BEST- LOCATION TECHNIQUE FOR MAMMOGRAM
AUTHENTICATION BY LSB BASED WATERMARKING
CIK FERESA BINTI MOHD FOOZY
UNIVERSITI TEKNOLOGI MALAYSIA
BEST- LOCATION TECHNIQUE FOR MAMMOGRAM AUTHENTICATION
BY LSB BASED WATERMARKING
CIK FERESA BINTI MOHD FOOZY
A thesis submitted in fulfilment of the
requirements for the award of the degree of
Master of Computer Science (Information Security)
Faculty of Computer Science and Information System
Universiti Teknologi Malaysia
OCTOBER 2009
iii
Special Dedicated to My Beloved Parents,
Mohd Foozy Bin Ghazali and Noraity Binti Nordin
My Brothers and Sisters,
Cik Farah Hani, Cik Fazleen, Mohd Farith and Fazly
My Supportive Friends,
Batch 9 Infosec, Shamala, ChaiWen, Sha, Zana, Huda, Ain, Ruby, Yus and Roslan.
iv
ACKNOWLEDGEMENT
IN THE NAME OF ALLAH, MOST GRACIOUS, MOST COMPASSIONATE
Alhamdullilah, thank you to Allah for giving me the blessing for health,
strength and earestness to accomplish and fulfill this project report.
I would like to take this opportunity to say the highest appreciation to Prof.
Dr. Azizah Abdul Manaf as my supervisor who always gives full sopport and
faithfulness in all guidance, advice and commitment upon on the effort for this
project. To Dr. Akram M. Zeki, who also gives full support and help me in this
project.
This appreciation also goes to my beloved family, lectures and all my friends
who always encourage and give me full support to finish this project. Indeed, the
courage and support is highly appreciated. Finally to Ministry of Higher Education,
University Tun Hussein Onn Malaysia (UTHM) and the family of Faculty of
Information Technology and Multimedia (FTMM) for both morale and financial
support for this course. May Allah Bless all of you.
Thank you.
v
ABSTRACT
This project has been developed to embed the patient information into 25
samples of mammogram images in a grayscale format. MATLAB Software has been
used as a tool to develop the algorithm where the algorithm has a capability to embed
and extract the hidden data. Since the algorithm will identify the breast area before
embedding process, the patient’s information will be embedded at the best location
which is at the background area without interrupting the breast area. Little distortion
at the background images are seen when embedding is using LSB 1 and LSB 2.
Meanwhile, few images are distorted when using LSB 3. However, there is no
distortion identified at all samples when using LSB 4 until LSB 8. Finally, based on
the PSNR result, it shows that the quality of watermarked mammogram images is in
high value and the watermarked mammogram images are in a good quality.
vi
ABSTRAK
Projek ini telah dibangunkan untuk memasukkan maklumat pesakit di dalam
25 jenis sampel imej mammogram yang berformat grayscale. Algoritma ini
dibangunkan dengan menggunakan perisian MATLAB untuk memasukkan
maklumat di dalam imej dan membaca kembali maklumat yang telah disembunyikan.
Oleh kerana, algoritma ini akan mengenalpasti kawasan dada sebelum prosess
memasukkan maklumat pesakit dilakukan, maka maklumat tersebut akan
dimasukkan di kawasan yang terbaik iaitu di kawasan latar belakang mammogram di
mana proses ini tidak meganggu kawasan dada pada imej mammogram. Sedikit
kerosakkan pada bahagian latar belakang imej boleh dilihat jika menggunakan LSB 1
dan LSB 2. Manakala beberapa imej mammogram akan mengalami kerosakkan jika
menggunakan LSB 3. Tetapi tiada kerosakkan dikenal pasti pada sampel apabila
menggunakan LSB 4 sehingga LSB 8. Akhir sekali, berdasarkan keputusan PSNR
yang diperolehi, ia menunjukkan bahawa kualiti imej mammogram yang telah
dimasukkan maklumat pesakit adalah tinggi dan kualiti mammogram tersebut adalah
baik.
vii
TABLE OF CONTENTS
CHAPTER
1
TITLE
PAGE
DECLARATION
ii
DEDICATION
iii
ACKNOWLEDGEMENT
iv
ABSTRACT
v
ABSTRAK
vi
TABLE OF CONTENTS
vii
LIST OF TABLES
x
LIST OF FIGURES
xi
LIST OF ABBREVIATIONS
xiii
LIST OF APPENDICES
xiv
INTRODUCTION
1
1.1
Introduction
1
1.2
Background of the Problem
6
1.3
Problem Statements
8
1.4
Project Aim
8
1.5
Project Objectives
9
1.6
Project Scope
9
1.7
Summary
10
viii
2
LITERATURE REVIEW
11
2.1
Introduction
11
2.2
Medical Image
11
2.3
Authentication of Medical
13
Image
2.4
Medical Image Watermarking
15
2.5
Medical Imaging Authentication
17
Application by Using Watermarking
Technique
3
4
2.6
Peak Signal to Noise Ratio (PSNR)
26
2.7
Issues in Typical mammogram
27
2.8
Summary
28
RESEARCH METHODOLOGY
30
3.1
Introduction
30
3.2
Project Methodology
30
3.3
Embedding Process
33
3.4
Extracting Process
34
3.5
Summary
35
PROJECT DESIGN AND TESTING
36
4.1
Introduction
36
4.2
Project Design
36
4.3
Activity Flow Diagram
38
4.3.1
Prototype Design
40
4.3.2
Browse Mammogram Module
40
4.3.3
Insert Data Module
41
4.3.4
Watermarked Module
41
4.3.5
Extracting Module
43
4.4
Algorithm Flow Chart
43
4.5
Best- Location Technique Algorithm for
48
Mammogram Authentication by LSB
Based Watermarking
4.6
Summary
49
ix
5
RESULT AND DISCUSSION
50
5.1
Introduction
50
5.2
Result in Identifying Breast Area
50
5.3
Result in Embedding Process
51
5.4
Result in Extracting Process
51
5.5
Result in PSNR Testing
51
5.5.1
60
Result in Attacking the
Watermarked of Mammogram
Image
5.6
6
Summary
60
CONCLUSION AND FUTURE WORK
61
6.1
Introduction
61
6.2
Contribution of the Project
61
6.3
Future Work
62
6.4
Summary
62
REFERENCES
APPENDICES A-B
64
70-75
x
LIST OF TABLES
TABLE NO.
TITLE
PAGE
2.1
Experiment Result of A.M et al. (2009).
24
2.2
Literature Review Table
25
5.1
Result of A Watermarked Mammogram Image of
52
LSB 1 until LSB 8 (Breast Area at Left side).
5.2
Result of A Watermarked Mammogram Image of
LSB 1 until LSB 8 (Breast Area at Right Side).
56
xi
LIST OF FIGURES
FIGURE NO.
TITLE
PAGE
1.1
Cranio-Caudal (CC) View
2
1.2
Mediolateral Oblique (MLO) Mammographic View
2
1.3
Medio Lateral (ML) Mammographic View
2
1.4
Radiology Department
4
2.1
Mammogram Medical Image
12
2.2
Symmetrical Organs (Molson Informatics Project,
13
McGill Faculty of Medicine Website)
2.3
MSE Formula
26
2.4
PSNR Formula
27
2.5
Radiograph Process for Medical Image
28
3.1
Project Methodology: Best- Location Technique For
32
Mammogram Authentication By Lsb Based
Watermarking
xii
3.2
Embedding Process: Best- Location Technique For
33
Mammogram Authentication By LSB Based
Watermarking
3.3
Extracting Process: Best- Location Technique For
34
Mammogram Authentication By LSB Based
Watermarking
4.1
Basic Watermarking Design
37
4.2
Best- Location Technique For Mammogram
38
Authentication By LSB Based Watermarking
Design
4.3
Activity Flow Diagram for the Prototype Design
39
4.4
Prototype Design
40
4.5
Patient Information
41
4.6
Watermarked
42
4.7
Extract Data
43
4.8
Algorithm Flow Chart
47
4.9
Categorize the Mammogram into Part A and Part B
44
4.10
LSB Bit Plane
45
4.11
Bit Plane that Represent Images
45
4.12
‘G’ in Binary
46
4.13
After Watermarked
46
xiii
LIST OF ABBREVIATIONS
NAME
DESCRIPTION
BMP
-
Bitmap
DWT
-
Wavelet Transform
DCT
-
Cosine Transform
JPEG
-
Joint Photographic Experts Group
LSB
-
Least Significant Bit
MSE
-
Mean Squared Error
PSNR
-
Peak Signal to Noise Ratio
xiv
LIST OF APPENDICES
APPENDIX
TITLE
PAGE
A
Summary Of PSNR Result
70
B
Watermarked Images
75
CHAPTER 1
INTRODUCTION
1.1
Introduction
As the technology drastically change especially in the medical environment,
the medical imaging technology such as mammogram, Magnetic Resonance Imaging
(MRI) and others are widely used in the medical domain for clinical purpose such as
medical procedure to diagnosis the patients’ internal problem without having the
surgery operation. For example, mammogram is used to identify the breast cancer
and it helps to detect the breast abnormality. Based on Pam Stephan (2007), all
women are recommended to have a baseline mammogram at the age of 40 and then
continued by a mammogram every couple of years until 50. Then after 50, a woman
should have a mammogram every year. Below are several types of mammogram
views that will be taken for screening:
i. Cranio-caudal (CC) is taken from above a horizontally compressed breast.
ii. Medio Lateral Oblique (MLO) is taken from the side and at an angle of a
diagonally compressed breast.
iii. Latero Medial (LM) is taken from the side towards the center of the chest.
iv. Mediolateral (ML) is taken from the center of the chest out. Spot
compression is compression on only a small area this is to get more detail
of the breast structure.
v. Cleavage view is a view from both breast compressed. This is to see the
tissue nearest center of chest.
vi. Magnification is to see borders of structures and calcifications.
2
Figure 1.1: Cranio-Caudal (CC) View
Figure 1.2: Mediolateral Oblique (MLO) Mammographic View
Figure 1.3: Medio Lateral (ML) Mammographic View
Since there are many advantages of medical images are discovered and it is
frequently used in the medical domain, most hospitals are facing with issues to
manage large amount of data storage such as administrative document, patient’s
information and medical images. Therefore, it is important to handle those data
accurately to avoid problem of lost, tampering and mishandling record at the hospital
(J. Nayak et al., 2008).
3
There are few ways to manage the patient’s information and the medical
images and one of them is using the watermarking technique in medical images. The
purpose of watermarking is to hide a message into the host of document or
multimedia format. As one to many communications technique, when sending the
watermarking image, only authorized user can read the hidden message. Thus, by
inserting watermarking in the medical image, it will help to authenticate the owner of
medical image. This is to ensure the medical image is issued from the right source
and to solve the missing medical document problem (G.Coatrieux et al.2008).
Moreover, there are many researches of medical watermarking images that
has been done such as C.S.Woo et al. (2006) proposed a multiple watermarking
method to store the medical images in a digital form and in a secure way in order to
avoid the data from being exposed to the unauthorized person. A. Giakoumaki et al.
(2006) proposed a wavelet based multiple watermarking schemes in the medical
images for secure and efficient health data management. D.C. Lou et al. (2009),
embedded large amount of data, maintained the quality of medical image and
restored the original image after extracted by using multiple layer data hiding in
spatial domain and Least Significant Bits (LSBs) technique.
Additionally, several advantages of watermarking in the medical image have
identified by S. Boucherkha and M. Benmohamed (2005) such as:
i. By embedding watermarking in medical image, the patient can protect
their information such as diagnosis result or personal details from being
viewed by unauthorized person.
ii. Medical image watermarking can help to authenticate the patient, if the
connection between the image and patient is lost.
iii. In addition, medical image watermarking will help the staff to identify or
search the old medical image in the Hospital Information System (HIS)
collection because there are some cases that medical images and patient’s
records need to be verified the integrity before use.
4
iv. Images may help to discover new findings in medical case. Thus, it is
needed to protect the copyright and integrity of the medical image by
digital watermarking.
v. Embedding the authentication code in the image will make it less
sensitive to attack than appending the information on medical image.
To get a clear scenario of medical imaging processes see Figure 1.4 for the
radiology department work flow. First of all, a patient will do a medical checkup
with the doctor or the medical assistant. If a patient has problem that needs to do a
medical imaging, the doctor will refer the patient to the consultant by writing a
request to him. Then, the consultant will accept the request and the patient can make
an appointment. If the consultant feels that patient needs to do medical imaging, the
patient will be referred to the physician by sending a request form to apply imaging
examination. The examination will be done by hospital personnel (nurse) under the
supervision of the radiologist. The result of the imaging examination is a set of films
that will be viewed by the radiologist. The radiologist will state his findings in a
written format and sent it to the referring physician. Finally, the referring physician
will perform the overall diagnosis and discuss the diagnosis treatment with the
patient (G.Muller, 2008).
Figure 1.4: Radiology Department (G.Muller, 2008).
5
As we know, hospitals are the place for the patients to get treatment and all
the medical records, patients’ information and others need to be managed properly.
Because of that, there are many systems are developed for the above purpose such as:
i.
Picture Archiving & Communications Systems (PACS)
ii.
Clinical Information Systems (CIS)
iii.
Radiology Information Systems (RIS)
iv.
Cardiology Information Systems (CIS)
v.
Radiation Therapy Systems
vi.
Patient Monitoring Systems
Picture Archiving and Communication Systems or PACS is a system for
storage, retrieval, distribution and presentation of images. PACS uses Digital
Imaging and Communication in Medicine (DICOM) format. The main purpose of
PACS is to manage hard copy of medical images like film archives. Next purpose is
to remote access for viewing and reporting or accessing the information from the
difference physical location. However, the medical images that stored on PACS have
been accessed every year thus the security and privacy of patients still in doubt.
Clinical Information system (CIS) is a system that built by IBM for patient’s
data management which is based on the patient’s information age. The system was
developed to replace the Medical Records Department of a medical institution, to
supporting the storage, manipulation, and distribution of clinical information
throughout the organization.
Moreover, Radiology Information Systems (RIS) is a solution which has
function of radiology billing services, comprehensive financial tracking and audit
trail, appointment scheduling, patient’s managed care data, powerful reporting and
patient’s database storage. Cardiology Information system (CIS) is a multi-modality
system that can be accessed by patients to access echo and nuclear cardiology, wave
forms, report and patient’s information. Radiation Therapy Systems is for radiation
treatment such as breast conservation therapy. Finally, for Patient Monitoring
6
system, the purpose of this system is to monitor the critical patient. The example of
the patient monitoring system is the heart monitor.
Most the applications that mention above are already implemented in some of
the hospitals and clinics. For authentication of medical image, watermarking
technique is proposed since there are few advantages that have been identified.
1.2
Background of the Problem
Currently, PACS has been used to handle the medical images and this system
integrates with several systems in the hospital but the result of diagnosis is written in
a piece of paper and the possibilities to lose the medical image or to be tampered by
unauthorized users that enter into the system may occur.
Since, the growing trends in the maintenance of medical records give an
impact to the security issues in the management of medical information. It creates the
interest among the researchers to explore the authentication of medical images.
Mohamed Kallelet et al. (2007) developed an algorithm for embedding multi
signatures using a multiple watermarking scheme to preserve the image history in the
medical field by using hash function and LSB technique to embed only the patient
diagnosis in medical images. Y.Li et al. (2007) did watermarking method using the
Discrete Cosine Transform Technique (DCT) to hide the information into the
mammogram background.
For patient authentication, A.W.T. Goh et al. (2008) studied about an
authentication of altering the image data and Rongrong Ni (2008) proposed new
algorithm to locate the tampered regions. Experiments show the result of watermark
medical image is PSNR=51.21dB. S.Boucherkha and M. Benmohamed, (2005),
examine 3 medical images of ultrasound in difference sizes and the result shows the
difference in PSNR (Peak Signal to Noise Ratio). G.Coatrieux et al. (2008), proposed
7
a method to ensure the authenticity and maintainability of data. From the findings
this method allows medical image managers to gather only the data of the same
patient without knowing the true identity of that patient. Since, patient identification
methods are different from country to country in Europe, the data that has been
embedded in medical images are the combination national health numbers and pivot
Id such as a family based identifier referring to family medical records.
In the research community, the watermarking technology is recognized by
them because of the medical confidentiality protection, origin and data authentication
that potentiality contributes to the medical information management systems (A.
Giakoumaki et al., 2006). Since, the medical data is stored in the system, the access
control to the system will be given only to the authorized user and when the access is
given, there is possibility that the patient’s information does not remain in one place.
Modification and lost can also be happen (G.Coatrieux et al., 2008). Thus, by
inserting the watermark in the medical image, the problem of lost, untraceable,
unintentional distortion and malicious modifications on the quality of medical
images can be prevented. Along with that, the medical image watermarking will be
able to trace the source or the owner of medical image (authenticate) and can detect
the changes that have been made in medical image (integrity). However, by
embedding the watermark on the medical image there will be possibilities that the
image will be affected when extracting the watermark is done. When the quality of
medical images is interrupted it can influence the diagnosis result.
There are few types of watermarking techniques such as Least Significant Bit
Hiding (LSB), Direct Cosine Transformation (DCT) and Wavelet Transformation.
LSB is the simplest method and S-Tools 4, HideBSeek, Steganos and StegoDos are
based on the LSB replacement (S. Huliiv et al., 2004). Implementation random LSB
technique hiddes the data in the least and second to least bits without being noticed
easily. A study on the LSB embedding technique by D.Neeta and K.Snehal (2005),
stated the result that LSB can hide the images in 24-bit, 8-bit or gray scale format of
a .png file and a .bmp file. Moreover, this paper mentions that the data can be hidden
in the least significant bits of the cover image and the human eyes would not be able
to notice the hidden image in the cover file. D.C. Lou et al. P (2009) also used the
8
LSB technique to embed a large amount of data and maintained the quality of
medical image.
1.3
Problem Statements
Based on the reviews on journals and papers that have done, it shows that the
authentication is one of the relevant elements that needs to be implemented in
watermarking of medical image application (G. Coatrieux et al., 2002) and the
quality of medical images are really important when embedding and extracting the
watermark. Below are the problem statements:
i.
Embedding the watermarking randomly will affect the quality of
mammogram image thus it will affect the diagnosis.
ii.
The breast area which contains important medical information should not
be modified during the watermarking.
iii.
Embedding the patient’s information in the best location would not affect
the critical areas or potion of mammogram.
1.4
Project Aim
The project aim for this research is to identify the best location on the
mammogram to embed the patient’s information without affecting the quality of the
image and to develop an authentication technique of watermarking mammogram
using the LSB technique.
9
1.5
Project Objectives
Below are the objectives of the project:
i.
To study and analyze the current technology of medical image
authentication.
ii.
To investigate the issues and problems in authentication technique in
mammogram.
iii.
To develop an authentication technique of watermarking mammogram
using the LSB technique.
iv.
To identify the best location on the mammogram to embed the
patient’s information without affecting the quality of the image.
1.6
Project Scope
Below are the project scopes of authentication medical image watermarking
project:
i. Twenty five (25) samples of mammogram images will be used to test the
algorithm for quality and authentication testing purpose at the extraction
process. The mammogram images samples are taken from Breast Cancer
Cases Molson Informatics Project, McGill Faculty of Medicine website.
ii. The mammogram images will be in grayscale format.
10
1.7
Summary
The advances in multimedia and communication technology nowadays have
provided new ways to store, access and distribute medical data in a digital format.
Hence, by identifying best location on the mammogram to embed the patient’s
information without affecting the quality of the image will be studied. The
development of authentication techniques for mammogram watermarking can be as
one of the alternatives to overcome the weaknesses of losing and mishandling of
medical images that will risk the patients. Moreover, comparison with old
mammograms is very important for breast cancer diagnosis (Diagnostic and
Interventional Radiology, 2008) and by embedding watermarking in medical image
will be easier to identify the medical image by authenticating the patient to prove that
the medical images belong to the correct person and the medical image is issued
from the trusted source.
CHAPTER 2
LITERATURE REVIEW
2.1
Introduction
In this chapter, the analysis will be done on medical image, authentication of
medical images, medical image watermarking and applications using several types of
watermarking technique. Other than that, review on current watermarking techniques
such as LSB, DWT and DCT, Peak Signal to Noise Ratio (PSNR) will be analyzed.
2.2
Medical Image
Medical imaging is one of the evaluations of the enhancement in health
industry where it gives many advantages to detect the disease earlier. There are many
types of medical imaging and one of them has their own purpose. For example MRI
is a medical imaging that scans inside human body and later it can show a tumor or
other problem in the brain where it can be viewed on a monitor or by hardcopy. A
special color can also be injected to show differences in the tissues of the brain.
Another example is CT scan that takes the picture of head and the function is to show
the tumors in the brain. This technology can also use special colors to inject the
picture to get the clear view of the tumor (Medindia.net, 2005)
.
12
Mammogram is one of the medical images that currently being used for
cancer diagnosis to detect symptoms such as lump, pain or nipple discharge. Based
on the medicinenet.com definition, mammogram is an image of the inner breast
tissues which visualizes on the film. Another definition of mammogram by Woman
to Woman website is an irregular or jagged white shape which is also called a mass
or a nodule that can be seen in the armpit on films. The white shape shows up best on
a background of gray or black, which is fat on a mammogram.
The mammogram will be in the grayscale format where the color value is in
range 0(black) to 256(white) and it has been identified as most of the efficient
screening method that can help to detect breast cancer earlier. In the earlier presented
of mammogram which was around 1969, this medical image was used for the breast
imaging only, then because of new development of technology during that time,
mammogram started to become a standard practice for a screening device in 1976 (S.
Luigi, 2005).
By doing mammogram, it can show or detect the symptoms or changes in the
breast up to two years before a patient or physician can feel it. Health officials are
recommending women to undergo mammogram tests every year beginning at the age
of 40. Additionally, women should also do self examination of the breasts for lumps
every month for early diagnosis of breast cancer (Medindia.net, 2005). It has been
agreed by the medicalimagingnews.com website where it mentions that the medical
imaging is also contributes major benefits to women.
Figure 2.1: Mammogram Medical Image
13
As mention in the previous chapter, there are few angles of breast image will
be taken such as Mediolateral Oblique View (MLO), Cranio-Caudal View (CC),
Medio-Lateral View (ML) and Latero-Medial View (LM) (Imaginis.com, 2008).
However, based on a website of Breast Cancer Cases Molson Informatics Project,
McGill Faculty of Medicine, there are six steps in viewing and analysis the
mammogram images. The first step is to place the correct orientation of the image
and the best view is in symmetrical organs where the right and left breast will be
compared in view box like Figure 2.2. This is an important step, which it will help
the physician doing mistakes that probably will happen when analysis the
mammogram image. Viewing the mammogram approach in the symmetrical organs
also has been agreed by Daniel B. Kopans on his book which is entitled Breast
Imaging.
Figure 2.2: Symmetrical Organs (Molson Informatics Project, McGill Faculty of
Medicine Website)
2.3
Authentication of Medical Image
Dictionary of Computing (2004) defined the authentication is a process by
which subjects, normally users, establish their identity to a system. It is also one of
the elements in the security to validate the authorized user since nowadays
technologies have become more advance and also provided alternative in storage and
distributing the information in digital format (Y. I. Khamlichi et al., 2003; Mingyan
Lia et. al., 2005). Many areas implement authentication in the application or system
such as e-commence, education, health and etc. this is because technologies also
14
affect on privacy and security so there are possibilities to get attacks or threats from
the communication technologies. Additionally based on Chen C.C and Ka D.S
(2008), digital images are easy to transmit, modify and reproduce. Ju-YuanHsiao et
al.(2009) said, since the digital forms are easy to be transmitted and altered, this can
contribute to the problems for people to authenticate the digital data’s content and its
origin as well as how to protect it from prohibited use and modification.
According to the article of website National E-Health Transition Authority
(NEHTA), currently the Internet has been used in the health sector as a medium to
transfer sensitive and important document such as patient’s details or diagnosis to the
other hospitals. Therefore, all sensitive documents which have been transferred need
to be ensured the authenticity and validity. In addition, medical imaging which is for
diagnosing internal disease of patient also has been embedded with authentication
techniques.
Authentications of medical images are also explored by many researches and
there are many advantages have been found until now especially in copyright
protection. The originality of an image can also be verified by detecting the
manipulations of malicious and the implementations of these techniques help to
ensure the integrity and confidentiality of medical folders (Jinshen Wang et. al.
(2004); S.Boucherkha and M. Benmohamed, (2005); Veysel Aslantas, (2008)).
Moreover, many types of authentication image have been applied and various studied
have shown, authentication can be implemented and beneficial in many areas like
P.M.Huang, et al., (2004) used half toning technique to develop a novel blockedbased binary image authentication. Chiou-Ting Hsu and Ja-Ling Wu (1996) did an
image authentication technique by embedding each image with a signature to avoid
unauthorized copying. Rongrong Ni et al. (2008) used watermarking which is based
on a chaotic system to propose a pinpoint authentication watermarking.
15
2.4
Medical Image Watermarking
Based on the Anastasios Tefas et al. (2005), watermarking is defined as a
practice of unnoticeably altering a piece of data in order to embed information about
the data. Watermarking is one of the techniques that can hide and hold the secret of
watermark for digital right management and copyright protection of the digital
images, videos or audios. This also has been agreed by Elham Shahinfard and
Shohreh Kasaei, (2001), Husrev T. Sencar and Nasir Memon, (2006) where they
identify that the purpose of watermarking can resolve the security issues such as
copyright protection, copy protection, fingerprinting and authentication.
Watermarking can be categorized in two types such as invisible watermark
and visible watermark. According to online dictionary askoxford.com website,
invisible is unable to be seen, either by nature or by concealed and the definition of
visible is able to be seen or noticed. Based on the study that has done by Veysel
Aslantas (2008), invisible watermark are unnoticeable. It is embedded as randomly
or nonrandom on the unknown places in the host data. The watermarked data have to
look similar with the original data and no one should know about the embedding.
Usually for invisible watermark, it will be used for copyright purpose. Furthermore,
for visible watermarks it is actually easily to be detected like company logos that
place into or cover on some products and it can be recognized by their owner.
Moreover, the owner can also do anything to the watermarking image such as
embedding, removing or destroying the watermark anytime that they want.
Additionally, invisible watermark can be divided into three types which are semifragile, fragile and robust.
Thus, watermarking is preferred as a favourite technique in hiding
information by the researches to enhance the security and privacy in the hospitals and
also to manage the patients’ data since it still poses some issues regarding the image
efficiency when designing such system. There are many advantages that have been
recognized and many areas have implemented watermarking techniques. Below the
applications that have been listed by N. Liu, P. Amin et al. (2006).
16
a. Copyright protection.
b. Unauthorized modification
c. Fingerprinting.
d. Annotation and indexing
e. Copy control
f. Medical applications
g. Broadcast monitoring
h. Covert communications
Study on the medical image watermarking has been done in medical
application regarding copyright protection, patient management and etc. Moreover,
there are many advantages of watermarking that are already known by other people
and currently are acknowledged as one of the important feature in the health care
system. Veysel Aslantas (2008) also agreed that watermarking has many advantages
in medical environment. Additionally, in future medical applications, it will be
predicted to have the integration access with the generalized databases that contain
the personal medical information of each patient (Ng Lee Ping et. al., 2003; Marie
Babel a et al.,2008).
Furthermore, there are many types of watermarking such as Spatial Domain,
Wavelet Transform (DWT) and Cosine Transform (DCT) each of these techniques
have their own advantages and some reviews have done regarding watermarking
techniques. Below are types of watermarking by N. Liu, P. Amin et al. (2006).
i. Spatial Domain
a. Modification Least Significant Bit (LSB)
ii. Transform Domain
a. Wavelet Transform (DWT),
b. Cosine Transform (DCT)
c. Fractal transform and others.
iii. Feature Domain
a. Spatial Domain
b. Transform Domain
17
LSB technique or Least Significant Bit is a technique that derives from the
Spatial Domain (N. Liu, P. Amin et al., 2006). According to Y. I. Khamlichi et al.
(2003), LSB technique is fragile and well known method plus efficient for inserting
data with a high embedding capacity. Moreover, LSB bit plane for each pixel of the
image are generally considered as noise because of the imaging device that has been
used. Thus, this technique can be used to embed secret message and patient’s
information without disturbing greatly the appearance of the image.
Many studies have done regarding the image watermarking by using LSB
technique for example F. Caoa et al. P. (2002) using LSB technique in the
development of medical image security using digital envelope. Y. I. Khamlichi et al.
(2003) determined if the watermarked image has been tampered with or modified.
S.Boucherkha and M. Benmohamed. (2005) embed the watermarking using this
technique. Mohamed Kallelet al. (2007) developed an algorithm for embedding
multi-signatures using a multiple watermarked scheme to preserve the history of the
genuine image. D.C. Lou et al. (2009) also used LSB technique in their project to
develop information hiding techniques for protection of medical information and
copyright. Therefore, LSB technique can be accepted to be used as the watermarked
technique since based on the result of the experiments that has done by the
researches above shows the successful result and they achieved the objectives of
their research.
2.5
Medical Imaging Authentication Application by Using Watermarking
Technique
In this section, techniques of watermarking that have been used by the
researchers for medical images will be reviewed. Boucherkha.S and Benmohamed.M
(2005) used LSB techniques where three medical images of ultrasound in difference
sizes of 256x256, 400x268 and 512x512 were examined. All images are in BMP
format. The watermarking that has been embedded in all ultrasound is 32 ASCII
characters of patient information which is 128 bits of Media Access Control (MAC).
18
The patient’s information consists of first name, family’s name, age and sex. Three
processes that involve in this research are:
i. Embedding process.
ii. Extraction process.
iii. Verification process.
From the experiment, information that embeds in all the images was
unnoticeable and it is successfully decoded. The result shows the difference in PSNR
(Peak Signal to Noise Ratio) where:
i. Ultrasound 1= 39.72,
ii. Ultrasound 2= 52.27
iii. Ultrasound 3= 41.33
A. Giakoumaki et al. (2006) implemented 4 levels of discrete wavelet
transform (DWT) in this research. Based on this research, there are 20 types of
medical images of CT, MRI, MRA and PET of 512x512 pixels used to test the
proposed scheme. The paper proposed a wavelet-based multiple watermarking
schemes to address the critical health information management issues that related to
the origin and data authentication, protection of sensitive data, and image archiving
and retrieval. Three types of watermarking were used such as:
i. Signature watermark
Contain physician’s digital signature or identification code
ii. Index watermark
Contain keyword such as ICD-10 diagnosis code, image
acquisition characteristics and etc.
iii. Caption watermark
Contain patient’s personal and examination data such as
diagnosis report.
19
iv. Reference watermark
Compare the extracted reference watermarked bit with the
original embedded watermark.
Moreover, the processes of watermarking for this paper consist of two
processes of embedding and extracting. The locations to embed the text watermarks
are based on a random key and the ROI map. There are three types of watermarks
that have been used such as signature, index and caption. For the watermark, there
are 52 and 208 characters of keywords have been used whereby the length is 364 and
1456 bits. Furthermore for signature length is 128 bits. For the future work, A.
Giakoumaki et al.(2006), will involves exploiting the wavelet-analytic nature of the
algorithm to derive image inherent characteristics in order to accommodate content
based image querying. The results show in PSNR (dB) measurement and the
percentage of bit errors are shown below:
i. CT 46.47 ± 0.06
ii. MRI 46.37 ± 0.05
iii. MRA 45.96 ± 0.04
iv. PET 46.66 ± 0.20
From the paper of C.S.Woo, J.Du, and B.Pham, (2006), these researches have
employed the digital watermarking technique and proposed a multiple watermarking
method to store the medical images in a digital form and in a secure way in order to
avoid the data from being exposed to the unauthorized person. The proposed method
consists of:
i. Annotation part
a. Annotation watermark is embedded into the border pixels.
b. Watermark message is placed in the frame pattern.
c. After that, embed using linear addictive method into three high
band of DWT.
d. Inverse DWT is done to get back the marked of image border.
20
ii. Fragile part
a.
Fragile watermark is embedded into the central of region of
the original image using LSB technique.
b. Combine annotation watermark and fragile watermark and
then the marks are embedded to be a multiple watermarked
image.
iii. Watermark detection
a. Detection will do separately which is similar with embedding
process.
b. For Annotation watermark, the border of watermarked
decomposed into DWT sub bands. Then calculate the
correlation value using three high pass bands.
c. For fragile watermark, it is detected by using LSB detection
method where each of the pixels is read to form the tiled
binary watermark pattern.
Encrypted patient’s data can be embedded in an annotation while the
tampering can be detected using fragile watermark. The embedded patient’s data not
only saves storage space, it also offers privacy and security. Furthermore the medical
images in digital form must be stored in a secured environment to preserve the
patient privacy (Woo et. al, 2006). Three experiments on the X-ray image of the
chest, MR image of the skull and CT image of the brain have done, however the
disadvantages can be found where there are possible to destroy the annotation
watermark on purpose using malicious attack technology. To solve it, annotation
watermark in textured regions of the image can be used as watermark instead of
image borders.
Mohamed Kallelet al. (2007) developed an algorithm for embedding multisignatures using a multiple watermarking scheme. The purpose of this scheme is to
preserve the history of the genuine image of medical image. The result from the
experiment shows PSNR is 51.21dB and a technique that used in this scheme is LSB.
Overall, there are 4 processes in this scheme such as:
21
i. Embedding process
ii. Multi-watermarking
iii. Signature extraction of the preceding doctor
iv. Embedding signature by the following doctor
Y.Li, C.T.Li, and C.H.Wei, (2007) studied about the framework for
combating security issues in PACS for mammograms that involve the watermarking
method to hide the patient’s information into the mammogram without changing her
important details and the watermark can also extract the hidden information from the
stegano-mammogram without the availability of the original image and only the
authorized review can remove the watermark. Other than that, this scheme is
compatible with the mammogram transmission and storage on the Picture Archiving
And Communication Systems (PACS). Based on the experimental that has done the
evaluation of capacity, embedding distortion and content masking have been
evaluated. There are two processes in this scheme such as:
i. Forward process
a. Segmentation
b. Segment mammogram into breast and background part
ii. Information Hiding
a. Mammogram is partition into blocks which each is 8x8
pixels.
b. Breast area is called medical blocks whereby background
is marked as background blocks.
c. Contain algorithm of block pairing and algorithm of
information embedding.
iii. Content Masking
a. This section will execute the process of mask content for
mammogram.
22
iv. Backward process
a. Distribute private keys
b. Apply inverse computation of equation content masking
c. Extract patient information
G.Coatrieux et al. (2008) discussed on the combination of watermarking with
different identifiers ranging from DICOM standard UID to an Anonymous European
Patient Identifier in order to improve medical. The combination of codes for
authenticity of DICOM Image identifiers codes (UID) and Patient identifiers (Id)
were reviewed. Since there are only reviewed the scheme, no experiment has been
done. The purpose of this paper is to combine an anonymized pivot number identifier
with national patient identifiers to guarantee the privacy and interoperability of data.
Moreover, this method helps to provide a solution of the problems in identification
and lost in medical documents.
D.C. Lou et al. (2009) has done a comparison study with Tian’s difference
expansion method to develop the information hiding technique. Both researches are
using the LSB technique but the diverse between this research is J. Tian(2002) which
he did the experiment on Leena image. Moreover, from the experimental results that
have done by D.C. Lou et al. (2009), it shows that the scheme should be able to
embed the watermarking better than J. Tian(2002). Purpose of D.C. Lou et al. (2009)
study is to protect medical information and copyright and a multiple-layer data
hiding technique in spatial domain (LSB). The experiment is using three different
types and textures of medical images in the same sizes of 512 x 512 are being used.
The processes in this experiment involve:
i.
Reviewed the Tian’s embedding method.
ii.
After reviewing has been done, D.C. Lou et al. (2009)
proposed the multiple-layer embedding method that consists of
three processes such as:
a. Embedding process
•
Uses a transformation function to reduce the
value of the expansion difference.
23
b. Extraction process
•
Has reversible characteristic after extracting the
data
c. Performance analysis
•
Uses PSNR to measure the quality of picture.
From the result by using D.C. Lou et al. (2009) proposed technique, a
remarkable improvement in the visual quality of the embedded can be satisfied
whereby the experimental results show the proposed scheme gives higher embedding
capacity and the level image quality compared than Tian's difference expansion
method. According to this paper, D.C. Lou et al. are focusing to enhance the
expansion method of Tian. Types of medical images that have been used in the
experiment are CT scan and MRI. The results show the advantages of the research
are more on the quality not on the security purpose. Moreover, in this study, type and
size of information that want to be embedded are not identified. However, this
research shows that by using LSB water marking techniques, it successfully embeds
the watermark.
A.M et al. (2009) did about an assuring authenticity of digital mammogram
where 32 of the same size of 512x512 pixels of images mammogram have been used
for experiment where watermarks are embedded at all the 128x128 blocks. The
watermarking technique that used for this study is DCT and DWT and the processes
related in this development are:
i. A set of 32 images are selected for testing.
ii. The extracted process of a sub-image of 512x512 pixels will be
executed
iii. Applying the watermarks to all the 128x128 blocks within the
images.
iv. Watermarked version at values equivalent are generated to
embedding the strengths of 1, 2 and 3.
24
The measurement of the quality has been done by using Peak Signal to Noise
Ratio (PSNR) metric, Steerable Visual Difference Predictor (SVDP) Sum metric
where the results are shown below:
Table 2.1: Experiment Result A.M et al. (2009).
WM configuration
PSNR metric
SVDP Sum metric
DCT ES=1
64.92
7.10
DCT ES=2
52.88
13.21
DCT ES=3
39.81
20.26
DWT ES=1
69.69
5.07
DWT ES=2
57.65
13.20
DWT ES=3
44.59
17.21
25
Table 2.2: Literature Review Table
No
Year
Author
Medical Image
Watermarking
Watermarking
Technique
1
2005
Boucherkha.S and
Ultrasound1=
First name, family name,
Benmohamed.M
39.72,
age and sex.
A Lossless
Ultrasound2=
Watermarking Based
52.27
Authentication
Ultrasound3=
System for Medical
41.33
LSB
Images.
2
2006
A. Giakoumaki et al.
CT
a. Physician’s digital
Discrete
MRI
signature or
wavelet
MRA
identification code
transform
PET
b. Contain keyword such
(DWT)
as ICD-10 diagnosis
code, image acquisition
characteristics and etc.
c. Contain patient’s
personal and
examination data such
as diagnosis report.
d. Reference watermark
3
2006
C.S.Woo, J.Du, and
X-ray, MR and
B.Pham
CT
Patient’s data
DWT and
LSB
4
2007
Mohamed Kallelet al.
Medical images
Multiple signature
LSB
5
2007
Y.Li, C.T.Li, and
Mammogram
Patient’s information
Not identified
Medical
DICOM Image identifiers
Not identified
Document
codes (UID) and Patient
C.H.Wei
6
2008
G.Coatrieux et al.
identifiers (Id)
7
2009
D.C. Lou et al.
CT scan and
Information
LSB
Information
DCT and
MRI
8
2009
A.M et al.
Mammogram
DWT
26
2.6
Peak Signal to Noise Ratio (PSNR)
It is a very important to measure the quality of the image. Mean Squared
Error (MSE) and Peak Signal-To-Noise Ratio (PSNR) are usually has been used to
measure the image (Eric A. Silva, 2008). There are many types of measurement of
image quality as below. However, Peak Signal to Noise Ration or PSNR is one of the
most popular measurements that have been used in medical images watermarking to
compute the quality of the image (Eric A. Silva, 2008).
i. Peak Mean Square Error (PMSE)
ii. Maximum Difference (MD)
iii. Average Difference(AD)
iv. Image Fidelity (IF)
v. Structural Content (SC)
vi. Laplacian Mean Square Error
(LMSE)
vii. N. Cross-Correlation(NK)
viii. N. Absolute Error (NAE)
ix. Correlation Quality (CQ)
x. N. Mean Square Error (NMSE)
Based on the study that was done by the researches, PSNR has been
commonly used to measure the medical image watermarking, Mingyan Lia et
al.,(2005). C.S.Woo, J.Du and B.Pham. (2006), evaluating the X-ray image of the
chest, MR image of the skull and CT image of the brain by using PSNR. B.Planitz
and A. Maeder (2006) also used PSNR to measure the quality of MRI, CTI and
CXR. It is followed by A. Giakoumaki et al.(2006), Mohamed Kallelet al.(2007), K.
A. Navaset Al. (2008), Rongrong Ni et al.(2008) and D.C. Lou et al. P (2009) have
used PSNR to measure the medical images. Thus, PSNR will be used in this project
to measure the quality of mammogram. Below is the formula of MSE and PSNR.
MSE = 1
N
N
∑
(xl-xl’)2
l=1
Figure 2.3: MSE Formula
27
PSNR =10 log 10 L2 db
MSE
Figure 2.4: PSNR Formula
Below are the descriptions about the formula above:
i. x1 is the pixel of the cover image which is based on coordinate
ii. x1’ is the pixel of the stego-image which is based on coordinate
iii. N is the size of the cover image and stego-image.
2.7
Issues in Typical mammogram
Currently, the mammogram that has been taken by the patient will be done a
process of diagnosis where the result and the medical image usually will be
transmitted in the difference hospital departments or room by room in the hospital.
After that, the medical image will be located in the reading room or in the file room
and if the doctor or the patient wants to get the mammogram or medical images, the
staff will get it for the doctor(J. Marc Overhage et al., 2005). Thus, possibilities to
lose or mishandle the medical image and problem to search the mammogram in the
hospital can happen. In mammography procedure, the history of mammogram is
really important and patients are responsible to the previous mammogram to the
doctor or radiologist before doing another mammogram.
Some hospitals implement the medical images application efficiently but
most of the applications are costly. However, this type of application actually gives
few advantages in distributing and managing the medical images. Nevertheless, since
the technology in the digital firm can be manipulated, tampered and altered thus, this
also becomes the issues of authentication and copyright of the owner of medical
images. Figure 2.5 shows common radiography processes of medical images.
28
Modality
Kodak DR
Agfa
Archive
CT
DICOM
Image
Processor
Image
Processor
CR
MRF
TRF
Medical
Image
JPEG
Image
RMRS
JPEG
Reports
Figure 2.5: Radiograph Process for Medical Image.(J. Marc Overhage et al., 2005)
2.8
Summary
As a summary, based on the reviews that have been done LSB technique is
already accepted and easy to be implemented. According to Y. I. Khamlichi et al.
(2003) this technique is fragile and a well-known method and efficient for inserting
data with a high embedding capacity. LSB technique is also well accepted as
watermarking technique and it has been proven based on the experiments results that
have done by the researches. Moreover, based on A.M et al. (2009) and Y.Li, C.T.Li,
and C.H.Wei, (2007) experiment, they used mammogram watermarking but the
techniques for watermarking that has been used are DCT and DWT techniques.
Additionally, based on the Diagnostic and Interventional Radiology (2008) article,
the comparison with old mammograms is very important for breast cancer diagnosis.
Mainly watermarking basic process consists of embedding and extracting
processes. The measurement of quality image will be done after the extracting
process by using Mean Squared Error (MSE) and Peak Signal-To-Noise Ratio
(PSNR) (Eric A. Silva, 2008). There are many types of measurement of image
29
quality. However, Peak Signal to Noise Ration or PSNR is one of the most popular
measurements that have been used in medical images watermarking to compute the
quality of the image (Eric A. Silva, 2008).
Based on the Table 2.2 above, it shows that LSB watermarking of medical
images are implemented on different types of images. For this project, an
authentication of mammogram image by using watermarking technique has been
proposed to identify the best location on the mammogram to embed the patient’s
information without affecting the quality of the image and also to authenticate the
mammogram image by using LSB technique. The authentication of watermarked
mammogram can help to authenticate the patients and capable to prove that the
mammogram belongs to the correct person and is issued from the trusted source.
CHAPTER 3
RESEARCH METHODOLOGY
3.1
Introduction
In this chapter, the project overflow is explained which consists of project
methodology, identifying breast area, embedding and extracting processes. For the
project development, MATLAB software is used.
3.2
Project Methodology
Figure 3.1 illustrates the project methodology for this project. It consists of
four stages such as Literature Review, Embedding process, Extracting process and
Testing.
Analysis Literature Review is the first stage to identify problem statements,
objectives, scope and analysis on the watermarked technique and type of medical
image will be done. Moreover, based on the literature reviews and studied that has
done, the process to develop this algorithm has been discovered. In addition,
techniques and type of the medical image will be identified.
31
The second stage is the embedding process. Before embedding process
executes, breast area will be identified at the left or at the right side of mammogram.
This can be done by calculating and comparing the dark and bright pixels intensity
for each image. After that, the watermark will be inserted into the mammogram
image. The embedding will be in the background area in a sequential order (pixel by
pixel). As mention in the chapter 1, patient’s information will be embedded at the
best location where it will be an empty place or unused area on mammogram by
using key.
The third stage will be the extracting process where the algorithm will read
the watermark (patient’s information) in mammogram. Key that has been used to
watermark the mammogram will be used back to read the patient’s information.
Finally at testing phase, comparison will be done between the extract
watermarks with the original watermark. Mammogram watermarking will be
compared with the original mammogram and the quality of image is measured by
using PNSR.
32
START
1.
2.
Literature Review
Identify objectives, scope, problem statements
Identify watermarking technique, watermark and type of
medical image
1.
2.
Embedding Process
Identify the area to embed watermark
Insert patient’s information into mammogram by using
key
1.
1.
2.
Extraction Process
Extract (read) the patient’s information
Testing
Compare the extract watermarks with original
watermark.
Compare the watermarked mammogram with original
mammogram by using PSNR.
NO
Mammogram
Authenticate?
YES
END
Figure 3.1: Project Methodology: Best- Location Technique For Mammogram
Authentication By LSB Based Watermarking
33
3.3
Embedding Process
Figure 3.2 below, shows the flow of embedding process. Before this process
is preceded, breast area and bit planes will be identified. After that, the extracting of
bit plane will be done. Next, the patient’s information will be inserted into the
mammogram bit. Then the watermarking will be embedded into mammogram by
using the key. Finally, the watermarked mammogram is generated.
EMBEDDING PROCESS
Mammogram
Watermarking
(w)
• Patient‘s
Information
Identify Bit
Planes
Extracting 8th bit
plane
Bit
Embedding Process
• Replacing the bit
of watermark
objects within 8th
bit plane of the
host image
Key
Watermarked
Mammogram
Figure 3.2: Embedding Process: Best- Location Technique for Mammogram
Authentication by LSB Based Watermarking
34
3.4
Extracting Process
The extracting process as shown as Figure 3.3 below, is the reversible
procedure of embedding process where first of all, to extract the watermark (patient’s
information), bit planes have to be identified and then by using key, the extracting
process will be completed. After that, the testing process will take place to
authenticate the owner of mammogram. The quality of the mammogram will be
measured using PSNR.
EXTRACTING PROCESS
Watermarked
Mammogram
Key
Identify Bit
Planes
Extracting
Process
Watermarking
(w’)
• The extracted
Patient‘s
Information
Figure 3.3: Extracting Process: Best- Location Technique for Mammogram
Authentication By LSB Based Watermarking
35
3.5
Summary
To summarize, the project methodology that has been developed in this
chapter has been used as guidance during the project development. It also helps in
minimizing the difficulties in the project development. Next chapter is the Project
Design and Testing that will describe the project development process in detail.
CHAPTER 4
PROJECT DESIGN AND TESTING
4.1
Introduction
This chapter will discusses the project design and tests on the 25
mammogram images. An algorithm and a prototype have been developed to execute
the embedding and extracting processes. This algorithm consists of embedding and
extracting the information from mammogram images.
4.2
Project Design
Based on the project scope, the algorithms were developed to test the quality
and authenticate the watermarked mammogram image at extraction process. This
project development was implemented two processes of embedding and extracting
the information from mammogram images. According to the literature review,
embedding and extracting processes were identified as the important part for
watermarking techniques. In this project development, the breast area will be
identified first before the watermarked process is done, the purpose of this approach
is to avoid embedding at the breast area.
As mention before in the literature review chapter, viewing the mammogram
is in symmetrical organs is the better approach to avoid mistakes in diagnosis the
mammogram. The physician will view and read the mammogram in side by side
37
orientation before doing further analysis on the mammogram image. Thus, the
algorithm will check whether the breast area at the left side or at the right side of
mammogram image before watermarking process is done. This can be done by
calculating and comparing the dark and bright pixel. Moreover, since different
medical image will have different background color level, the background color of
each mammogram samples will be checked by using data cursor tool in MATLAB
software to get the suitable range for dark value. The development of algorithm and
the prototype has been done by using MATLAB software.
Figure 4.1 shows the basic watermarking design that implemented in this
project development where the inputs will be mammogram image and patient’s
information as watermarking. After that, both of the inputs will go through to the
embedding process. Then, it will be followed by the extracting process and the
output is a watermarked mammogram image.
Mammogram
Watermarking
Embedding
Process
Extracting
Process
Watermarked
Mammogram
Image
Figure 4.1: Basic Watermarking Design
Figure 4.2 shows the overview design that has been used to develop an
authentication of mammogram image by using watermarked technique. The input
will be the mammogram image and patient’s information. Identifying breast area will
be done before the embedding process. After that, extracting process will be
executed and finally, for the output will be a watermarked mammogram images.
38
Mammogram
Identify Breast
Area
Embedding
Process
Watermarked
Mammogram
Image
Extracting
Process
Patient’s
Information
Figure 4.2: Best- Location Technique for Mammogram Authentication by LSB
Based Watermarking Design
4.3
Activity Flow Diagram
Based on the Figure 4.3, shows that the Activity Flow Diagram for the
prototype. The description will be explained in the next title.
39
Start
Click Browse
Mammogram Button
Click Insert Patient‘s
Information Button
End
No
Click Embedding
Button
Success Embeds the
Watermark into
Mammogram?
Identify Breast
Location, Bit Planes
and Extract 8th bit
plane of Mammogram
Replace the Bit of
Watermark Objects
within 8th Bit Plane of
the Host Image
Yes
Watermarked
Mammogram
Embedded
End
No
Success Extract the
Watermark into
Mammogram?
Click Extracting
Button
Identify Bit Planes to
Extract Patient
Information from
Mammogram
Yes
Compare the
watermarked
mammogram with the
original mammogram
by using PSNR.
End
Figure 4.3: Activity Flow Diagram for the Prototype Design
40
4.3.1 Prototype Design
Below is Figure 4.4, that shows the prototype design that has been developed
by using MATLAB software. This prototype has 4 buttons to execute the below
module. The mammogram image will be displayed on the axes at the right screen.
Figure 4.4: Prototype Design
4.3.2 Browse Mammogram Module
This module is to browse the mammogram images in the bitmap format.
Browsing can be done by clicking on the ‘Browse Mammogram’ button. User can
select the mammogram image that wants to be watermarked in the directory. The
selected watermarked image will be displayed in the axes at right screen.
41
4.3.3 Insert Data Module
Type of watermark that will be used is in text format. The watermark will be
the patient’s information such as Patient’s Name, Patient’s ID, Age, MyKad (Identity
Card of Malaysia), Date, Physician’s Name and Diagnosis. The data can be inserted
by clicking on the button of ‘Insert Data’ as shown in the Figure 4.5 below. All
information must be inserted and will be saved in .txt format.
Figure 4.5: Patient Information
4.3.4 Watermarked Module
In this module, the embedding process will be continued after the
‘Watermarked’ button is clicked. At this phase, the patient’s information will be
embedded in the mammogram image. During the embedding process, the best
location of embedding watermark will be involved to differentiate the background of
mammogram with the breast area. Thus, the embedding of watermarking will be
done in the background area and type of watermark that will be implemented is
invisible watermark and from the literature review that has done, invisible watermark
is suitable for copyright purpose.
42
After the watermarked process completed, the watermarked image will be
displayed at the right screen and the message dialog will be displayed to show that
the watermarked process completed. Finally the watermarked image will be saved in
the directory that has been chosen by the user. Refer Figure 4.6 below.
Additionally, from the studies that have been done, watermarking is also
preferred to be chosen as a favourite technique in hiding information by the
researches. This is because it will help to enhance the security and privacy in the
hospitals and also to manage the patients’ data since issues regarding the image
efficiency still occurred. Furthermore, LSB is already well accepted as watermarking
technique and from the result of the experiments that has been successfully done by
the researches, this technique can be used to accomplish the watermarked process
and also be able to embed large information.
Figure 4.6: Watermarked
43
4.3.5 Extracting Module
The extracting process is a process that will read the watermarking that has
been embedded into watermarked mammogram. The patient’s information that has
been embedded in the mammogram image will be read back and displayed at the text
field Figure 4.7 below. Since PNSR is commonly used to measure the quality of
image. Thus, it will be used to measure the quality of the extracting mammogram.
After watermarked process accomplished, the PSNR value will be displayed to show
the image quality value after watermark.
Figure 4.7: Extract Data
4.4
Algorithm Flow Chart
Figure 4.8 is the flow chart of the algorithm for this prototype. Before
embedding the watermarking, the selected mammogram will be split into two parts,
A and B like Figure 4.9. The size of A and B will be identified based on the
mammogram image size. The reason to split the mammogram into two parts is for
the pixel comparison purpose in the next process.
44
Next step is to identify the location of breast area. This can be done by
calculating the black (dark) and white (bright) pixel at both parts. The dark pixel has
been set from 0 to 26. Based on the literature review, the background of
mammogram will be in the black and gray colors. Thus, Data Cursor Tool on the
MATLAB Software has been used to check the range of dark pixel of mammogram
samples since it is hard to differentiate because the intensity of color looks similar.
Part A
Part B
Figure 4.9: Categorize the Mammogram into Part A and Part B
Then, for A part of mammogram, bright and dark pixel are calculated. The
purpose to calculate the pixel color is to ensure whether the bright pixel is more than
the dark pixel or not. The same activity is done to the B part.
45
After that the amount of dark pixel at A is compared with the amount of dark
pixel at B. If the quantity of the dark pixel at A more than B, the breast area will be
at right side and the patient’s information will be embedded at the left side. However,
if the quantity of the black pixel at A is less than Right, breast area will be at the left
side and the patient’s information will be embedded at the right side.
As mention before, Least Significant Bit (LSB) watermarking is used in this
algorithm. After the selected image and the input patient’s information has been
changed into the binary format, there will be 8 bit plane for each pixel like shown as
Figure 4.10. The patient’s information will be hidden in the LSB image. The
insertion of LSB will be together with key to define the starting embedding location
at the background area. About 25 samples of images have been tested to embed the
same size patient’s information with the different LSB value and the result will be
discussed later.
LSB 1
LSB 2 LSB 3
LSB 4
LSB 5
LSB 6
LSB 7 LSB 8
Figure 4.10: LSB Bit Plane
Patient’s information will be embedded to the image based on the LSB
chosen. Below is the example for LSB 1. Mammogram image will be converted into
binary as shown in the Figure 4.11. ‘G’ will be represented as patient’s information,
which will convert into 01000111 as shown in the Figure 4.12. The character 'G' will
be across the following eight bytes of a mammogram file. These eight bits will be
inserted into the least significant bit of each of the eight image bytes as shown in
Figure 4.13. The same process will be applied to LSB 2, LSB 3, LSB 4, LSB 5, LSB
6, LSB 7 and LSB 8 respectively.
10010101
00001101
11001001
10010110
00001111
11001011
10011111
00010000
Figure 4.11: Bit Plane that Represent Images
46
0
1
0
0
0
1
1
1
Figure 4.12: ‘G’ in Binary
10010100
00001101
11001000
10010110
00001110
11001011
10011111
00010001
Figure 4.13: After Watermarked
After embedding process has done, the next activity will be the extracting
process where patient’s information will be read on the watermarked mammogram.
The watermarked mammogram has been inserted with the patient’s information at
the LSB of mammogram image. Thus, to read back the information from the
watermarked mammogram, the process when embedding the data will be turned
around where the same key will be used to find the starting location embedding at the
background area and extract the patient’s information in the watermarked
mammogram.
Finally, checking the quality of the watermarked mammogram, PSNR
formula has been used to measure the quality of mammogram. Please refer
Appendix A to see the PSNR result of 25 samples mammogram.
47
Separate the mammogram image into two parts:
A and B
Identify the pixel color whether it is dark or
bright pixel.
No
Consider as Bright pixel
Intensity less than 26?
Yes
Calculate bright and dark pixel
at B part. Compare pixel color
at both part
Consider as Dark pixel
Calculate bright and dark pixel at A part. Compare
pixel color at both parts
Black pixel at part A is
more than Black pixel at
B?
Yes
Breast Area at Right Side
No
Breast Area at Left Side
Embed patient’s information at
Right Side
Embed patient’s information at Left Side
Extract watermark
Test the quality of watermarked image by using
PSNR
Figure 4.8: Algorithm Flow Chart
48
4.5
Best- Location Technique Algorithm for Mammogram Authentication by
LSB Based Watermarking
Below is the algorithm that has been developed in this project:
1. Algorithm to Identify Breast Area.
i.
Calculate dark and bright pixel at A part:
a. Identify size of row and column for A part.
b. Calculate the pixel intensity for mammogram image that LESS than
26.
c. Calculate the pixel intensity for mammogram image that MORE than
26.
ii.
Calculate dark and bright pixel at B part:
a. Identify size of row and column for B part.
b. Calculate the pixel intensity for mammogram image that LESS than
26.
c. Calculate the pixel intensity for mammogram image that MORE than
26.
iii.
Comparing black pixel at A and B part:
a. If black pixel at A part MORE than B part, breast area at the right
side.
b. If black pixel at A part LESS than B, breast area at the left side.
2. Algorithm to Embedded Information.
i.
For breast area at the left side:
a. Information will be embedding at the right side when the pixel
intensity of pixel is less 30.
b. If information is less than value image bits, the embedding will stop.
c. All the information will be embedded into mammogram image by
replacing the binary of watermarked mammogram image based on
LSB and key.
49
ii.
For breast area at the left side:
a. Information will be embedded at the left side when pixel intensity of
pixel is less 30.
b. If information in less than value image bits, the embedding will stop.
c. All the information will be embedded into mammogram image by
replacing the binary of watermarked mammogram image based on
LSB and key.
3. Algorithm to Extract Information.
a. Information will extract the data by using key and compare the text
size and logo size of information between original image and
watermarked image.
4. Check quality of watermarked image by using PSNR.
a. Get mean squared error (MSE) value by using the MSE formula:
b. Get PSNR value by using PSNR formula
4.6
Summary
As a summary, best- location technique for mammogram authentication by
LSB based watermarking is developed using MATLAB software. According to the
studies about the current technology and issues in medical image authentication, it
shows that the further research is still needed. Thus, by identifying the best location
on the mammogram image it will help to embed the patient’s information without
affecting the quality of the image. Furthermore, by developing an authentication
technique on mammogram image, the patient’s information that embeds into the
mammogram background can extract patient information without distorting the
medical image.
CHAPTER 5
RESULTS AND DISCUSSION
5.1
Introduction
This chapter discusses about the results of an authentication of mammogram
image by using watermarked technique. There are 25 samples of mammogram
images are being used to test the algorithm for quality and authentication testing
purpose at the extraction process and the samples will be in grayscale format.
5.2
Results in Identifying Breast Area
This process identifies the breast area, by dividing the image into two parts
such as part A and B. After dividing the image into these two parts, the pixel
intensity of mammogram images is calculated to identify the breast area. The pixel
intensity that is less than 26 will be considered as dark pixel and 27 until 255 will be
consider as bright pixel.
After calculating dark and bright pixel between these two parts, the
comparison dark pixel between these parts is done to check which side has darker
pixel. If A part has darker pixel than B part, it means that the breast area is at the
right side and the embedding patient’s information is done at the background area
which is at the left side. This algorithm has been tested to the mammogram samples
and it can identify the breast area acceptably.
51
5.3
Results in Embedding Process
After breast area has been identified, embedding process is done, embedding
process uses Least Significant Bit (LSB) where the bit of patient’s information is
replaced in the LSB of mammogram image. Based on the algorithm, only LSB 1 and
2 shows a little distorted at the background images while only some images are
distorted when using LSB 3. For LSB 4 until LSB 8, there are no distorted images
identified at all. Refer Appendix B to see the watermarked images that used LSB 1
to LAB 8.
5.4
Results in Extracting Process
Extracting process is a process that can extract the patient’s information on
watermarked image. Based on the result, all information that has been embedded in
the samples of mammogram images can be read back without any interruption or
data lost. The text file size before embedding and after extraction is also in the same
size. This shows that the algorithm has successfully extracted the data and enables to
authenticate the mammogram image.
5.5
Results in PSNR Testing
PSNR is being used to test the quality of image. The samples of watermarked
mammogram images show the results are more than 45 dB. Please refer Appendix
A, for the summary of PSNR result. The PSNR has been tested to the watermarked
image of LSB 1 until LSB 8. Appendix B shows some of the watermarked
mammogram images with PSNR result. Table 5.1 and Table 5.2 show the result of a
watermarked mammogram image of LSB 1 until LSB 8.
52
Table 5.1: Result of A Watermarked Mammogram Image of LSB 1 until LSB 8
(Breast Area at Left side).
Data
Mammogram
1
Name: CIK FERESA BINTI MOHD FOOZY ID: ID001
MyKad: 840102067541
Age: 25
Date: 18/09/2009 , 11:01:48 AM
Physician Name: NORAITY BINTI NORDIN
Diagnosis: Highly suggestive of malignancy (cancer) means that
there are findings that look like and probably are cancer. Patient
requires biopsy.
Size character: 2044
LSB 1
49.9075
Dimension:
305 x 620
Mammogram
1
Dimension:
305 x 620
LSB 2
55.9281
53
Mammogram
1
LSB 3
61.9487
Dimension:
305 x 620
Mammogram
1
Dimension:
305 x 620
LSB 4
67.9693
54
Mammogram
1
LSB 5
79.8256
Dimension:
305 x 620
Mammogram
1
Dimension:
305 x 620
LSB 6
80.6174
55
Mammogram
1
LSB 7
89.6483
Dimension:
305 x 620
Mammogram
1
LSB 8
94.5295
Dimension:
305 x 620
Label:
Shows the distorted location on the mammogram.
56
Table 5.2: Result of A Watermarked Mammogram Image of LSB 1 until LSB 8
(Breast Area at Right Side).
Data
Mammogram
16
Name: CIK FERESA BINTI MOHD FOOZY ID: ID001
MyKad: 840102067541
Age: 25
Date: 18/09/2009 , 11:01:48 AM
Physician Name: NORAITY BINTI NORDIN
Diagnosis: Highly suggestive of malignancy (cancer) means that
there are findings that look like and probably are cancer. Patient
requires biopsy.
Size character: 2044
LSB 1
49.7793
Dimension:
306x600
Mammogram
LSB 2
16
55.7999
Dimension:
306x600
57
Mammogram
LSB 3
16
61.8205
Dimension:
306x600
Mammogram
LSB 4
16
67.8411
Dimension:
306x600
58
Mammogram
LSB 5
16
73.8617
Dimension:
306x600
Mammogram
LSB 6
16
82.0383
Dimension:
306x600
59
Mammogram
LSB 7
16
87.7592
Dimension:
306x600
Mammogram
LSB 8
16
94.0795
Dimension:
306x600
Label:
Shows the distorted location on the mammogram.
60
5.5.1 Results in Attacking the Watermarked of Mammogram Image
The advantage of using JPEG format, it will help in minimizing the disk
storage and the file image storage. This format will reduce the image size and quality
because of the compression. Thus, this intentional attack has been done to the
algorithm to check the quality of watermarked images after the attack. From the
experiments that have done, the quality of the image is still in a good condition and
the PSNR result shows it is more than 30dB (Refer Appendix A for the PSNR
result).
5.6
Summary
For the summary, the requirement of the project has been achieved where an
authentication technique of watermarked mammogram using the LSB technique has
been developed and the best location on the mammogram has been identified to
embed the patient’s information without affecting the quality of the image.
Moreover, PSNR result shows that, the quality of image is still in a good condition
after embedding the patient’s information and this process has been done at the
background area so the quality of breast area will not interrupt.
CHAPTER 6
CONCLUSION AND FUTURE WORK
6.1
Introduction
This chapter discusses about the contribution and the future work of this
project. There are 25 samples of mammogram images being used in Bitmap format
to test the algorithm for quality and authentication testing purpose at the extraction
process and the samples are in grayscale format.
6.2
Contribution of the Project
A prototype for this project has been developed where this algorithm has the
ability to embed more information on mammogram images by using LSB technique.
This algorithm is also able to identify the best location by calculating and comparing
the dark and bright pixel intensity in the mammogram images. This is to differentiate
the breast and background area of the mammogram images. Normally, embedded
images are considered as a good quality if the PSNR is more than 30dB. Seems, the
experiment shows the result of PSNR of this technique is more that 45dB, thus the
images are still in a good quality (refer Appendix B).
62
For the extraction process, all patient’s information can be read and this
shows that the mammogram images have been authenticated with the patient’s
information.
6.3
Future Work
For the future work of algorithm, artificial intelligent techniques are
suggested to enhance the algorithm. By implementing this technique, pixel intensity
in mammogram images can be checked precisely to differentiate the critical area and
the background area. Moreover, different attempts can be tested to study the
outcomes of the experiment.
6.4
Summary
LSB watermarking is a technique that available to embed large information,
thus by using this technique in the mammogram image, it can embed large patient’s
information. As mention in the previous chapter, this algorithm consists of two
processes such as embedding the patient’s information in text format where the best
location will be identified based on the pixel color intensity and another process is
extracting process. Based on the pixel color intensity, the area of breast location can
be identified at the left or at the right side of mammogram image. By identifying the
breast location, the patient’s information will not embed at the breast area and the
breast area will not distort.
63
Based on the result (refer Appendix B), it shows that by using LSB 1, LSB 2
and LSB 3 the mammogram images will distort the embedded area. However by
using LSB 4, LSB 5, LSB 6, LSB 7 and LSB 8, there is no distortion traced. For
extracting process, the algorithm will read the patient’s information that has
embedded on the watermarked mammogram images. Based on the result, the
algorithm can extract (read) the patient’s information.
The algorithm for an authentication of mammogram image was developed to
study the implementation of LSB watermarking technique on the mammogram
images and to test the quality of image and extract the patient’s information. A
prototype also has been developed to run the algorithm. Moreover, this algorithm
also has been tested with JPEG compression attack and the PSNR result of JPEG
compression is more than 30 dB (refer Appendix A).
64
References
A. Giakoumaki et al. P. (2006). Secure and Efficient Health Data Management
through Multiple Watermarking On Medical Images. Springer
Med.Bio.Eng.Comput. 44: 619-631.
Anastasios Tefas et al. (2005). Watermarking Techniques for Image Authentication
and Copyright Protection. Elsevier Academic Press.
A.W.T. Goh et. al. P. (2008). On the Security of Zhou et al.’s Authentication and
Integrity Scheme for Digital Mammography Images. Springerlink: Biomed
2008, Proceedings 21. 871–874.
BBC. (9th January 2009). Prisoners' medical details lost. BBC News.
http://news.bbc.co.uk/2/hi/uk_news/england/lancashire/7820338.stm
Bits per pixel - Image Processing with LEADTOOLS. Retrieved 2nd August 2009,
from
http://www.imaging-components.com/Imaging-Components/ImageProcessing/bits-per-pixel.shtm
B.Planitz and A. Maeder. P. (2006). Medical Image Watermarking: A Study on
Image Degradation. Brisbane: e-Health Research Centre, ICT CSIRO
Breast cancer cases Molson Informatics Project, McGill Faculty of Medicine website
Interactive Mammography Analysis Web Tutorial. Retrieved 2nd August 2009, from
http://sprojects.mmi.mcgill.ca/Mammography/index.htm
65
Chen C.C and Ka D.S. P. (2008).Dct-Based Zero Replacement Reversible Image
Watermarking Approach.
Chiou-Ting Hsu and Ja-Ling Wu. P.(1996). Hidden Signatures in Images. Taiwan:
National Taiwan University, Taipei.
C.S.Woo et al. P. (2006). Multiple Watermark Method for Privacy Control and
Tamper Detection in Medical Images. Australia: Information Security
Institute, Faculty of Information Technology, Queensland University of
Technology.
D.C. Lou et al. P (2009). Multiple Layer Data Hiding Scheme For Medical Images.
ScienceDirect Computer Standards & Interface. Page: 329-335.
Center of Medicare & Medicaid Services (CMS). P. 2005.Selecting a Development
Approach.
C. M. Kung, et al. P (2003). A Robust Watermarking and Image Authentication
Technique. IEEE. 400-404.
Daniel B. Kopans. P(2007) . Breast imaging. Lippincott Williams & Wilkins,
Philadephia USA.
Dennis A. Wixom (2003). Systems Analysis Design. Second Edition John Wiley &
Sons, Inc.
Diagnostic and Interventional Radiology. Basic Imaging : Mammography. Creighton
University Medical Center. Retrieved January 31, 2009, from
http://radiology.creighton.edu/mammo.htm
D.J Manning et al. P. (2005). Perception Research in Medical Imaging. British
Journal of Radiology. Page:683-685.
D.Neeta and K.Snehal. P. (2005). Implementation of LSB Steganography and Its
Evaluation for Various Bits. India: Computer Science Dept K.K.Wagh
Institute of Engineering Education & Research, Nashik. Page:173-178.
66
Elham Shahinfard and Shohreh Kasaei. P. (2001).Digital Image Watermarking using
Wavelet Transform.
Eric A. Silva. P. (2008).Quantifying Image Similarity Using Measure Of
Enhancement By Entropy. University of Texas at San Antonio.
F. Caoa et al. P. (2002). Medical image security in a HIPAA mandated PACS
Environment. Sciencdirect : Computerized Medical Imaging and Graphics 27
(2003). Page:185–196.
G.Coatrieux et al. P. (2002). Relevance of Watermarking in Medical Imaging. Paris:
ENTS,Department TSI.
G.Coatrieux et al. P. (2008).Watermarking Medical Images with Anonymous Patient
Identification to Verify Authenticity. Inserm U650, LaTIM; GET ENST
Bretagne, Dpt. ITI.
G.Muller. P. (2008).Medical Imaging Workstation: CAF Views. Embedded Systems
Institute. Page:8-16.
Husrev T. Sencar and Nasir Memon. P. (2006) Watermarking and Ownership
Problem: A Revisit. New York.
How Mammography is Performed: Imaging and Positioning. Retrieved 2nd August
2009, from
http://www.imaginis.com/breasthealth/mammography_imaging.asp
Jinshen Wang et. al. P. (2004).A Feature-Watermarking Scheme for JPEG Image
Authentication.Springer: Verlag Berlin Heidelberg. 212–222.
J. Marc Overhage et al. P. (2005).Integration of Radiographic Images with an
Electronic Medical Record.Indianapolis.
J. Nayak et al. P. (2008). Secure and Efficient Health Data Management through
Multiple Watermarking On Medical Images. Springer Science Business
Media.
67
Ju-YuanHsiao et al. P.(2009).Block-Based Reversible Data Embedding .
Sciencedirect: Signal Processing.Page: 556–569.
K. A. Navaset Al. P. (2008).EPR Hiding In Medical Images for Telemedicine.
Proceedings Of World Academy Of Science, Engineering And Technology
.Page: 307-6884.
Marie Babel a et al. P. (2008). Joint source–channel coding: Secured and
progressive transmission of compressed medical images on the Internet.
Sciencedirect: Computerized Medical Imaging and Graphics. Page: 258–269.
Medindia.com. Digital Mammogram For Detection Of Breast Cancer In Women
With Dense Breasts. The New England Journal of Medicine. Retrieved 16th
September 2005.
http://www.medindia.net/news/view_news_main.asp?str=2&x=4963
Mingyan Lia et. al. P. (2005). Protecting Patient Privacy Against Unauthorized
Release Of Medical Images In A Group Communication Environment.
Sciencedirect: Computerized Medical Imaging and Graphics 29 (2005). Page:
367–383.
Mohamed Kallelet al. P. (2007). A multiple Watermarking Scheme for Medical
Image in the Spatial Domain. GVIP Journal. Page: 37-42
M. R. M. Rizk et al. P. (2006). Adaptive Watermarking Techniques Based On MultiScale Morphological Image Segmentation. Egypt
N. Liu, P. Amin et al. P. (2006).Chapter 7: An Overview of Digital
Watermarking.Journal Of Innovative Computing, Information And Control.
N. L.Ping. P. (2003). A Study of Digital Watermarking on Medical Image. IFMBE
Proceedings.Page: 2264-2267.
Pam Stephan. P.(2007). What Are the Two Angles of View for a Routine
Mammogram?.Medical Review Board.
68
Rongrong Ni et al. P. (2008). Pinpoint Authentication Watermarking Based On A
Chaotic System. ScienceDirect: Forensic Science International . Page: 54–62.
S.Boucherkha and M. Benmohamed. P.(2005). A Lossless Watermarking Based
Authentication System for Medical Images. Proceedings of World Academy
of Science, Engineering and Technology. Page: 100-103.
S. Huliiv et al. P. (2004). A Variable Depth LSB Data Hiding Technique in Images.
IEEE: Proceedings of the Third International Conference on Machine
Learning and Cybernetics, Shanghai. Page: 3990-3994.
S. Luigi. P. (2005). Mammogram. Italy: Department of Dentistry and Surgery.
Section of General Surgery, Medical and Dentistry School, State University
at Bari.
Veysel Aslantas. P. (2008). An Optimal Robust Digital Image Watermarking Based
On SVD Using Differential. Sciencedirect: Computerized Medical Imaging
and Graphics. Page: 258–26
Woman to Woman: Dixie Mills. Mammograms — what’s best for you?. Retrieved
2nd August 2009, from
http://www.womentowomen.com/breasthealth/mammograms.aspx
Y.Li et al. P. (2007). Protection of Mammograms Using Blind Steganography and
Watermarking. United Kingdom: Department of Computer Science,
University of Warwick.
Y. I. Khamlichi et al. P(2003).Authentication System For Medical Watermarked
Content Based Image.
Zhou X Q, Huang H K, Lou S L (2001) Authenticity and Integrity of Digital
Mammography Images. IEEE Trans Med Imaging.
69
APPENDIX A
SUMMARY OF PSNR RESULT
70
LSB
Mammogram
LSB
1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Side of
Breast Area
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Right
Right
Right
Right
Left
Right
Right
Right
Right
Right
Right
Right
LSB
2
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Right
Right
Right
Right
Left
Right
Right
PSNR(dB)
49.9075
50.5470
50.5121
49.6284
50.5539
51.3289
51.0522
49.5045
51.2420
50.4778
50.4941
50.4700
50.6665
50.8613
51.1359
49.7793
50.4602
49.9622
50.5470
50.0311
50.6665
51.2420
50.4778
50.6665
50.7580
PSNR(dB)
(JPEG Attack)
44.6367
43.8638
43.2719
42.0953
42.5077
41.7875
42.5839
43.1760
43.0104
44.1941
43.2633
44.0444
44.6917
44.7401
41.5258
44.0209
41.8908
34.1179
43.1418
42.9403
45.1666
42.7561
42.8833
44.7804
43.8241
55.9281
56.5676
56.5327
55.6490
56.5745
57.7635
57.0728
55.5251
57.2626
56.4984
56.5147
57.5551
56.6871
56.8819
57.1565
55.7999
55.4162
55.9828
56.5676
56.0517
45.0613
44.0966
43.4509
42.3064
42.6872
41.9468
42.7116
43.3724
43.1925
44.5080
43.4130
44.3175
45.0507
45.1007
41.7063
44.5040
42.2069
34.1568
43.4172
43.3074
71
21
22
23
24
25
Right
Right
Right
Right
Right
56.8802
57.2626
56.4984
56.6871
56.7786
45.6388
42.9593
43.1862
45.1921
44.2059
LSB
3
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Right
Right
Right
Right
Left
Right
Right
Right
Right
Right
Right
Right
61.9487
62.5882
62.5533
66.8359
62.5951
66.7944
68.9292
67.3815
63.2832
62.5190
64.1114
63.3409
62.7077
62.9025
64.7532
61.8205
61.4368
62.0034
68.4239
62.2653
62.7077
63.2832
62.5190
62.7077
62.7992
45.0746
44.1314
43.4813
42.3364
42.7111
41.9589
42.7515
43.4459
43.2099
44.5575
43.4580
44.3337
45.1150
45.1934
41.7605
44.5887
42.2879
34.1576
43.5032
43.3582
45.7656
43.0088
43.2430
45.3250
44.2770
LSB
4
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Right
Right
Right
67.9693
74.4445
68.5739
67.8832
74.4515
71.6755
69.1140
67.5663
69.3038
70.1157
71.0337
68.7248
69.1234
68.9231
71.6755
67.8411
45.0999
44.1347
43.4870
42.3348
42.7141
41.9603
42.7493
43.4435
43.2181
44.5633
43.4634
44.3457
45.1232
45.2157
41.7639
44.6142
72
17
18
19
20
21
22
23
24
25
Right
Left
Right
Right
Right
Right
Right
Right
Right
73.2932
68.0240
70.7648
68.0929
68.9214
75.1395
68.5396
68.7283
68.8198
42.3014
43.5018
43.5018
43.3775
45.7839
43.0238
43.2589
45.3418
44.2922
LSB
5
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Right
Right
Right
Right
Left
Right
Right
Right
Right
Right
Right
Right
79.8256
80.4651
77.0724
76.1886
80.4721
76.7944
80.9704
73.5869
81.1601
77.0381
77.0543
78.1696
79.3353
77.7691
78.8356
73.8617
73.4780
79.8804
76.4857
79.2798
75.5787
75.3244
80.3960
76.0617
79.4268
45.0999
44.1341
43.4870
42.3367
42.7139
41.9604
42.7515
43.4461
43.2187
44.5632
43.4634
44.3459
45.1232
45.2206
41.7647
44.6148
42.3014
34.1597
43.5037
43.3793
45.7862
43.0238
43.2608
45.3457
44.2939
LSB
6
1
2
3
4
5
6
7
8
9
10
11
12
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
80.6174
82.8060
82.1912
82.2092
84.2742
85.3137
81.5503
79.6075
81.9520
83.0587
83.0749
81.6375
45.0999
44.1341
43.4880
42.3367
42.7139
41.9603
42.7512
43.2187
43.2187
44.5632
43.4634
44.3468
73
13
14
15
16
17
18
19
20
21
22
23
24
25
Left
Right
Right
Right
Right
Left
Right
Right
Right
Right
Right
Right
Right
83.5950
82.2771
82.5517
82.0383
80.3284
80.0652
82.5063
80.5292
82.9255
84.1704
83.0587
83.5950
81.2561
45.1232
45.2208
41.7646
44.6151
42.3016
34.1597
43.5041
43.3797
45.7866
43.0238
43.2609
45.3460
44.2940
LSB
7
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Left
Right
Right
Right
Right
Left
Right
Right
Right
Right
Right
Right
Right
89.6483
88.8266
90.2530
87.0647
87.7420
89.4155
90.0013
86.9409
89.8434
88.4578
89.4431
87.9064
88.6465
88.2977
89.7373
87.7592
88.7226
91.9216
89.4960
87.7308
87.8547
88.1953
88.7574
90.4074
87.2767
45.0999
44.1340
43.4880
42.3368
42.7139
41.9603
42.7514
43.4461
43.2186
44.5632
43.4636
44.3468
45.1232
45.2208
41.7647
44.6152
42.3016
34.1597
43.5041
43.3794
45.7860
43.0239
43.2607
45.3460
44.2940
LSB
8
1
2
3
4
5
6
7
8
Left
Left
Left
Left
Left
Left
Left
Left
94.5295
94.5475
94.2324
93.9286
94.8541
94.3446
96.3998
93.8047
45.0999
44.1340
43.4880
42.3368
42.7139
41.9603
42.7514
43.4461
74
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Left
Left
Left
Left
Left
Right
Right
Right
Right
Left
Right
Right
Right
Right
Right
Right
Right
94.9623
93.9348
93.7028
95.0920
93.6405
94.8618
95.1365
94.0795
93.1159
97.9422
94.0040
93.7514
94.6671
95.8640
95.8254
98.6465
93.7319
43.2186
44.5632
43.4636
44.3468
45.1232
45.2208
41.7647
44.6152
42.3016
34.1597
43.5041
43.3794
45.7861
43.0239
43.2608
45.3460
44.2940
75
APPENDIX B
WATERMARKED IMAGES
76
Data
Mammogram 3
Dimension:
345x630
Mammogram 3
Dimension:
345x630
Name: CIK FERESA BINTI MOHD FOOZY ID: ID001
MyKad: 840102065174
Age: 25
Date: 18/09/2009 , 11:01:48 AM
Physician Name: NORAITY BINTI NORDIN
Diagnosis: Highly suggestive of malignancy (cancer) means
that there are findings that look like and probably are
cancer. Patient requires biopsy.
Size character: 2044
LSB 1
50.5121
LSB 2
56.5327
77
Mammogram 3
Dimension:
345x630
Mammogram 3
Dimension:
345x630
Mammogram 3
Dimension:
345x630
LSB 3
62.5533
LSB 4
68.5739
LSB 5
77.0724
78
Mammogram 3
Dimension:
345x630
Mammogram 3
Dimension:
345x630
Mammogram 3
Dimension:
345x630
Label:
LSB 6
82.1912
LSB 7
90.2530
LSB 8
94.2324
Shows the distorted location on the mammogram.
79
Data
Name: CIK FERESA BINTI MOHD FOOZY ID: ID001
MyKad: 840102067541
Age: 25
Date: 18/09/2009 , 11:01:48 AM
Physician Name: NORAITY BINTI NORDIN
Diagnosis: Highly suggestive of malignancy (cancer) means that
there are findings that look like and probably are cancer. Patient
requires biopsy.
Size character: 2044
LSB 1
Mammogram
4
49.6284
Dimension:
345x514
Mammogram
4
LSB 2
55.6490
Dimension:
345x514
80
Mammogram
4
LSB 3
66.8359
Dimension:
345x514
Mammogram
4
LSB 4
67.8832
Dimension:
345x514
Mammogram
4
LSB 5
76.1886
Dimension:
345x514
Watermark Mammogram Image
81
Mammogram
4
LSB 6
82.2092
Dimension:
345x514
Mammogram
4
LSB 7
87.0647
Dimension:
345x514
Mammogram
4
LSB 8
93.9286
Dimension:
345x514
Label:
Shows the distorted location on the mammogram.
82
Data
Name: CIK FERESA BINTI MOHD FOOZY ID: ID001
MyKad: 840102067541
Age: 25
Date: 18/09/2009 , 11:01:48 AM
Physician Name: NORAITY BINTI NORDIN
Diagnosis: Highly suggestive of malignancy (cancer) means that
there are findings that look like and probably are cancer. Patient
requires biopsy.
Size character: 2044
LSB 1
Mammogram
5
50.5539
Dimension:
350x627
Mammogram
5
LSB 2
56.5745
Dimension:
350x627
83
Mammogram
5
LSB 3
62.5951
Dimension:
350x627
Mammogram
5
LSB 4
74.4515
Dimension:
350x627
Mammogram
5
LSB 5
80.4721
Dimension:
350x627
84
Mammogram
5
LSB 6
84.2742
Dimension:
350x627
Mammogram
5
LSB 7
87.7420
Dimension:
350x627
Mammogram
5
LSB 8
94.8541
Dimension:
350x627
Label:
Shows the distorted location on the mammogram.
85
Data
Name: CIK FERESA BINTI MOHD FOOZY ID: ID001
MyKad: 840102067541
Age: 25
Date: 18/09/2009 , 11:01:48 AM
Physician Name: NORAITY BINTI NORDIN
Diagnosis: Highly suggestive of malignancy (cancer) means that
there are findings that look like and probably are cancer. Patient
requires biopsy.
Size character: 2044
LSB 1
Mammogram
7
51.0522
Dimension:
387x636
Mammogram
7
LSB 2
57.0728
Dimension:
387x636
86
Mammogram
7
LSB 3
68.9292
Dimension:
387x636
Mammogram
7
LSB 4
69.1140
Dimension:
387x636
Mammogram
7
LSB 5
80.9704
Dimension:
387x636
87
Mammogram
7
LSB 6
81.5503
Dimension:
387x636
Mammogram
7
LSB 7
90.0013
Dimension:
387x636
Mammogram
7
LSB 8
96.3998
Dimension:
387x636
Label:
Shows the distorted location on the mammogram.
88
Data
Name: CIK FERESA BINTI MOHD FOOZY ID: ID001
MyKad: 840102067541
Age: 25
Date: 18/09/2009 , 11:01:48 AM
Physician Name: NORAITY BINTI NORDIN
Diagnosis: Highly suggestive of malignancy (cancer) means that
there are findings that look like and probably are cancer. Patient
requires biopsy.
Size character: 2044
Mammogram 49.9622
18
Dimension:
383x500
Mammogram
18
Dimension:
383x500
55.9828
89
Mammogram
18
62.0034
Dimension:
383x500
Mammogram
18
68.0240
Dimension:
383x500
Mammogram
18
Dimension:
383x500
79.8804
90
Mammogram
18
80.0652
Dimension:
383x500
Mammogram
18
91.9216
Dimension:
383x500
Mammogram
18
97.9422
Dimension:
383x500
Label:
Shows the distorted location on the mammogram.
91
Data
Mammogram
19
Name: CIK FERESA BINTI MOHD FOOZY ID: ID001
MyKad: 840102067541
Age: 25
Date: 18/09/2009 , 11:01:48 AM
Physician Name: NORAITY BINTI NORDIN
Diagnosis: Highly suggestive of malignancy (cancer) means that
there are findings that look like and probably are cancer. Patient
requires biopsy.
Size character: 2044
LSB 1
50.5470
Dimension:
350x626
Mammogram
19
LSB 2
56.5676
Dimension:
350x626
92
Mammogram
19
LSB 3
68.4239
Dimension:
350x626
Mammogram
19
LSB 4
70.7648
Dimension:
350x626
Mammogram
19
LSB 5
76.4857
Dimension:
350x626
93
Mammogram
19
LSB 6
82.5063
Dimension:
350x626
Mammogram
19
LSB 7
89.4960
Dimension:
350x626
Mammogram
19
LSB 8
94.0040
Dimension:
350x626
Label:
Shows the distorted location on the mammogram.
94
Data
Mammogram
20
Name: CIK FERESA BINTI MOHD FOOZY ID: ID001
MyKad: 840102067541
Age: 25
Date: 18/09/2009 , 11:01:48 AM
Physician Name: NORAITY BINTI NORDIN
Diagnosis: Highly suggestive of malignancy (cancer) means that
there are findings that look like and probably are cancer. Patient
requires biopsy.
Size character: 2044
50.0311
Dimension:
320x608
Mammogram
20
Dimension:
320x608
56.0517
95
Mammogram
20
62.2653
Dimension:
320x608
Mammogram
20
68.0929
Dimension:
320x608
Mammogram
20
Dimension:
320x608
79.2798
96
Mammogram
20
80.5292
Dimension:
320x608
Mammogram
20
87.7308
Dimension:
320x608
Mammogram
20
93.7514
Dimension:
320x608
Label:
Shows the distorted location on the mammogram.
97
Data
Name: CIK FERESA BINTI MOHD FOOZY ID: ID001
MyKad: 840102067541
Age: 25
Date: 18/09/2009 , 11:01:48 AM
Physician Name: NORAITY BINTI NORDIN
Diagnosis: Highly suggestive of malignancy (cancer) means that
there are findings that look like and probably are cancer. Patient
requires biopsy.
Size character: 2044
LSB 1
Mammogram
25
50.7580
Dimension:
374x615
Mammogram
25
LSB 2
56.7786
Dimension:
374x615
98
Mammogram
25
LSB 3
62.7992
Dimension:
374x615
Mammogram
25
LSB 4
68.8198
Dimension:
374x615
Mammogram
25
LSB 5
79.4268
Dimension:
374x615
99
Mammogram
25
LSB 6
81.2561
Dimension:
374x615
Mammogram
25
LSB 7
.2767
Dimension:
374x615
Mammogram
25
LSB 8
93.7319
Dimension:
374x615
Label:
Shows the distorted location on the mammogram.