IEEE TRANSACTIONS ON INFORMATION TECHNOLOGY IN BIOMEDICINE, VOL. 13, NO. 5, SEPTEMBER 2009 721
Overview of Advanced Computer Vision Systems
for Skin Lesions Characterization
Ilias Maglogiannis, Member, IEEE, and Charalampos N. Doukas, Student Member, IEEE
Abstract—During the last years, computer-vision-based diagno-
sis systems have been used in several hospitals and dermatology
clinics, aiming mostly at the early detection of skin cancer, and
more specifically, the recognition of malignant melanoma tumour.
In this paper, we review the state of the art in such systems by
first presenting the installation, the visual features used for skin
lesion classification, and the methods for defining them. Then, we
describe how to extract these features through digital image pro-
cessing methods, i.e., segmentation, border detection, and color and
texture processing, and we present the most prominent techniques
for skin lesion classification. The paper reports the statistics and
the results of the most important implementations that exist in the
literature, while it compares the performance of several classifiers
on the specific skin lesion diagnostic problem and discusses the
corresponding findings.
Index Terms—Classification methods, computer vision, der-
moscopy, melanoma, pattern analysis, skin cancer.
I. INTRODUCTION
T
HE INTEREST of the biomedical scientific community
for computer-supported skin lesion inspection and charac-
terization has been increased during the last years. Skin cancer
is among the most frequent types of cancer and one of the
most malignant tumors. Its incidence has increased faster than
that of almost all other cancers, and the annual rates have in-
creased on the order of 3%–7% in fair-skinned population in
recent decades [1]. Currently, between 2 and 3 million non-
melanoma skin cancers and 132 000 melanoma skin cancers
occur globally each year. One in every three cancers diagnosed
is a skin cancer, and according to the Skin Cancer Foundation
Statistics, one in every five Americans will develop skin can-
cer in their lifetime [2]. The cutaneous melanoma, which is the
most common type of skin cancer, is still incurable. However,
when it is diagnosed at early stages, it can be treated and cured
without complications. The differentiation of early melanoma
from other pigmented skin lesions (e.g., benign neoplasms that
simulate melanoma) is not trivial even for experienced der-
matologists; in several cases, primary care physicians seem to
underestimate melanoma in its early stage [3]. The latter has
attracted the interest of many researchers, who have developed
systems for automated detection of malignancies in skin lesions.
The main design issues for the proper characterization of skin
Manuscript received December 4, 2007; revised July 23, 2008. First pub-
lished March 16, 2009; current version published September 2, 2009.
I. Maglogiannis is with the Department of Informatics with Appliances in
Biomedicine, University of Central Greece, Lamia 35100, Greece (e-mail:
imaglo@ucg.gr).
C. N. Doukas is with the Department of Information and Communication
Systems Engineering, University of the Aegean, Samos 83200, Greece (e-mail:
doukas@aegean.gr).
Digital Object Identifier 10.1109/TITB.2009.2017529
lesions concern the image acquisition, the image processing and
analysis, the feature extraction, and the classification method-
ology. This paper presents an overview of existing systems that
address the aforementioned issues. In addition, an evaluation
of state-of-the-art classifiers is presented in the context of skin
lesion characterization, and performance metrics are discussed
as well.
This review paper is organized as follows. Section II pro-
vides background information on the pathogenic mechanisms
of skin cancer in regards to visual differentiations, while
Section III presents the image acquisition and feature ex-
traction methods utilized in the literature. Existing classifica-
tion systems and their corresponding results are discussed in
Sections IV and V. Section VI presents results from the con-
ducted experiments concerning the performance evaluation of
different classifiers, and Section VII concludes the paper.
II. SKIN CANCER BACKGROUND INFORMATION
The skin consists of a number of layers with distinct func-
tion and distinct optical properties. White light shone onto the
skin penetrates superficial skin layers, and while some of it is
absorbed, much is remitted back and can be registered by a
digital camera. The stratum corneum is a protective layer con-
sisting of keratin-impregnated cells, and it varies considerably
in thickness. Apart from scattering the light, it is optically neu-
tral. The epidermis is largely composed of connective tissues.
It also contains the melanin-producing cells, the melanocytes,
and their product, melanin. Melanin is a pigment that strongly
absorbs light in the blue part of the visible and the UV spec-
trum. In this way, it acts as a filter that protects the deeper layers
of the skin from harmful effects of UV radiation. Within the
epidermal layer, there is very little scattering, with the small
amount that occurs being forward directed. The result is that all
light not absorbed by melanin can be considered to pass into the
dermis. The dermis is made of collagen fibers, and in contrast
to the epidermis, it contains sensors, receptors, blood vessels,
and nerve ends (see Fig. 1).
Pigmented skin lesions appear as patches of darker color on
the skin. In most cases, the cause is excessive melanin concentra-
tion in the skin. In benign lesions (e.g., common nevi), melanin
deposits are normally found in the epidermis [see Fig. 3(c)]. In
malignant lesions (i.e., melanoma), the melanocytes reproduce
melanin at a high, abnormal rate (see Fig. 2). While they and
their associated melanin remain in the epidermis, melanoma is
termed “in situ.” At this stage, it is not life threatening, and its
optical properties make it conform to those of the normal, highly
pigmented skin. When malignant melanocytes have penetrated
1089-7771/$26.00 © 2009 IEEE
Authorized licensed use limited to: The University of Manchester. Downloaded on November 4, 2009 at 05:26 from IEEE Xplore. Restrictions apply.