Abstract
Skin cancer, which is one of the most common types of cancer in the world, is a malignant growth seen on the skin due to various reasons. There was an increase in the number of the cases of skin cancer nearly 200% between 2004-2009. Since the ozone layer is depleting, harmful rays reflected from the sun cannot be filtered. In this case, the likelihood of skin cancer will increase over the years and pose more risks for human beings. Early diagnosis is very significant as in all types of cancers. In this study, a mobile application is developed in order to detect whether the skin spots photographed by using the machine learning technique for early diagnosis have a suspicion of skin cancer. Thus, an auxiliary decision support system is developed that can be used both by the clinicians and individuals. For cases that are predicted to have a risk higher than a certain rate by the machine learning algorithm, early diagnosis could be initiated for the patients by consulting a physician when the case is considered to have a higher risk by machine learning algorithm.
TopIntroduction
The cancer is defined as the emergence of malignant tumors when the division of cells or tissues are seen aberrantly and multiply. The uncontrolled growth of cancer cells in the body of an individual is the main reason of cancer including distinct illnesses (Lancaster Cancer Center, 2020). Despite the wide variety of cancer types, these distinct diseases are seen as the result of abnormal cell growth. When a treatment is not provided for the disease, it can cause severe health problems and loss of lives. One of these types is skin cancer (Medical News Today, 2020). Skin cancer is usually divided into two groups. These are non-melanoma and melanoma types (Armstrong and Kricker, 1995). According to the report of the World Health Organization (WHO), a range between 2 and 3 mil. people experience non-melanoma type skin cancer every year, while more than 100000 people have melanoma type skin cancer (WHO Report-1, 2020). Although melanoma is not seen very commonly, the disease is the deadliest of skin cancer types (Kasap et al., 2015).
According to another report of the WHO, it has been predicted more than 9 mil. individuals died from cancer in 2018 (WHO Report-2, 2020). Skin cancer is commonly encountered as a cancer causing death. According to the data obtained from the report, the most common cancers are given in Figure 1. When Figure 1 is examined, skin cancer cases are seen very much over the world. As with any cancer case, skin cancer is likely to cause death. Considering this situation, skin cancer is one of the serious problems that need to be addressed.
Figure 1. Globally highly seen cancer types
Skin cancer occurs when an error (mutation) is seen in the cells’ DNA in the skin. Such mutations are the reason of growing uncontrollably and create a huge amount of cancerous cells (Narayanan et al., 2010). Much of the DNA damage in skin cells comes from ultraviolet (UV-ultraviolet light) radiation found in sunlight and the lights used in solariums. Besides, some factors will increase the risk of skin cancer (Saladi and Persaud, 2005). These include light skin color, sunburn history, excessive sun exposure, sunny or high-altitude climates, skin moles, precancerous skin lesions, family skin cancer history, personal skin cancer history, weakened immune system, and radiation exposure.
Motivation and Contribution
Dermatologists make the diagnosis of skin cancer mainly by visual evaluation of pathological skin. However, since this is a subjective assessment, it is mostly based on the experience of the dermatologist. With the advances in technology, computer-aided systems have started to be used in the determination of skin cancer as in many diseases. Especially with the development of image processing technologies, the detection of skin cancer can be performed more easily and accurately compared to painful and costly methods such as biopsy. In the literature, for the detection of skin cancer many methods based on image processing and application of computer algorithms have been preferred. However, mobile application-based diagnostic systems are limited and mobile applications are needed. In this study, a skin cancer detection system working on a mobile application system based on deep learning was developed. In this way, the skin cancer detection system, which can work directly on Android devices, is developed and contributed to the literature. When compared to other studies in the literature, the main contribution of this study can be summarized as follows:
- •
The most important advantage of the proposed diagnostic system is that can work on tablets and phones with Android operating system. The fact that the Android operating system is the most used mobile operation in the world will enable many patients or doctors to access the system.
- •
As patients and doctors are active on the developed system, early diagnosis will be provided. Thus, patients who are likely to have skin cancer detected by the system will be encouraged to apply to the nearest health institution. Thereby, the number of people who die from a dangerous disease, skin cancer, will be reduced.
- •
Skin cancer diagnosis system specific to the proposed Android operating system is enriched with CNN, one of the deep learning techniques. Since there is CNN in the infrastructure of the system, there is no direct feature extraction step based on image processing techniques. Because the feature extraction phase is automatically done by CNN.
- •
This study sheds light on how readers can design such a system and how to construct its infrastructure.
Key Terms in this Chapter
Deep Learning: It is an advanced version of artificial neural networks from machine learning techniques.
Segmentation: It is usually the first stage of image analysis. Image segmentation can be described as dividing an image into meaningful regions in which different features are held.
Melanoma: It is a skin cancer that begins in cells called melanocytes, which give the skin its color.
Machine Learning: It is the modeling of systems that make predictions by using mathematical and statistical processes on data.
Lesion: It is the general name given to any abnormal tissue in the organism that is often destroyed by disease or trauma.
Expert System: It is computer software used to solve problems in an information field. The logic of these software; when information is stored in databases and then problems are encountered, it is tried to reach results with inferences made on these databases.
Image Processing: It is a method that can identify with different techniques to obtain useful information based on digitized images according to the relevant need.