Paper Menu >>
Journal Menu >>
J. Biomedical Science and Engineering, 2010, 3, 576-583
doi:10.4236/jbise.2010.36080 Published Online June 2010 (http://www.SciRP.org/journal/jbise/
Published Online June 2010 in SciRes. http://www.scirp.org/journal/jbise
A robust system for melanoma diagnosis using heterogeneous
Khaled Taouil1, Zied Chtourou1, Nadra Ben Romdhane2
1UR CMERP National Engineering School of Sfax, Sfax, Tunisia;
2Faculty of Sciences of Sfax, Sfax, Tunisia.
Received 5 April 2010; revised 8 May 2010; accepted 15 May 2010.
Early diagnosis of melanoma is essential for the fight
against this skin cancer. Many melanoma detection
systems have been developed in recent years. The
growth of interest in telemedicine pushes for the de-
velopment of offsite CADs. These tools might be used
by general physicians and dermatologists as a second
advice on submission of skin lesion slides via internet.
They also can be used for indexation in medical con-
tent image base retrieval. A key issue inherent to
these CADs is non-heterogeneity of databases ob-
tained with different apparatuses and acquisition
techniques and conditions. We hereafter address the
problem of training database heterogeneity by de-
veloping a robust methodology for analysis and deci-
sion that deals with this problem by accurate choice
of features according to the relevance of their dis-
criminative attributes for neural network classifica-
tion. The digitized lesion image is first of all seg-
mented using a hybrid approach based on morpho-
logical treatments and active contours. Then, clinical
descriptions of malignancy signs are quantified in a
set of features that summarize the geometric and
photometric features of the lesion. Sequential for-
ward selection (SFS) method is applied to this set to
select the most relevant features. A general regres-
sion network (GRNN) is then used for the classifica-
tion of lesions. We tested this approach with color
skin lesion images from digitized slides data base se-
lected by expert dermatologists from the hospital
“CHU de Rouen-France” and from the hospital
“CHU Hédi Chaker de Sfax-Tunisia”. The perform-
ance of the system is assessed using the index area
(Az) of the ROC curve (Receiver Operating Charac-
teristic curve). The classification permitted to have
an Az score of 89,10%.
Keywords: Melanoma; Computer Aided Diagnosis System;
Segmentation; Feature Selection; Classification; Generalized
Regression Neural Network
Melanoma is the most deadly form of skin cancer. The
World Health Organization estimates that more than
65000 people a year worldwide die from too much sun,
mostly from malignant skin cancer .
The five-year survival rate for people whose mela-
noma is detected and treated before it spreads to the
lymph nodes is 99 percent. Five-year survival rates for
regional and distant stage melanomas are 65 percent and
15 percent, respectively . Thus the curability of this
type of skin cancer depends essentially on its early di-
agnosis and excision.
The ABCD (asymmetry, border, colour and dimension)
clinical rule is commonly used by dermatologists in visual
examination and detection of early melanoma . The
visual recognition by clinical inspection of the lesions by
dermatologists is 75% . Experienced ones with spe-
cific training can reach a recognition rate of 80% .
Several works has been done on translating knowledge
of expert physicians into a computer program. Computer-
aided diagnosis (CAD) systems were introduced since
1987 . It has been proved that such CAD systems can
improve the recognition rate of the nature of a suspect
lesion particularly in medical centres with no experience
in the field of pigmented skin lesions [7,8]. For these
systems to be efficient, the shots of the suspected lesion
have to be taken using the same type of apparatuses than
the one used for the learning database  and with iden-
tical lighting and exposure conditions. This could be-
come very challenging in the majority of cases.
In order to overcome the lack of standardization in
stand alone CADs and to provide an open access to der-
matologists, web-based melanoma screening systems
were proposed [10,11]. These systems have to consider
the heterogeneity in databases collected in different cen-
K. Taouil et al. / J. Biomedical Science and Engineering 3 (2010) 576-583 577
Copyright © 2010 SciRes. JBiSE
This work describes an enhanced CAD system that
addresses the problem of robustness of such tools under
the use of different databases.
2. PROPOSED CAD SYSTEM
The proposed software combines automated image seg-
mentation and classification procedures and is designed
to be used by dermatologists as a complete integrated
dermatological analysis tool. CAD systems in melanoma
detection are usually based on image processing and data
classification techniques. Five steps are generally needed:
data acquisition, pre-processing, segmentation, feature
extraction and classification (Figure 1).
2.1. Data Acquisition
The main techniques used for this purpose are the
epiluminence microscopy (ELM, or dermoscopy), tra-
nsmission electron microscopy (TEM), and the image
acquisition using still or video cameras. The use of
commercially available photographic cameras is also
quite common in skin lesion inspection systems, par-
ticularly for telemedicine purposes .
2.2. Segmentation of Lesion Images
Image segmentation is the most critical step in the entire
process. It consists of the extraction of the region of in-
terest (ROI) which is the lesion. The result of segmenta-
tion is a mask image. This mask is the base for the com-
putation of several shape and colour features.
The computer has a great difficulty in finding lesion
edge accurately. This task alone has formed the basis of
much research . The difficulty of segmentation is
due to low contrast between the lesion and the sur-
rounding skin and irregular and fuzzy lesion borders.
Artefacts (light reflections, shadows, overlapping hair,
Segmentation Feature extraction
Figure 1. CAD system in melanoma detection.
etc) can also give a false segmentation result. Some
works rely on the physician to outline the suspicious
We use a hybrid segmentation approach based on two
steps. The first consists in applying morphological
pre-processing filters to facilitate the extraction of the
approximate region of the lesion from the safe skin. The
second consists in applying active contour method on the
approximate mask to have the final contour of the lesion
Active contours or snakes are curves defined within
an image domain that can move under the influence of
internal forces coming from within the curve itself and
external forces computed from the image data. Snakes
were introduced by Kass et al. . Snakes are param-
v(s)x(s), y(s),s0,1 (1)
This curves move through the spatial domain of an
image to minimize the functional energy :
E(x,y)I(x, y) (4)
v(s) is a set of coordinates to form a snake contour.
v’(s) and v”(s) denote the first and second deriva-
tives of v(s) with respect to s .
α and β are weighting parameters that control re-
spectively the snake’s tension and rigidity.
is the gradient of grey-level image I.
A snake that minimizes Esnake must satisfy the Euler
αv"(s) βv""(s) E0
The internal force Fint discourages stretching and
bending while the external potential force Fext pulls the
snake toward the desired image edges.
To find a solution to (4), the snake v(s) is made dy-
namic by adding the parameter of time t to the equation
of the curve that becomes:
v(s,t)αv"(s,t) βv""(s,t)E xt
indicating how the snake must be modified at the instant
t+1 according to its position at the instant t.
When v(s, t) stabilizes, we achieve a solution of (6).
2.3. Features Extraction from Lesion Images
To characterize the different types of lesions we consider
a parametric approach. In such approach, the skin lesion
is resumed in a vector of features which dimension de-
pends on the number of extracted primitives. We use
578 K. Taouil et al. / J. Biomedical Science and Engineering 3 (2010) 576-583
Copyright © 2010 SciRes. JBiSE
The SFS  is an ascending research method (bot-
tom-up) of the set of most discriminative parameters
from an initial set of parameters (Ei) with:
quantitative parameters from the descriptions of derma-
tologists based on the ABCD rule to model clinical signs
The preliminary developed set of parameters under-
went a series of tests to evaluate their robustness when
quantifying multiple shots of the same lesion acquired
under different lighting conditions with different appa-
ratuses as is the case when using different slides from
heterogeneous databases . Our set of parameters was
besides used by  to develop an automatic recognition
of melanoma which reached a correct classification rate
The most important clinic signs that were kept to
characterize melanoma and the different lesions are the
irregularity of the contour, the asymmetry of colour and
shape, as well as the heterogeneity of the colour. We
classify parameters in two categories: geometric and
2.3.1. Geometric Parameters
Geometric parameters are extracted from the binary
shapes obtained after image segmentation. These pa-
rameters permit to characterize the shape of the lesion,
its elongation and the regularity of its contour. All these
parameters are standardized and independent of transla-
tion, zoom and rotation effects and therefore compensate
for rigid transformations introduced by the optics condi-
tions and the scene selection and framing.
2.3.2. Photometric Parameters
Photometric parameters are calculated from true colour
and binary images. These parameters permit to describe
homogeneity and symmetry of the colour as well as the
deviation between the mean colour of the lesion and the
mean colour of the surrounding safe skin. We tested dif-
ferent colour spaces representations calculated from the
red green blue components. We reflect the correlation
between the level of a colour component of a pixel and
its position on the digital image witch is supposed to be
independent of lighting conditions and spectral ٛensiti-
veity of the camera sensor.
2.4. Feature Selection
Feature selection allows choosing the most relevant pa-
rameter subset to perform the classification step. This
subset must contain the more robust and the most dis-
criminative primitives . Three criteria’s must be
fixed: the assessment method of the variables set rele-
vance, the research procedure to follow and the stopping
criteria of the selection. In this work we report the use of
a sequential forward selection with a stopping criteria
based on the minimum error generated by the classifier.
2.4.1. Sequential Forward Selection
j, j1,2, , N
EEi , nN
For this method one parameter pj is added at a time to
the ESFS subset. If the assessment criterion is an artificial
neural network, for each step, we insert one by one re-
maining parameters pj of Ei in ESFS and we calculate the
corresponding classification error (Err) with:
with q: total image number of the training database.
di: desired output.
ai: real output.
Initially, the subset =
For every step, parameter pj that will be selected is the
one for which the new ESFS subset permits to minimize
the classification error:
Err(E U)Err(E U')pj (8)
Thus, the first selected parameter is the most dis-
criminative one of the initial set of parameters. The se-
lection of parameters stops when while adding a new
parameter to ESFS, the classification error increases.
2.5. Classification of Lesion Images
After having summarized information contained in the
different images of our databases in vectors of parame-
ters we use a classifier based on a general regression
network (GRNN) . GRNN network is much faster to
train than a multilayer perceptron network (MPN).
GRNN gave better recognition rates than MPN for
melanoma classification . The architecture of this
network is illustrated on Figure 2. The GRNN is com-
posed of four layers: the input layer, the first intermediate
Hidden layer Summarized Layer
Figure 2. Architecture of the used GRNN.
K. Taouil et al. / J. Biomedical Science and Engineering 3 (2010) 576-583 579
Copyright © 2010 SciRes. JBiSE
layer constituted of radial units, the second intermediate
layer constituted of summarized units and the output
3. ASSESSMENT OF THE
The performance of our CAD system is evaluated in
term of sensitivity and specificity. These measures are
defined as follow:
# True Positives
#True Positives +#False Negatives
#TrueNegatives +#False Positives
With #true-Positive and #False-Negative corresponds
respectively to the number of malign lesions well classi-
fied and badly classified. #True-Negative and #False-
Positive corresponds respectively to the number of be-
nign lesions well classified and badly classified.
A ROC curve consists in representing the value of the
sensitivity according to (1-specificity) . The area
under the ROC curve or area index (Az) represents the
probability to correctly identify the image with anomaly
when an image with anomaly and an image without are
presented simultaneously to the observer.
We use two databases of image lesions whose malign
or benign nature is perfectly known after histological
analysis. The first one has been collected in CHU Rouen
France with the collaboration of the research laboratory
PSI-INSA Rouen and has been supported by the French
National League against Cancer. This database was digi-
tized in true colours by a 35 mms slides Nikon LS-
1000S scanner. It was used in previous works [19,25]
and [26,27]. We divide this database in a training (B0)
and test (B1) sets used in the first assessment of classifi-
The second image database has been collected in Tu-
nisia from the dermatology service of CHU Hédi
CHAKER in Sfax. Images were digitized in true colours
with a HP Scanjet 3570c scanner (cf. Table 1).
Our approach for efficiency assessment of the devel-
oped tool has been achieved in three steps:
3.1. First Assessment of the System
For the first step, we evaluate the diagnosis results of our
system while using two sets of images (B0 for training
and B1 for test) from the same CHU Rouen image data-
Table 1. Distribution of images databases.
base Set Class
B0 Training59 27 86
Rouen B1 Test 34 15 49
Sfax B2 Test 31 20 51
Total 124 62 186
3.2. Comparison of Our System’s Diagnosis with
Dermatologist’s Visual Diagnosis
This step consists in comparing diagnostic results of our
system with the opinion of four expert dermatologists for
the same test database (B1). Dermatologists are part of
the dermatology service of the CHU Sfax. We asked
every dermatologist to give his diagnosis for each lesion.
3.3. Third Assessment of the System
For this step, we evaluate our system while using the
second image database (B2) collected at CHU Sfax. This
test has been done while using the artificial neural net-
work having the best recognition rate according to the
4. RESULTS AND DISCUSSION
4.1. Results of the Segmentation Step
For image lesion segmentation, we propose a hybrid
method that combines the advantages of morphological
treatments, histogram thresholding and active contours
First the contrast of the gray level original image is en-
hanced using top-hat and bottom-up filtering (Figure 3).
The extraction of the image mask is based on the de-
tection of regional minima of the complementary image
of the contrast enhanced image. The detection of these
regions requires the application of a threshold. This
threshold is obtained with histogram thresholding using
Otsu method . We apply then a morphological ope-
ning on the obtained image. Lakes of the resulting image
are eliminated by filling holes. The approximate zone or
approximate mask of the lesion is finally obtained fol-
lowing a labelling, conservation of the biggest element.
We initialize a snake at the approximate boundary of
the safe skin (cf. Figure 4). The snake begins with the
calculation of a field of external forces over the image
domain. The forces drive it toward the boundary of the
lesion. The process is iterated until it matches the con-
tour of the lesion. We superpose the obtained contour on
the original color image.
Figures 5 and 6 show some examples of segmentation
results of lesions collected from both CHU Rouen and
HU Sfax databases. C
580 K. Taouil et al. / J. Biomedical Science and Engineering 3 (2010) 576-583
Copyright © 2010 SciRes.
(a) original image (b) contrast enhancement (c) Approximate mask of the safe skin
Figure 3. Extracting the approximate mask of the safe skin.
(a) Snake initialization (b) Snake progression (c) segmented image
Figure 4. Application of the active contour on the approximate mask.
ferent subsets of selected variables.
The chosen set of variables is the one that generates
the minimal error. We pursued research until the selec-
tion of all parameters. Then we chose the smallest subset
gotten with the minimal error.
The result of this selection method is illustrated on
Figure 7. It illustrates the variation of the classification’s
mean square error (MSE) according to the number of
included parameters, during trainings and tests of the
According to the test curve, we note that the set of the
Figure 5. Segmentation results of lesions collected from CHU
4.2. Results of Features Selection
For features extraction, a set of 68 parameters are ex-
tracted for every lesion. Through correlation and ro-
bustness study, a set of 42 parameters have been kept. To
find the most discriminative ones for the classification
step, we apply the sequential forward selection (SFS)
method. Training and test databases of images have been
For the SFS method, the assessment of parameter selection
is based on the comparison of the error generated by the
general regression neural network (GRNN) for the dif-
Figure 6. Segmentation results of lesions collected from CHU
K. Taouil et al. / J. Biomedical Science and Engineering 3 (2010) 576-583 581
Copyright © 2010 SciRes. JBiSE
Figure 7. Parameter selection using SFS method.
Figure 8. The ROC curve obtained using GRNN classifier and
the selected parameters.
Table 2. Order of the selected parameters using SFS method.
Rg N° Description
01 24 scG symmetry of the green component
02 09 rmoyr mean of the normalized red in the
03 03 DeltaD expanse of the distance to the
04 35 Beta spherical coordinate beta
05 23 scR symmetry of the red component
06 01 comp compactness of the shape
07 39 RB Ratio between lesion and safe skin
for the red
08 29 gamma1B Fisher coefficient for the blue
09 16 sigmamoyLb Standard deviation of normalized
bleu in the safe skin
10 06 scB symmetry of the blue component
most discriminative parameters permitting to minimize
the MSE of test database are gotten after the insertion of
the first ten parameters (MSE = 0.1434).
The selection by the SFS method permitted a reduc-
tion of 76.19% of the total number of parameters. The
list of parameters selected is presented in Table 2.
4.3. Results of the Classification Step
The classification is based on parameters kept in features
selection. Figure 8 illustrates results of the classification
while using the 10 most discriminative parameters se-
lected by the SFS method.
The performances of classification using B1 set of
images extracted from CHU Rouen database are based
on the comparison of the value of the area index (Az) of
the ROC curve. We obtain a value of Az equal to 89.1%.
The recognition rate of the system is 92.05%.
To validate the efficiency of our system, we compare
the obtained classification results with the diagnosis of
four dermatologists from CHU Sfax. We asked every
dermatologist to give his diagnosis for every lesion of
the same test set B1. Results of their diagnoses are given
in Table 3.
According to these results, we notice that the mean
value of sensitivity provided by dermatologists is equal
to the mean percentage of visual recognition of the true
positives by dermatologists that is 75% .
The recognition rate of our CAD system (92.05%) ex-
ceeds that obtained by dermatologists. It even exceeds that
of experimented trained ones which is about 80% .
Results of the second assessment of the system are
given in Table 4. We note that even when using a test set
of image lesions selected randomly from a different da-
tabase the recognition rate our system is 90.15%. It re-
mains better than the one of the visual diagnosis of ex-
In this paper we have described the different steps used in
the CAD system that we propose for melanoma detection.
To make this tool useful by the dermatologist commu-
nity outside specialized centres, each stage of processing
had to be automatic and robust to different conditions of
acquisition and apparatus. The system segments and
extracts parameters of description of the lesion. These
parameters are normalized and used as inputs for the
neural network classifier which decides if the lesion is
suspicious. We have also described the different steps
used for the evaluation of our system. This evaluation
had proved the robustness of our system when using
different databases in training and test. This property
makes it a suitable and an efficient candidate for use in a
context of a telemedicine dermatological application.
582 K. Taouil et al. / J. Biomedical Science and Engineering 3 (2010) 576-583
Copyright © 2010 SciRes. JBiSE
Table 3. Recognition rate dermatologists (D).
RR / D D A D B D C D D
Sensitivity 60 53,3 93,3 93,3 74,9
Specificity 64,7 100 6 4,7 76,5 76,4
RR 62,3 76,6 79,0 84,9
- FP (False Positive)
- TP (True Positive)
- FN (False Negative)
- TN (True Negative)
- RR: Recognition Rate
Table 4. Assessment of the system with database B2.
Recognition rate 90,15
We sincerely thank:
- Dermatology Services of CHU Rouen, France and CHU Hédi
Chaker, sfax, Tunisia.
- Laboratory PSI (Perseption Systems Information), INSA Rouen,
- Research Unit: Sciences and Technologies of Image and Tele-
communications, High Institute of Biotechnology, Sfax.
 (2006) Solar ultraviolet radiation: Global burden of disease
from solar ultraviolet radiation. Environmental Burden of
Disease Series, World Health Organization, 13.
 (2009) Cancer Facts and Figures. American Cancer So-
 Friedman, R.J. (1985) Early detection of malignant mela-
noma: the role of the physician examination and self
examination of the skin. Cancer Journal for Clinicians,
 Schmid-Saugeon, P., Guillod, J. and Thiran, J.P. (2003)
Towards a computer-aided diagnosis system for pig-
mented skin lesions. Computerized Medical Imaging and
Graphics, 27, 65-78.
 Blanzieri, E. (2000) Exploiting classifier combination for
early melanoma diagnosis support. Proceedings of the
11th European Conference on Machine Learning, Bar-
celona, 31 May-2 June 2000, 55-62.
 Cascinelli, N., Ferrario, M., Tonelli, T., et al. (1987) A
possible new tool for clinical diagnosis of melanoma.
Journal of the American Academy of Dermatology, 16,
 Ercal, F., Chawla, A., Stoecker, W.V., Lee, H.C. and
Moss, R.H. (1994) Neural network diagnosis of malig-
nant melanoma from color images. IEEE Transactions
on Biomedical Engineering, 41(9), 837-845.
 Ganster, H., Pinz, A., Röhrer, R., et al. (2001) Auto-
mated melanoma recognition. IEEE Transactions on
Medical Imaging, 20(3), 233-238.
 Burroni, M. (2004) Melanoma computer-aided diagnosis:
Reliability and feasibility study. Clinical Cancer Re-
search, 10, 1881-1886.
 Guillod, J., Schmid-Saugeon, P., Déccaillet, F., et al.
(2003) An open internet platform to distributed image
processing applied to dermoscopy. Studies in Health
Technology and Informatics, 95, 107-112.
 Oka, H., Hashimoto, M., Iyatomi, H. and Tanaka, M.
(2004) Internet-based program for automatic discrimina-
tion of dermoscopic images between melanoma and
Clark nevi. British Journal of Dermatology, 150(5), 1041.
 Maglogiannis, I. and Charalampos, N.D. (2003) Overview
of advanced computer vision systems for skin lesions
characterization. IEEE Transaction on Information Tech-
nology in Biomedicine, 13(5), 721-733.
 Hall, P.N., Claridge, E. and Morris Smith, J.D. (1995)
Computer screening for early detection of melanoma－
Is there a future? British Journal of Dermatology, 132,
 White, R., Rigel, D.S. and Friedman, R.J. (1991) Com-
puter applications in the diagnosis and prognosis of ma-
lignant melanoma. Dermatologic Clinics, 9, 695-702.
 Taouil, K. and Romdhane, N.B. (2006) A new automatic
approach for edge detection of skin lesion images. Inter-
national Conference on Information & Communication
Technologies: From Theory to Applications, Damascus.
 Kass, M., Witkin, A. and Terzopoulos, D. (1987) Snakes:
Active contour models. International Journal of Com-
puter Vision, 1, 321-331.
 Xu, C.Y. and Prince, J.L. (1998) Snakes, shapes, and
gradient vector flow. IEEE Transactions on Image Proc-
essing, 7(3), 359.
 Taouil, K., Bouhlel, M.S., Elloumi, M. and Kamoun, L.
(2002) Quantification des caractéristiques de mélanomes
en vue d'une classification. JTEA'2002, Sousse, 179-185.
 Zagrouba, E. and Barhoumi, W. (2004) A prelimary ap-
proach for the automated recognition of malignant mela-
noma. Image Analysis and Stereology Journal, 23(2),
 Taouil, K. and Romdhane, N.B. (2006) Automatic seg-
mentation and classification of skin lesion images. The
2nd International Conferences on Distributed Frame-
works for Multimedica Applications, Pulau Pinang.
 Pudil, P., Novoviovca, J. and Kittler, J. (1994) Floating
search methods in feature selection. Pattern Recognition
Letters, 15, 1119-1125.
 Specht, D.F. (1991) A general regression neural network.
IEEE Transactions on Neural Networks, 2, 568-576.
 Romdhane, N.B., Taouil, K., Boudaya, S. Turki, H. and
Bouhlel, M.S. (2007) Sélection des Variables et Classi-
fication par Réseaux de Neurones des Lésions Derma-
K. Taouil et al. / J. Biomedical Science and Engineering 3 (2010) 576-583 583
Copyright © 2010 SciRes. JBiSE
tologiques. 4th International Conference: Sciences of
Electronic, Technologies of Information and Telecom-
 Tokan, F., Türker, N. and Yıldırım, T. (2006) ROC
analysis as a useful tool for performance evaluation of
artificial neural networks. Artificial Neural Networks
ICANN 2006, 4132, 923-931.
 Zagrouba, E. and Barhoumi, W. (2005) An accelerated
system for melanoma diagnosis based on subset feature
selection. Journal of Computing and Information Tech-
nology, 13(1), 69-82.
 Vannoorenberghe, P., Colot, O. and de Brucq, D. (1999)
Dempster-shafer’s theory as an aid to color information
processing application to melanoma detection in derma-
tology. Proceedings of the 10th International Conference
on Image Analysis and Processing, Venice, 774-779.
 Lefevre, E., Colot, O., Vannoorenberghe, P. and de Brucq,
D. (2000) Knowledge modeling methods in the frame-
work of evidence theory: An experimental comparison
for melanoma detection Systems. 2000 IEEE Interna-
tional Conference on Man and Cybernetics, Nashville, 4,
 Sezgin, M. and Sankur, B. (2004) Survey over image
thresholding techniques and quantitative performance
evaluation. Journal of Electronic Imaging, 13(1), 146-