Mobile-Based Image Analysis System for Cervical Cancer Detection
Chapter One
Research Objective and Scope
This research work aims to design a mobile based digital image analysis system for cervical cancer detection. We seek to achieve the following objectives;
- To critical study several algorithms involved in digital image processing and analysis; To select an appropriate approach suitable for cervical cancer image analysis;
- To develop an Android application that serves as a demonstrator for the designed approach; and
- To integrate the applications and tools on an Android-based mobile device
CHAPTER TWO
LITERATURE REVIEW
Related Works
Much research work has been done in the area of digital image analysis for cervical cancer detection. This ranges from defining several algorithms and methods for the various stages involved in the digital image analysis. Also, great effort has been made in coming up with several Computer Assisted Design (CAD) solutions in this realm of digital imaging for cervical cancer detection.
In 2009, Yeshwanth Srinivasan et al. [7] proposed a unified framework for a fully automated system for the diagnosis of Cervical Intraepithelial Neoplasia (CIN). In their approach, several algorithms that are based on mathematical morphology, and clustering based on Gaussian Mixture Modelling (GMM) in a joint colour and geometric feature space are used to segment the micro regions of the cervix (mosaicism, vasculature and punctuation which made up the Acetowhite region [AW]). Not only are these algorithms of great importance in helping experts to evaluate the inter-capillary distance, which is the most important indicator of severe CIN, but they also help in quantifying precisely the extent of the abnormality.
In their methodology, the whole process of converting the raw cervix image data into a thorough diagnosis of CIN is broken down into six different modules: Segmentation of cervix ROI from the raw cervix image; removal of specular reflection (SR) segmentation of the cervix ROI into acetowhite (AW); columnar epithelium (CE) and squamous epithelium; classification of AW regions into AW, mosaic or punctuation tiles; segmentation of mosaic and punctuation from AW tiles; and assessment of the disease severity.
In their results, the segment obtained using the GMM based clustering correlates with visual perception. Also, the AW region observed, which was detached from the central AW region, is similar to the central AW regions (opalescent region – less intense AW region and which are clinically significant).
In 2014, Abhishek Das et al. in research work [6], proposed that the elimination of specular reflection and the identification of the ROI as the first step in the automated detection of cervical cancer using digital colposcopy. In their system, the whole scheme was decomposed into four different modules. Specular reflection removal (SR) from raw cervigram; segmentation of cervix ROI; segmentation of cervix ROI into Acetowhite (AW); columnar epithelium (CE) and squamous epithelium(SE); classification of AW regions into AW, mosaic or punctuation tiles. However, they were able to present the first two steps in their work. They took the advantage of advances in vision chip technology to enable high quality image processing in real time. They also took advantage of the fact that automated analysis algorithms based on modern image processing techniques have the potential to substitute clinical expertise, as a result of which there could be possible reduction in the cost of screening.
The specular reflections in the raw cervigram, which occurred as a as bright spots heavily saturated with white light, resulting from the light illuminating of moisture on the uneven surface of cervix, was removed by finding an interpolating function y that satisfies Laplace’s equation in n dimension.
For x € R, | | 0 is a solution of Laplace’s equation in n R – {0}. We notice that the function y defined in (1.1) satisfies Δy(x) = 0 for x 0, but at x = 0, Δy (0) is undefined.
The reason for choosing Laplace’s equation (among all possible partial differential equations) is that the solution to Laplace’s equation selects the smoothest possible interpolant.
K-means Clustering algorithm was used in segmenting the cervix ROI from the raw image. However, it is possible that the resulting ROI will consist of several disjoint areas in images, and the large one is chosen in this case and others ignored.
A dataset of 200 normal cervigrams and 40 acetowhite cervigrams was used. On visual inspection by domain experts (Gyneco-oncologists) of previous research results, their approach of removing specular reflection performs much better, as their algorithm smoothly interpolates the speckles.
Previous methods have considered only replacing the speckles in the image by blobs. This is a new technique.
CHAPTER THREE
CERVICAL CANCER
Introduction
All forms of cancer start when cells in the body (parts or organs) begin to grow out of control. Cells in nearly any body part can become cancerous, and if not promptly detected and cured, it can spread to other areas of the body [1].
Cancer of the cervix starts in the cells lining the cervix – the lower part of the uterus (womb). This is sometimes called the uterine cervix. The fetus grows within the upper part of the woman’s uterus. The cervix links the body of the uterus to the vagina, forming the birth canal. The part of the cervix closest to the body of the uterus is called the endocervix. The part that is next to the vagina is called the exocervix (or ectocervix) [15].
The two main types of cells covering the cervix are squamous cells (on the exocervix) and glandular cells (on the endocervix). These two cells meet at a place called the transformation zone. The exact location of the transformation zone changes as with age and after giving birth.
CHAPTER FOUR
APPROACH AND METHODS
Design Methodology
In any software project, a software development methodology is crucial. In fact it does not only determine the efficiency and the effectiveness of the system, it also helps in planning different activities that are involved in the development processes [23].
The Agile software development methodology was adopted for this project. It can simply be defined as a conceptual framework for undertaking software engineering projects [23]. Many varieties of this method exist with their peculiar features. These includes; Crystal methods, Dynamic System Development Models (DSDM) Extreme programming, Adaptive Software Development, Future Driven Development and Scrum [23, 24].
The desirable features of this development approach made it the best candidate of choice. Some of these include:
- Stakeholder engagement;
- High level of transparency;
- Early and predictable delivery;
- Predictable cost and schedule;
- Room for frequent changes ;and
- Focus on the business and the
CHAPTER FIVE
RESULT AND DISCUSSIONS
Data Set
The images were sourced from an online image database; Geneva Foundation for Medical Education and Research (GFMER) [24]. A total of 24 classified images was used in this research work. The classification was made by VIA by experts in the field of oncology. Out of these images, six images show no reaction with acetic acid, 15 indicate probably an early stage of cervical cancer (CIN I and few CIN II) while the last three indicate a chronic stage of the disease (CIN III and infiltrating cancer).
Performance and Measurement
In the last few decades, a large increase in the number of diagnostic tests in the field of medicine had been witness, and thus the need for personalized medicine will bring about rapid increase in this in the nearest future [25]. As a result of this, there is great need for careful evaluation of any potential testing procedure, so as to limit their potential danger to humans.
There are two basic diagnostic evaluations of these tests: Sensitivity
Specificity.
Some key terms will be used in defining the above terms:
True Positive: Detecting no disease/condition when it is actually present
True Negative: Detecting no disease when it is not present
False Positive: Detecting a disease/condition when it is not present
False Negative: detecting no disease when it is actually present.
CHAPTER SIX
SUMMARY, CONCLUSION AND FUTURE WORK
Summary
In this research work, we have been able to bring image analysis computation to a mobile platform. This is indeed a novel realm in cervical cancer imaging. The outcome of this is a portable tool that runs on Android platform. In our approach, we have used edges as a feature in classifying cervical cancer images. These images were taken with a low-resolution camera after the application of 5% acetic acid.
The irrelevant part of the image was segmented by using thresholding technique. The canny edge detection algorithm was implemented for the edge detection.
In addition to a robust algorithm being implemented, a simple and effective user interface was provided for both image acquisition and result display.
Conclusion
A thorough review of literature was done to ascertain the level of work done in this research field.
We have been able to devise a portable Android mobile application for the analysis of cervical cancer images. This system is limited to Android devices and some hybrid versions of Blackberry devices, thus leaving out the iOS devices user. The Android mobile platform was given preference due to its large user base.
All our set objectives in the introductory chapter of this research work have also been satisfactorily met.
Future Work
This study is limited by the number of data sample used (images). Therefore further research work is recommended with more image samples. This will not only assist in adequate testing and evaluation, it will aid using more sophisticated model that requires more images in both analysis and classification stages.
More efficient knowledge based approach especially in the image classification stage should be considered. This can be an online-based system which can be used as a decision support system in case of any ambiguity in the result offered by our system, which may be due to limited resources present on the host mobile devices, among other reasons.
REFERENCES
- American Cancer Cancer Facts & Figures 2015. Atlanta: American Cancer Society; 2015. Available at/www.cancer.org/acs/groups/content/@editorial/…/acspc-044552.pdf
- Suleiman Mustafa, Steve Adeshina, Mohammed Dauda and Wole Soboyejo “Classification of cervical cancer tissues using a novel low cost methodology for effective screening in rural settings” IEEE,2014
- World Health Organization Global Health Observatory Data Repository, Mortality and Global Health Estimates2012
- Abhishek Das, Avijit Kar, Debasis Bhattacharyya; “Detection of abnormal regions of precancerous lesions in Digitised Uterine Cervix images”; IEEE, ISBN 978-1-4799-3174- 3/14,2014
- Usman Arshad, Cecilia Mascolo and Marcus Mellor; “Exploting mobile computing in health-care”, IWSAWC,2014
- Abhishek Das, Avijit Kar, Debasis Bhattacharyya, “Elimination of Specular reflection and identification of ROI: The first step in automated detection of cervical cancer using digital colposcopy” IEEE, ISBN 978-1-61284-896-9/11,2014
