Abstract
Cancer is one of the most common diseases nowadays, and it is a very heterogeneous disease that consists of several different subtypes. According to data from the World Health Organization (WHO), this disease caused death to approximately 10 million people during 2020, and in the same period, 19.3 million new cases were identified. Breast cancer is the most common cancer diagnosed for women, and lung cancer is the most detected cancer in men. Artificial intelligence has many different applications, and specifically, machine learning techniques are used for detecting and treating cancer. The methods associated with machine learning are computer algorithms that are considering different types of logic, and therefore, those types can be classified into supervised, unsupervised, reinforcement learning, self-supervised learning, active learning, etc. The main purpose of this work is to review and evaluate the different techniques associated to machine learning used by medical professionals but also by researchers with the main objectives of detecting and treating cancer.
Top2. Background
Cancer is a very heterogeneous disease and consists of several different subtypes and is one of the most challenging barrier to increase life expectancy (Bray et al. 2021). Cancer is a disease that starts with a transformation from normal cells to tumor cells and those continue growing to a pre-cancerous lesion and potentially into a malignant tumor. That critical transformation can be caused by 2 different categories of agents: genetic factors and external agents. Those external agents can be classified are physical carcinogens (for example, ionizing or ultraviolet radiation), chemical carcinogens (for example, tobacco smoke, food contaminants or drinking water contaminant) and biological carcinogens (for example, baterias, viruses or parasites) (WHO, 2021).
In the last 20 years, the cancer research has been in constant evolution and the scientific interest in the disease has increased exponentially (Hanahan & Weinberg, 2011).
According to information from the World Health Organization (WHO, 2021), cancer was the leading cause of death in 2020, causing around 10 million deaths in the mentioned year. In 2020, occurred a around 19.3 million in terms of new cases of cancer. The most common types are: breast, lung, colon and rectum, prostate, skin and stomach.
Cancer’s early detection is critical for decreasing mortality and this activity has two important approaches:
- 1.
Early diagnosis: It is key to diagnose cancer in an early phase to facilitate the treatment response and therefore, having a greater probability of survival. This includes also applying a less expensive treatment. The early diagnosis has 3 different components:
- a.
Being concerned about the symptoms of the different cancer types so the individual can identify if a visit to the specialist is needed.
- b.
Having access good clinical services that could diagnose the disease
- c.
Treatment referral when it is needed.
- 2.
Screening: The screening activity is crucial to identify different cancer types in an early phase. The screening programs are mainly based on the age and the risk factors, and aim to identify individuals with findings that could be indicating a potential tumor in a cheap and easy way, even before the patient develop symptoms. Once those individuals have been identified, then, more tests and exams are developed to confirm or reject the cancer diagnosis.
Key Terms in this Chapter
SVM: Support Vector Machines. It is a specific supervised learning model, that is usually applied in regression problems or in classification analysis.
Carcinoma: This is related to the cancer initiated in the organs, glands or even in the body structures. The carcinomas are linked to 80-905 of all the cancer cases detected. Some of those carcinoma’s types includes the melanoma, the Merkel cell carcinoma or the basal cell carcinoma.
Cancer: It is a group of diseases characterize by the abnormal growing process of cells in an uncontrollable manner.
CNN: Convolutional Neural Networks. It is a specific class of neural networks that are normally used to evaluate images.
Machine Learning: It is defined as group of techniques that combine mathematics with computational processing to learn patterns in a set of data with different purposes: classification, prediction, etc.
Diagnose: To identify or recognize the patterns that describe a specific disease.