A FUZZY APPROACH FOR CELL COUNTING IN POORLY-ILLUMINATED IMAGES APPLIED TO A CELL-PHONE MICROSCOPE

weakassuredAI and Robotics

Nov 6, 2013 (3 years and 9 months ago)

73 views

A FUZZY APPROACH
FOR CELL COUN
TING

IN POOR
LY
-
ILLUMINATED


IMAGES
APPLIED

TO

A
CELL
-
PHONE MICROSCOP
E







A Thesis




Presented to the faculty of the Department of

Electrical and Electronic Engineering


California State University, Sacramento




Submitted in partial satisfaction of


the requirement
s

for

the

degree of




MASTER OF SCIENCE



in



Electrical and Electronic Engineering



b
y



Mehdi Rahimzadeh Soumesaraei






SUMMER

201
2
ii


A FUZZY APPROACH
FOR CELL COUNTING IN POORLY
-
ILLUMINATED

IMAGES APPLIED TO

A
CELL
-
PHONE MICROSCOP
E



A Thesis



b
y



Mehdi Rahimzadeh

Soumesaraei










Approved by:


______________________________, Committee Chair

Warren D. Smith


______________________________,
Second Reader

Stephen M. Lane


______________________________, Third Reader

V.

Scott

Gordon


_________________

Date

iii












Student:
Mehdi Rahimzadeh Soumesaraei



I certify that this student has met the requirement
s

for format contained in the University
format manual, and that this thesis is suitable for shelving in the Library and credit is to
be awarded for the thesis.




____________________________, Graduate Coordinator ________________

Preetham B. Kumar






Date




Department of Electrical and Electronic Engineering

iv


Abstract


of


A FUZZY APPROACH
FOR CELL COUNTING IN POORLY
-
ILLUMINATED



IMAGES APPLIED TO

A

CELL
-
PHONE MICROSCOP
E



b
y


Mehdi Rahimzadeh Soumesaraei




A b
lood cell count

is
a

common diagnostic tool in medicine
, and one way to
obtain such a count is
from
an image

of

a

blood s
mear
. Researchers at the Center for
Biophotonic
s

Science and Technology (CBS
T)
at

the
U
niversity of
C
alifornia,
Davis
have developed an attachment
to convert a cell phone to a microscope. The images
provided by this cell
-
phone microscope suffer from several artifacts
,

such as radial
distortion and non
-
u
niform illumination.
It is desired to
develop a

software
application

for a smart phone
to perform image processing and pattern recognition that can
return

an
approximate blood count.

In this work,
prototype

software has been
developed

on a personal compute
r (PC)
that performs the whole procedure of image processing and pattern recognition to
provide

a
n

approximat
e

red blood
cell
count
. To
do

the red blood cell count
, images

that are

taken
of a blood sample
by a smart phone are transferred to a PC for proces
sing.

Radial
distortion correction and cropping the
de
focused area of

the

image are done as pre
-
processing steps
in preparation

for robust
cell
recognition.
A
daptive multi
-
level
segmentation is performed as the second step to transf
orm

the image to a fuzzy scene
,

followed by
the
red cell recognition step.

A f
uzzy approach is
taken

for red cell recognition
.
The fuzzy approach presented
in this work utilized fuzzy sets and not fuzzy logic.
Adaptive image fuzzification and
v


fuzzy criterion

functions proposed in this thesis
have

higher performance than
conventional counting methods.
The proposed approach

is robust against fuzziness of the
image due to the poor

quality of
a
cell

phone
image, taken
under non
-
laboratory
conditions.
The

recognit
ion process in this application is a blind search method that is
independent of manual calibration and
learning.

Most of
this work

has been dedicated to enhanc
ing

the algorithm of cell
recognition even in poor
ly
-
illuminated images. This work
focuses on

red blood cell
counting. H
owever
,

the concept
can be extended to

other bloo
d smear

counting
,

such as
white blood cells and pla
telets
.
This algorithm is tested
on

seven blood
smear
images
,

and
the average
values for

precision and recall are 95.6
percent
an
d 95.4

percent
,
respectively.



__________________________________, Committee Chair

Warren D. Smith


______________________

Date





vi


ACKNOWLEDG
E
MENTS






I wish

to express
my

gratitude to
my

graduate advisor
, Prof. Dr.
Warren D. Smith
who
offered invaluable assistance, support
,

and guidance.
This
research

would not have
been possible without the guidance and the help of
Prof. Dr. Stephen Lane and
Assoc.
Prof.
Dr. Sebastian Wachsmann
-
Hogi
u
.
Deepest gratitude
is

also due to the member of
the
supervisory committee,
Prof.
Dr.
Scott Gordon.

I also wish to express my gratitude to
the graduate coordinator
,

Prof.
Dr.
Preetham Kumar
,

for his invaluable help and guidance
during the years of my
m
aster
’s

degree
.

Special thanks to all my dear friends who

have always been helpful to me.
Finally
, I would like thank my family for their support and encouragement throughout my
life.










vii





TABLE OF CONTENTS



Page


Acknowledgments ………………………………………………………………….



i楳i映 a扬e猠
……………………………………………………………………….



List of Figures ………………………………………………………………………

x

C桡灴敲


1. INTRODUCTION …………………………………………………………….

N

††††††
ㄮN

Overview ……………………………………………………………


N

††††††
ㄮN
.

B
汯潤⁃e汬⁃潵湴
……………………………………………………


O

††††††
ㄮN
.

佲ga湩na瑩潮o⁴桥⁒ m

n
楮i⁃桡灴e牳r
………………………….


.

P

㈮††⁂䅃䭇o何ka
……………………………………………………………..
.

R

††††††
㈮O
.

m牥癩潵v⁗o牫

………………………………………………………


R

††††††
㈮O
.

周楳⁗潲o

……………………………………………………………..

T

P
⸠†.
o䕖f䕗⁏c⁔ 䔠
䱉呅T
A
T啒r
……………………………………………

V

††††††
P

.

oa摩慬d
a
istortion Correction ………………………………………
⸮.
⸮.

V

††††††
P

.

c潣畳⁃物瑥物


Functions ……………………………………………
.

N
N

††††††††††††
P
⸲⸱
.

䝲ay sa物r湣e

……………………………………………….

N
P

††††††††††††
P
⸲⸲
.

呥ne湧牡d
⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮⸮.

N
4

††††††††††††
P
⸲⸳
.

p畭⁍潤u晩f搠ia灬ac楡渠
………
…………………………….

N
R

††††††††††††
P
⸲⸴
.

p畭⁍潤畬畳⁄楦ue牥nce
……
………………………………

N
S

††††††
P

.

fmage⁓
egmentation …………………………………………………..

N
T

††††††††††††
P
⸳⸱
.

䕤b
e
-
Based Methods ………………………………………...



††††††††††††
P
⸳⸲
.

oe杩g
n
-
Based Methods ……………………………………
⸮.

O
M

††††††††††††††††††
P
⸳⸲⸱
.

佴獵⁍整桯l

………………………………………
.

O
4

††††††††††††††††††
P
⸳⸲⸲
.

Minimum Error Method…..

……………………….

O
U

††††††††††††††††††
㌮㌮㈮P
.

c畺zy⁃
-
Mean Method ……………………………

P
O

††††††
㌮P
.

Image Fuzzification ……………………………………………………

P
P

viii





3
.
5
.

Object Recognitio
n and Counting ……………………………………..

4
2


3
.
5
.1
.

Pixel
-
Counting

Method ……………………………………...

4
4


3
.
5
.
2
.

Masking Method ……………………………………………..

4
6

4
.
ALGORITHM DEVELOPMENT …………………………………………….

5
0


4
.1
.

Pre
-
p
rocessing ………………………………………………………


5
0


4
.1.1
.

Radial Dis
tortion Correction ………………………………...

5
0


4
.1.2
.

Selecting the Most
-
Focused

Region

…………………………

5
2


4
.2
.

Image Segmentation
and Fuzzification
…………………
……………...

5
4


4
.3
.

Cell Co
unting …………………………………………………………..

6
0


4
.3.1
.

Affini
ty
F
unction …………………………………………….

6
4


4
.3.2
.

Homog
eneity Function ………………………………………

66


4
.3.3
.

Cell Recog
nition ……………………………………………..

68

5
. RESULTS AND DISCUSSION

………………………………………………

7
4


5
.1
.

Preprocessing
………………………………………………………….
..

7
4


5
.2
.

Image Segmentation and Fuzzification ……………………………….
.

77


5
.3
.

Cell Recognition ………………………………………………………
.

79


5
.
4
.

Performance
…………………………………………………………….

8
0

6
. SUMMARY, CONCLU
SIONS
, AND RECOMMENDATIONS …………..

86


6
.1
.

Su
mmary ……………………………………………………………….

86


6
.2
.

Conc
lusions ……………………………………………………………

86


6
.3
.

Recomme
ndations ……………………………………………………..

87

Appendix 1.
M
atlab Codes ……

..
…………………………………………

….

89

Appendix 2
.

The Cell Images ……
……………………………………………...
....

103

References
………………………………………………………………………….

1
07


ix




LIST OF TABLES


Tables



Page

1.

Table
5
.1

The comparison of the elapsed time for different

focusing criteri
on

functions .………………………………




O
.

呡扬攠
R
.
O

周T⁣潮晵獩潮慴o楸⁦潲⁴桥⁣e汬⁲ cog湩n楯渠

a
lgorithm …………………………………………………..


U
N

P
.

呡扬攠
R
.
P

䕶b汵慴楯渠潦⁴桥⁡lg潲楴桭

developed in this work ………

U
4

x




LIST OF FIGURES


Figures



Page

1
.

Figure
2
.1

The complex attachment developed at
UCB


to convert
a

cell phone to a microscop
e …………………
….

S



c楧u牥
O


周T敮
-
晲ee a瑴ac桭敮琠摥癥汯灥搠l琠啃i䄠景f
桯汯h牡phy⁩ ag楮g映扬潯搠獡浰mes

…………………
…..

S



c楧u牥
O


周T⁳y獴敭⁤sa杲g洠潦⁴me⁣e汬
-
灨潮p c牯獣潰o
摥癥汯灥搠l琠tBpT

…………………………………………

T



c楧u牥′
4

䅮⁩A桯湥′䜠 e汬⁰桯湥⁥煵楰灥搠q楴栠h⁢ 汬敮猠瑯s
c潮癥牴



瑯⁡楣t潳o潰eⰠ摥癥汯灥搠a琠tBpT

……………

U



c楧u牥
P
.
N

oe灲p獥湴慴楯渠潦⁲a摩慬d摩獴潲瑩潮渠o⁲ c瑡tg畬u爠

睩牥敳

…………………………………………………..

V



c楧u牥
P
.
O

䝲ay
-
sca汥⁶l物r瑩潮⁩渠摩g楴a氠l浡ges

……………………...

N
V



c楧u牥
P
.
P

周T⁲ 獵汴映 獳sg湭敮琠潦⁤楦晥re湴⁴桲h獨潬摳⁩渠

業age⁳eg浥湴m瑩潮

………………………………………..

O
O



c楧u牥
P
.
4

周T⁩ 灡c琠潦tgray
-
獣a汥lc潮瑲o獴⁩渠業a来

s
eg浥湴a瑩潮

……………………………………………….

O
4



c楧u牥
P
.
R

周T⁳ g浥湴m搠業a来映瑨攠獣a湮n搠扯潫d

⡆ig畲u″⸴ aF

潢瑡楮od

by a灰py楮g⁴桥⁴桲e獨潬d


T

=

㈱O

….
………………………………………………
⸮⸮


O
R

㄰N

c楧u牥
P
.
S

A

灬潴映瑨 ⁣物瑥物潮⁦u湣瑩潮⁩渠o桥⁍䔠浥瑨潤

………


P
M

ㄱN

c楧u牥
P
.
T

B牥a歩kg⁡⁨楳 潧牡洠m湴漠瑷漠䝡畳獩a渠摩獴物扵瑩潮猬o
牥灲pse湴n湧⁴桥 杲gy
-
sca汥⁤楳l物扵瑩潮o⁴桥扪 c瑳t

a湤⁢nckg牯r湤

…………………………………………….

3
1

12.

Figure
3
.
8

A chest x
-
ray image

………………………………………..

P
4

ㄳN

c楧u牥
P
.
V

周T⁲ 獵汴映獥g浥湴a瑩潮⁵獩湧⁡⁴桲e獨潬搠瑯d

楬汵獴牡瑥⁴桥⁥晦ec琠潦⁩湨潭潧ene楴y⁩渠 桥 gray

獣a汥

癡汵敳

o映瑨f扪散ts

…………………………………
.

3
6

xi




14.



Figure
3
.1
0



A linear fuzzy function that is extended from

the minimum to the maximum gray scale

values

…………
..

3
7





15.

Figure
3
.1
1

A fuzzy function that is linear in a limited range,

called
the
fuzzy range

………………………………………

39

16.

Figure 3.1
2

An illustration of the fuzzy region in the histogram

………..

4
0

17.

Figure 3.
13

A step
-
wise fuzzy
function

…………………………………

4
2

18.

Figure 3.
14

An

illustration of
(a)
broken and
(b) connected blood cells


after the segmentation process

……………………………...

4
4

19.

Figure 3.
15

A segmented image that contains 19
similar

rounded
o
bjects

……………………......................................

4
5

20.

Figure 3.
16

Illustration of
the masking method

…………………………

4
7

21.

Figure 3.
17

Identifying

cell
s

that are not circular

……………………
….

48

22.

Figure 3.
18

Over
-
counting
,

when a small mask is applied

…………
…...

49

23.

Figure 3.
19

A
guard distance




to provide

minimum distance

between

the centers of the detected cells

…………………
..

49

24

Figure
4
.
1

An

i
mage of the pattern of parallel

dark and bright straight
bars captur
ed by the cell
-
phone microscope
developed


at CBST

…………………………………………
…………

5
1

25
.

Figure
4
.
2

An image

captur
ed by the cell
-
phone microscope


developed at CBST

from a blood sample

………………….

5
5

26
.

Figure 4
.
3

Effect of

uneven illumination
……………………………...

5
6

27
.

Figure 4
.
4

The effect of
illumination variation on the histogram

……...

57

28
.

Figure 4.5

Partitioning an image
in
to nine sub
-
images and

assigning a distinct threshold to each of them

……………...


59

29
.

Figure
4
.
6

The threshold surface for the image shown


in Figure 4.3


……………………………………………

6
0





xii




30
.



Figure 4
.
7



A small portion of the red cell

image captured by

the cell
-
phone microscope developed
at

CBST

……………




6
2

31
.

Figure
4
.
8

A link in a fuzzy scene

……………………………………..

6
4

32
.

Figure
4
.
9

A one
-
pixel
-
thick ring
mask on a fuzzy scene

……………..

6
5

3
3
.

Figure
4
.1
0

A situation in which

the mask
spans across
cells …
………..

67

34
.

Figure
4
.1
1

The mask

sweeping over the fuzzy scene

……………….....

69

35
.

Figure
4
.
12

The feature space for

cell recognition formed


by the fuzzy functions, affinity and homogeneity
…………

7
0

36
.

Figure
4
.
13

A

flowchart of the algorithm that is developed

in this thesis (part

1)

………………………………………..

7
2

37
.

Figure
4
.
14

A

flowchart of the algorithm that is developed

in this thesis (part

2)

………………………………………


7
3

38
.

Figure
5
.
1

The distortion
-
corrected image of the bright and


dark bars

depicted in Figure 4.1

…………………………...

7
5

39
.

Figure
5
.2

The output images of the focusing criteri
on


functions using different methods

………………………….

76

40
.

Figure
5
.
3

The fu
zzified image of the blood cell image


shown in Figure 5.2
(d)

……………………………………

78

41
.

Figure
5
.
4

A

guard distance as long as the diameter of

the small masks

……………………………………………..

80

42
.

Figure
5
.
5

Finding the optimum affinity and homogeneity

thresholds for
the
cell detection region


in the feature space

……
…………………………………...

8
4

43
.

Figure
A
1
.1

Blood cell image #1

………………………………………...

103

44.

Figure A1.
2

Blood cell image #
2

………………………………………...

103

45.

Figure A1.
3

Blood cell image #
3

………………………………………...

104

46.

Figure A1.
4

Blood cell image #
4

………………………………………...

104

47.

Figure A1.
5

Blood cell image #
5

………………………………………..

105

xiii




48.



Figure A1.
6



Blood cell image #
6

………………………………………..



105

49.

Figure A1.
7

Blood cell image #
7

………………………………………...

106