Anda di halaman 1dari 4

The 2011 Biomedical Engineering International Conference (BMEiCON-2011)

Counting Number of Sweat Glands Using


Image Processing
Narutchai Pakkaselevat, Supaporn Kiattisin, Waranyu Wongseree, Adisorn Leelasantitham
Technology of Information System Management Program, Faculty of Engineering, Mahidol University
25/25 Puttamonthon, Nakorn Pathom 73170, Thailand
narutchai1412@gmail.com, tom_kiattisin@hotmail.com, waranyu1979@hotmail.com, egadisorn@mahidol.ac.th
researched papers for related works or counting objects. For
example, medical device helping treatments of skin disease is
Optical Coherence Tomography (OCT) [4]. This device can
see objects in a high resolution to be clear but it is very
expensive. In case of microscope, it uses to inspect small object
on plate but it is not convenient for dislocation [5-10]. Many
papers have presented the use of camera for counting number
of the objects [11-24]. Even though, the resolution of the
camera is less than OCT but it has cheaper and easier than the
microscope for dislocation. However, counting number of the
SGs has not been reported using the image from camera.

AbstractThis paper presents counting number of the sweat


glands using image processing. This method uses a process of
image processing as follows. The cropped image of the sweat
glands is resized for an extensive size and it is converted to gray
image. Then, unsharp filter is used for this image to adjust
clearly the sweat glands. Quality of the image will be improved
for image enhancement using adaptive histogram equalization.
Finally, extended minima transform, i.e. the regional minima of
the H- minima transform, is used for counting number of the
sweat glands. The experimental results show that the accuracy,
sensitivity and precision are approximately at 70.31%, 77.38%
and 88.89%, respectively. It saves time to analysis of the sweat
glands.

In this paper, counting number of the SGs for following


treatment of disease of the SGs is presented through the use of
image processing. This method uses a process of image
processing as follows. The cropped image of the SGs is resized
for an extensive size and it is converted to gray image. Then,
unsharp filter is used for this image to adjust clearly the SGs.
Quality of the image will be improved for image enhancement
using adaptive histogram equalization. Finally, extended
minima transform, i.e. the regional minima of the H- minima
transform, is used for counting number of the SGs. It saves
time to analysis of the SGs.

Keywords-counting; sweat glands; image processing;

I.

INTRODUCTION

From statistics of Institute of Dermatology (2008 - 2010),


there have reported more than 50000 people per year [1] who
are abnormally skin diseases (e.g. dermatitis, acne, seborrheic,
sweat glands and other). Especially, sweat glands (SGs) are
one of these diseases for a danger of human body. Function of
the sweat glands will release sweat which is a heat out from
human body. This is an excellent method for mechanisms of
the human body. If the body cannot eject the sweat, then he
will get the effects of heat exhaustion, heat stroke,
hyperthermia or death. In another way, if the body is
hyperhidrosis, then he will cause a lose salt of the body. There
are many sweat glands in human body about two or three
millions which palms and soles have the most of sweat glands
but it cannot found at nail, bed lip, margins glands penis and
ear drum [2]. There are two types for the abnormal sweat
glands. Firstly, more flow of sweat will cause serious diseases
(sweat on palm) e.g. toxic thyroid gland, goiter, tuberculosis,
diabetes, heart disease and others. Secondly, less flow of sweat
will cause, for example, psoriasis, prickly heat, migraine and
others. In other words, if patients feel that they have the flow of
sweat for more or less than normally, then it will be the
aforementioned diseases [3]. Therefore, it has a very important
for treatments of the skin from the doctor.

II.

BASIC KNOWLEDGE FOR FOLLOWING TREATMENT OF


DISEASE OF SWEAT GLANDS

In Institute of Dermatology, Figure 1 shows a manual


method for finding SGs from patients whose foot is painted by
iodine and sprinkle powder. The patients stand to affix the
footprint on a paper. After that, medical technician captures an
image of the affixed paper, as shown in Figure 2, using a
camera. Then, such an image is loaded into a computer for
cropping the image size approximately at 1 cm2. Finally, the
medical technician uses his eye and his experiences for finding
and manually counting number of the SGs and he sends the
result of them to the doctor for diagnose. However, this manual
process has still some problem. For example, it may an error of
counting number of the SGs using the human eye in case of the
use for more time.

Nowadays, numbers of the SGs reported from Institute of


Dermatology normally are counted by human eye from an
image captured by a camera. However, there are two problems
of counting number of the SGs. Firstly, the captured image is
too small and not too clear, therefore; it may mistake for
counting them. Secondly, if glands of sweat overlap, then it
will cause difficult to count the glands. There have reported

978-1-4577-2190-8/11/$26.002011 IEEE
147

Figure 3. A proposed method of autoomatically counting number of the SGs


using imagee processing.
Figure 1. A manual method for the patientts of SGs.

A. Resizing image and conveerting RGB to gray image


It can be seen from Fig.33 that the cropped image is an
input of the proposed methodd. This image is resized for an
extensive size, as shown in Fiig. 4. It is converted from RGB
color image to gray image. Thee equation is
Gray = (0.2989 x R) + (0.55870 x G) + (0.1140 x B)

(1)

where R, G and B are red, greeen and blue colors, respectively.


p
of SGs.
Figure 2. An image of the affixed paper for the patients

III.

OGY
PROPOSED METHODOLO

Figure 3 shows a proposed method of automatically


counting number of the SGs using image proocessing. It can be
seen from Fig. 3 that the process can be desccribed as follows.
(a)

(b)

mage for (a) an extensive size and


Figure 4. Example of SG gray im
(b) its histogram.

B. Unsharp filter and adaptivve histogram equalization


The gray mage from Sectioon IIIA is processed by unsharp
filter to adjust clearly it. After
A
that, the contrast-limited
adaptive histogram equalization helps to enhance the contrast
of the gray-scale image. This histogram of the output region
approximately matches the histogram specified by the
distribution parameter. It uses a bilinear interpolation to
eliminate artificially induced boundaries. The contrast,
a
can be limited to avoid
especially in homogeneous areas,

148

amplifying any noise that might be present in the image [25],


as shown in Figs. 5(a) and 5(b).

(
(a)
(a)

(b)

Figure 5. Example of SG gray image using (a) adaaptive histogram


equalization and (b) its histogram
m.

C. Extended minima transform and labeliing


Figure 6 shows a result of using extended minima
transform which is a regional minima of
o the H-minima
transform [25]. H is a nonnegative scalar. This
T step considers
regional minima and adds H value into thaat regional. When,
addition of regional has intensity more than around. It decides
equal 0. It resembles background. But adddition of regional
has intensity less than around. It decides equal
e
1. Minima
means object is white color. Background iss black color [26].
As a result from the process of extended minima
m
transform,
the output result from Fig. 6 will be applied to a labeling
method for marking number of the SG
Gs. This method
considers the importance middle ground between
b
individual
foreground pixels and the set of all foregrround pixels [25].
The connection of each component is just defined
d
in terms of
a path and the definition of the path depennds on adjacency.
This method chooses an 8-adjacency for a mask
m
of image [25].
Figure 7 (a) shows an example of the result for manually
counting of number of the SGs. Figure 7 (b) shows an
example of the result for automatically couunting of the SGs
using image processing and the labeling methhod.

(
(b)
Figure 7. Examples of the results for (a)
( manually counting of number of the
SGs, and (b) automatically counting of
o the SGs using image processing and
the labeling method.

D. Verification of performan
nces
Table I shows verificationn of the SGs for the proposed
method which is measured in terms of four values as (1) the
true positive (TP) rate (numberr of the SGs correctly detected),
(2) the false positive (FP) ratee (number of non-SGs wrongly
detected as the SGs), (3) the faalse negative (FN) rate (number
of the SGs not detected) and (4) the true negative (TN) rate
(number of non-SGs correctlyy identified as non-SGs). This
paper will use three indicatorss to verify the performances in
this proposed method as folllows. Firstly, accuracy is the
overall successful rate of counting
c
number of the SGs.
Secondly, sensitivity is the percentage of the actual SGs that
are detected. Thirdly, precision is the accurate percentage of
the actual SGs that are dettected. From these quantities,
accuracy, sensitivity and preccision are defined as equations
(2), (3) and (4), respectively.
TABLE I.

Positive
Negative

Figure 6. The output result for using extend minima transform to count
number of the SGs.

VERIFICATTION OF THE SWEAT GLANDS

The Proposed Method


Present
Absent
True Positive (TP)
False Positive (FP)
False Negativee (FN)
True Negative (TN)
Accuracy = (TP+
+TN)/(TP+TN+FP+FN)

149

(2)

Sensitivity = TP/(TP+FN)
Precision = TP/(TP+FP)
IV.

REFERENCES

(3)
(4)

[1]
[2]
[3]
[4]

EXPERIMENTAL RESULTS

From the experimental results, we use ten images of the


SGs to test the proposed method based on the MATLAB
program. Table II shows the overall results of performances in
terms of three indicators i.e. accuracy, sensitivity and precision
using equations (2), (3) and (4), respectively. It can be seen
from Table II that the accuracy, sensitivity and precision are
approximately at 70.31%, 77.38% and 88.89%, respectively.
TABLE II.
Images

TP

[6]

[7]

THE OVERALL RESULTS OF PERFORMANCES


FP

FN

[8]

TN

Accuracy
(%)

Sensitivity
(%)

Precision
(%)

H28

110

31

32

63.58

77.46

78.01

H34

88

47

62.86

65.19

94.62

H43

71

15

77.17

82.56

92.21

H50

156

16

29

77.61

84.32

90.70

106.25

14.50

30.75

70.31

77.38

88.89

Average

[5]

[9]
[10]
[11]
[12]
[13]
[14]

V.

CONCLUSION

[15]

The proposed method can count the number of the SGs


using image processing. The results show that the accuracy,
sensitivity and precision are approximately at 70.31%, 77.38%
and 88.89%, respectively. The image processing can help to
improve clearly the SG image for suitable analysis to count the
total number of SGs. It also saves the time to count them for
recording and sending the data to the doctor analysis. However,
this proposed method can be improved for higher accuracy
than those results through the use of a new technique.

[16]
[17]
[18]
[19]
[20]
[21]

ACKNOWLEDGMENT

[22]
[23]
[24]

The authors would like to specially thank the Institute of


Dermatology, Bangkok for their helps in this project and
supports of the information of sweat glands.

[25]
[26]

150

http://www.inderm.or.th
http://www.cai.md.chula.ac.th/lesson/lesson4410/data/sweat.htm
C. Matene, Chewaje, magazine.
Y. Ueda, Maximum-intensity-projection images of human sweat gland
using optical coherence tomography,
K. Ishizu, Image Processing of Particle Detection for Asbestos
Qualitative Analysis Support Method -Particle Counting system Based
on Classification of Background Area,
K. Ishizu, Image Processing of Particle Detection for Asbestos
Qualitative Analysis Support Method Particle Counting by Using Color
Variance of Background,
S. Mao-jun, A New Method for Blood Cell Image Segmentation and
Counting Based on PCNN and Autowave,
K. Sonal, Automated Cell Counting And Cluster Segmentation Using
Detection And Ellipse Fitting Techniques,
H. Mohammad, Automatic Counting of Leukocytes in Giemsa-Stained
Images of Peripheral Blood Smear,
Geisa M. Faustino, Automatic Embryonic Stem Cells Detection and
Counting Method in Fluorescence Microscopy Images,
X. Pengyun, Computer assistance image processing spores counting,
Y. H. Toh, Automated fish counting using image processing,
Chandrappa D N, Face Detection in Color Images using Skin Color
Model Algorithm based on Skin Color Information,
Z. Li-hua, A New Algorithm of Counting Human Based on
Segmentation of Human Faces in Color Image,
D. Xingjing, Analysis of Detection and Track on Partially Occluded
Face,
M. Hu, Application of Rough Sets to Image Pre-processing forFace
Detection,
V. Boonyarat, Automatic Counting Method Using Fast Radial
Symmetry Transform,
J. Lippold, Automatic Counting of Fission Tracks using Object-Based
Image Analysis for Dating Applications,
H. Zhong-zhi, Counting Ear Rows in Maize Using Image Process
Method,
P. Wijethunga, Digital Image Analysis Based Automated Kiwifruit
Counting Technique,
H. Men, Counting Method of Heterotrophic Bacteria Based on Image
Processing,
Tim K. Lee, Counting Moles Automatically From Back Images,
Z. Ping, Grain Counting Method Based On Image Processing,
X. Ying, Research on an automatic counting method for steel bars
image,
Gonzalez, R.C., Woods, R.E., and Eddins, S.L., Digital Image
processing using MATLAB, New Jersey: Pearson, 2004
The Mathworks, Contrast Enhancement Techniques, July 30, 2008,
http://www.mathworks.com/help/toolbox/images/ref/imextendedmin.ht
ml