Melanoma-nevus differentiation by multispectral imaging

ranchocucamongabrrrΤεχνίτη Νοημοσύνη και Ρομποτική

6 Νοε 2013 (πριν από 3 χρόνια και 10 μήνες)

116 εμφανίσεις

Melanoma
-
nevus
differentiation
by
multispectral imaging


Ilze Diebele
a
, Ilona Kuzmina
a
, Janis Kapostinsh
b
, Alexander Derjabo
b
, Janis Spigulis
a


a
Inst.
o
f

Atomic Phys. and Spectroscopy, Univ. of Latvia, Raina Blvd. 19, Riga, Latvia, LV
-
1586
;

b
Latvian

Oncology
Centre
,

Hipokrata str. 4, Riga,
Latvia,
LV
-
1006



ABSTRACT


A clinical trial on multi
-
spectral imaging studies of malignant and non
-
malignant skin pathologies comprising 16
melanomas and 25 pigmented nevi was performed in Latvian Oncology Cente
r. Analysis of data obtained in the
spectral range 450

950 nm have led to a novel image processing algorithm capable to distinguish melanoma from
pigmented nevi. The proposed methodology and potential clinical applications are discussed.

Keywords:

melanoma
-

nevus differentiation,
multi
-
spectral imagin
g


1.

INTRODUCTION

Cutaneous melanoma currently represents 5% of newly diagnosed cancer in men and 6% in women. It is the leading
fatal illness arising in the skin and is responsible for 80% of deaths from skin
cancer.

Melanoma arises from the
malignant transformation of melanocytes at the dermal
-
epidermal junction or from the nevomelanocytes of dysplastic
melanocytic nevi
,

or
congenital nevomelanocytic nevus that become invasive and metastasize
1
.

By now only bi
opsy can determine exact malformation diagnose. Though biopsy can rise metastasizing
, therefore

noninvasive diagnostics is
preferable
.

The penetration depth of optical radiation
in the skin tissues depends on
wavelength
2
.

Diffuse reflectance from skin
prov
ides morphological information from different depths
,

and using

m
ultispectral imaging

camera it is
possible

to
get this information for
various

wavelength
s
.

B.
U
sing
the
multispectral imaging
,

Farina et al.
3
obtained that

at 940 nm
the reflectance of the
benign nevus is close to that of the surrounding skin, whereas pigmentation of the melanoma is
still detectable.

Consequently, NIR
-
image analysis might help to distinguish melanomas from other skin
malformations.

2.

METHODS AND EQUIPMEN
T


2.1

Experimental set
-
up

and processing algorithm

This
clinical trial comprised
41

pigmented lesions and was performed
in
Latvian Oncology Center
by multispectral
imaging in the spectral range 450


950 nm with 10 nm step.

Overall 16 melanomas and 25 nevi were investigated in
thi
s study.

The set
-
up contain
ed

multispectral imaging camera
Nuance EX

with additional halogen lamps and
Nicon

objective.
The
Nuance Imaging Module

contained the principal imaging components in a single compact enclosure: high
-
resolution, scientific
-
grade CC
D imaging sensor, solid
-
state
liquid crystal filter

with a polarizer, wavelength tuning
element, spectrally optimized lens and internal optics
4
.
The
Nuance program CRi was used to
acquire

average spectra
of pigmented and normal skin areas.

Overall

51
images

at different wavelengths were taken from
each
pigmented malformation. Before
the

measurement,
the multispectral image of white etalon reproducing the illumination spectrum was taken.

Optical density was
calculated by the
CRi Nuance

program as

OD(

)=
-
log(
I(

)/I
0
(

)
)







(1),

where I(

)


intensity of the skin
-
reflected light and I
0
(

)


intensity of the light reflected from the white etalon (at
the same distance).

Fig. 2 presents
the
spectra of optical density for healthy

skin

(red c
urves
) and melanoma
s

(blue c
urves
)
. To create an
algorithm for image processing
,

the f
o
llowing wavelength were
chosen
: 540 nm, 650 nm and 950 nm. 540 nm
corresponds to maximum
absorption

of blood

5
-
6
,
while at
650 nm
the greatest difference
between

melano
ma and
healthy skin
was observed
and 950 nm
is the longest available wavelength corresponding to the deepest penetration
under the skin surface
3
.


Figure 2.Optical density spectra for melanoma (blue
curves) and healthy skin (red curves
).

For differentiati
on
between melanoma and nevus,
the following algorithm
of parametric imaging
was
proposed
:






p=OD
650
+OD
950
-
OD
540






(2),

where

p

is
the
differentiation parameter,

OD
540

is
the
optical density at 540 nm, OD
650

is
the
optical density at
650 nm
and OD
950

is
the
optical density at 950 nm
.


3.

RESULTS AND DISCUSSI
ON


To illustrate results of this algorithm, Fig.3
compares

ordinary color RGB and
parametric

images of melanoma
s,

and
Fig.4
compares

RGB and parametric images of nevi
.
In the first c
ase of
Fig
.3
,

near melanoma the nevus is also seen
.
In all cases a
fter processing melanoma
had

notably
higher p va
lues than
the
skin around

malformation
,

but nevus
is

vanished

or has lower p values than healthy skin
.

Consequently, the criterion p>p
0
, where

p
0

is related to the
surrounding healthy skin, may be regarded as indication to melanoma.

After analysis of the 16 available melanoma multi
-
spectral image sets, 12 showed
clear response accordingly to the
above
-
mentioned criterion, and the 4 remaining cas
es also generally corresponded to that, but with some deviations.


Fig.3

RGB images and parametric
p
-
images of two melanomas.

Melanoma

Nevus

1

2


Fig.4 RGB images and paramet
ric

p
-
images

of
two pigmented
nev
i
.


4.

CONCLUSIONS


To conclude, the newly developed image processing methodology may be helpful for
in

vivo

melanoma
differentiation from nevus by distant optical biopsy. The first results show that sensitivity of this approach
may be up
to

90 %

or even higher
.
Further development of software being able to replace
the
visual assessment (Fig.3
, Fig.4
) by
automatized
selection
is necessary.
Additional studies by use of this methodology

would promote better understanding
of
the photophysiological

processes that take

place in
human skin and
its lesions.


A
CKNOWLEDGMENTS

The financial support of European Social Fund (grant
#2009/0211/1DP/1.1.1.2.0/09/APIA/VIAA/077) is highly appreciated.


REFERENCES

1.

Klaus, W. and Allen, J.R., [Fitzpatrick’s Color Atlas & Synopsis of
Clinical Dermatology],

McGraw
-
Hill
Professional,

New York, 308
-
309 (2009).

2.

Anderson, R., Parrish, B.S., Parrish, J.A., M.
D., “The Optics of Human Skin”,

J.
Invest
.

Dermat
.

77(1), 13
-

19
(1981).

3.

Farina, B., Bartoli, C., Bono, A., Colombo, A., Lualdi, M.,
Tragni, G. and Marchesini, R., “Multispectral imaging
approach in the diagnosis of cutaneous melanoma: potentiality and limits” Phys
.

Med
.

Biol
.

45, 1243
-
1254 (2000).

4.

[
User’s Manual for Nuance 2.4.],
Cambridge Research & Instrumentation, Woburn, 13
-
19

(200
7).

5.

http://omlc.ogi.edu/spectra/hemoglobin/summary.html

6.

http://omlc.ogi.edu/spectra/melanin/eumelanin.html


1

2