
>>> offers extremely fast publication
typically less than 14 days from
submission to online publication
. This
is definitely a world record for Letter
journals in solid state physics! Peer
review by two independent referees
guarantees strict quality standards.
physica status solidi RRL
Rapid Research Letters,
the fastest peer-reviewed
publication medium in
solid state physics!
2010.Volume 4, 12 issues.
Print ISSN: 1862-6254
Online ISSN: 1862-6270
The latest member
of the physica status solidi journals
For more information please contact Wiley Customer Service:
With a brilliant first Impact Factor of 2.147
(2008 Journal Citation Reports®, Thomson Reuters 2009)
pss RRL is the most highly cited journal publishing exclusively Letter articles in
Condensed Matter Physics.
(Germany, Austria, Switzerland)
(Americas, Europe, Middle East and Africa, Asia Pacific)
J. Biophoton. 3, No. 12 (2010) 43
2010 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Page 10
    • "In particular, SHG microscopy can provide morphological and functional images of anisotropic biological structures with high hyperpolarizability, such as collagen. Thanks to its high sensitivity to collagen, SHG microscopy has been used to reveal collagen organization in a variety of tissues, including human dermis, keloid, and tumor microenvironment [7] [8] [9]. Nevertheless applications to human cardiac tissues [5;10], and in particular to the relation between atrial collagen and AF [11;12], remain sparse. "
    [Show abstract] [Hide abstract] ABSTRACT: The assessment of collagen structure in cardiac pathology, such as atrial fibrillation (AF), is essential for a complete understanding of the disease. This paper introduces a novel methodology for the quantitative description of collagen network properties, based on the combination of nonlinear optical microscopy with a spectral approach of image processing and analysis. Second-harmonic generation (SHG) microscopy was applied to atrial tissue samples from cardiac surgery patients, providing label-free, selective visualization of the collagen structure. The spectral analysis framework, based on 2D-FFT, was applied to the SHG images, yielding a multiparametric description of collagen fiber orientation (angle and anisotropy indexes) and texture scale (dominant wavelength and peak dispersion indexes). The proof-of-concept application of the methodology showed the capability of our approach to detect and quantify differences in the structural properties of the collagen network in AF versus sinus rhythm patients. These results suggest the potential of our approach in the assessment of collagen properties in cardiac pathologies related to a fibrotic structural component.
    No preview · Conference Paper · Aug 2015
  • Source
    • "Grey-level co-occurrence matrix (GLCM) analysis was performed to assess stromal collagen fibre structure and organisation in HMD and LMD tissue samples. This method allows assessment of the level of structural collagen organisation, as previously described [28, 29]. GLCM was performed using a macro in ImageJ software (National Institutes of Health, Bethesda, MD, USA). "
    [Show abstract] [Hide abstract] ABSTRACT: Mammographic density (MD) adjusted for a women's age and body mass index (BMI) is a strong and independent risk factor for breast cancer (BC). While the BC risk attributable to increased MD is significant in the normal population, the biological basis of high MD (HMD) causation and how it raises BC risk remains elusive. We assessed the histological and immunohistochemical differences between matched HMD and low MD (LMD) breast tissues from healthy women to define which cell features may be mediating the increased MD and MD-associated BC risk. Tissue was accrued from 41 women undergoing prophylactic mastectomy due to a high BC risk profile between 2008 and 2013. Tissue slices resected from the mastectomy specimens were X-rayed, then HMD and LMD regions were dissected based on the radiological appearance. The histological composition, aromatase immunoreactivity, hormone receptor status and proliferation status were assessed, as were collagen amount and orientation, epithelial subsets, and immune cell status. HMD tissue had a significantly greater proportion of stroma, collagen and epithelium, and less fat than LMD tissue. Second harmonic generation imaging demonstrated more organised stromal collagen in HMD compared to LMD. There was significantly more aromatase immunoreactivity in both the stromal and glandular regions of HMD than LMD tissues, although no significant differences in levels of oestrogen receptor (ER), progesterone receptor (PR), and Ki-67 expression were detected. The number of macrophages within the epithelium or stroma did not change, however HMD stroma exhibited less CD206+ alternatively activated macrophages. Epithelial cell maturation was not altered in HMD samples, and no evidence of epithelial mesenchymal transition was seen, but there was a significant increase in vimentin+/CD45+ immune cells within the epithelial layer in HMD. We confirmed increased proportions of stroma and epithelium, increased aromatase activity, and no changes in the hormonal receptor or Ki-67 markers status in HMD tissue. The HMD region showed increased collagen deposition and organisation, and decreased alternatively activated macrophages in the stroma. The HMD epithelium may be a site for local inflammation as we observed a significant increase in CD45+/vimentin + immune cells in this area.
    Full-text · Article · Jun 2015 · Breast cancer research: BCR
  • Source
    • "Experimental techniques providing volume 3D images of the collagen structure of tissues do already exist in the literature. For example, we note the multiphoton fluorescence microscopy techniques678, which have the advantage to avoid any preparation of the sample for the observation of collagen due to its autofluorescence. Another technique allowing such observations is second harmonic imaging microscopy91011. "
    [Show abstract] [Hide abstract] ABSTRACT: The assessment of the three-dimensional architecture of collagen fibers inside vessel walls constitutes one of the bases for building structural models for the description of the mechanical behavior of these tissues. Multiphoton microscopy allows for such observations, but is limited to volumes of around a thousand of microns. In the present work, we propose to observe the collagenous network of vascular tissues using micro-CT. To get a contrast, three staining solutions (phosphotungstic acid, phosphomolybdic acid and iodine potassium iodide) were tested. Two of these stains were showed to lead to similar results and to a satisfactory contrast within the tissue. A detailed observation of a small porcine iliac vein sample allowed assessing the collagen fibers orientations within the medial and adventitial layers of the vein. The vasa vasorum network, which is present inside the adventitia of the vein, was also observed. Finally, the demonstrated micro-CT staining technique for the three-dimensional observation of thin soft tissues samples, like vein walls, contributes to the assessment of their structure at different scales while keeping a global overview of the tissue. Copyright © 2015 Académie des sciences. Published by Elsevier SAS. All rights reserved.
    Full-text · Article · May 2015 · Comptes rendus biologies
Show more