Applied Machine Learning and Multi-Criteria Decision-Making in Healthcare

Machine learning (ML) provides computational approaches for an updated knowledge that assists in processing ideas such as data precision. Studies using ML methods are driven by the use of technological approaches to assist the healthcare system. This work reports different significant studies on the applications of machine learning algorithms as alternatives to healthcare challenges. The goal was to identify the research areas concern with possible solutions.

Machine learning (ML) as a subset of Artificial intelligence is gradually getting accepted in the healthcare industry. Thousands of data are revealed to be analyzed from different sources in healthcare through machine learning algorithms. ML is unarguably essential in disease diagnoses and a variety of healthcare application services. In this study, the application of ML in healthcare was focused on using artificial intelligence techniques based on different scientific studies. ML is essential in carrying out special, efficient healthcare services with ease for health professionals.

Artificial intelligence tries to imitate human intelligence with powerful computer skills and aims to solve the problems that people have difficulty solving. Therefore, artificial intelligence and machine learning have begun to be applied thoughtfully in the field of healthcare and other lives. Remarkable results have been obtained in the performed research. These results have paved the way for applications in different healthcare areas and increase the frequency of the studies to achieve even more successful results in the same field. Prediction applications of machine learning are widely applied in healthcare as both class prediction and regression and support doctors and independent decision-making mechanisms. Even if classification studies of these applications have been performed, their diversity and excess numbers make them difficult and cause them to be considered on a subject basis. In this chapter, the usage of the terms “prediction,” “regression,” and “classification” in the literature is explained, and evaluation metrics used in all kinds of problem domains are defined. In addition to these, the problem areas of using machine learning techniques are summarized, and the literature search is performed in three scientific databases. Finally, the number of publications in the considered databases and an overview of the healthcare examples are presented. The data obtained and presented show that the applications using machine learning continue to increase significantly in healthcare and continue to be applied unlimitedly in all healthcare problems.

Health, considered the state of individual physical well-being, is greatly affected by factors, such as the physical environment and everything within an individual. Individuals in poor health require monitoring to be observed in the hospital, home, or anywhere being cared for. Many healthcare applications are being recorded and classified considering the health of humans, using Machine Learning algorithms. The healthcare system has rapidly advanced in technology. Applications in healthcare mainly comprise the IoT (Internet of Things), which are smart mobile devices, and the healthcare applications required for keeping track of individual health status, medical staff prescription, and medical history. In this chapter, modalities of Medical imaging and machine learning methods used in prediction will be studied. The classification in medical applications will also be studied and discussed in detail.

In recent years, the use of classification methods in machine learning, which is very popular among artificial intelligence methods, has been increasing due to the increase in the availability of health data. There are several classifier algorithms or methods in both supervised and unsupervised machine learning algorithms. Major unsupervised learning methods can be listed as cluster analysis and principal component analysis. Some notable examples of supervised machine learning algorithms are logistic regression, discriminant analysis, decision trees, nearest neighbor, neural network, naive Bayes, random forest, and support vector machine. In this chapter, binary logistic regression, one of the classification techniques in artificial intelligence, supervised machine learning, and econometrics, is theoretically discussed. In addition, the place and importance of this method in empirical applications in the field of health are briefly mentioned. In this section, an experimental study has also been conducted using the logit model to classify patients living in Adana province in Turkey according to their hospital services preferences. The data used in the study were collected by surveying in Adana. In the study, a binary logit model was used to classify patients and investigate the effects of many factors that may affect patient classification. Also, many tests have been conducted to investigate the classification ability of the model. As a result, the test results show that the model has good performance.

The state of individual physical well-being, which is considered as health, can be affected by the physical environment, as well as the factors going on within an individual. Poor health requires continuous monitoring for the individuals, and the process needs monitoring in hospitals or individuals’ homes. Furthermore, the recorded health data should be analyzed by specialists to be prepared for further processing, such as artificial intelligence (AI), using the latest environment of technology. Many healthcare applications considering AI have been studied, especially in cancer diagnosis and treatment, in recent years. Deep learning, expressed as the future of healthcare, is driven by increased computational power and huge datasets. In this chapter, an introduction to machine learning (ML), AI, and deep learning in healthcare will be studied.

In analyzing the Human immunodeficiency virus (HIV) epidemic dynamics, the biggest problem is uncertainty when planning for the future. In future evaluations, predicting what might happen will make the decisions’ results more realistic. Policymakers will have the opportunity to take precautions against any negative changes that may occur. Machine learning methods that produce good and effective predictive results are needed to plan future policies, eliminate the negativities and overcome deciding in an uncertain environment. In this study, seven machine learning models used to make time-series analysis for medical purposes are theoretically explained. Machine learning methods such as Linear Regression, RepTree, Alternating Model Trees, M5, k Nearest Neighbor (kNN), Autoregressive Integrated Moving Average (ARIMA), and Random Forest were used. The dynamics of the HIV epidemic in Turkey have been made stationary time series, considering compliance of the correlation. Then, the time series were preprocessed using the Moving Average technique, and the time series was softened. The time series is divided into 2/3 training and 1/3 test sets. Machine learning methods were trained using these sets, parameter optimization of models was made and tested. Then these models were used to forecast the HIV epidemic Dynamics in Turkey in 3 years between 2019-Q4 and 2022-Q3. The Random Forest method has been successful as the model that produces the least error rate (Mean Absolute Percentage Error, MAPE) among these seven models. According to the estimation results of the Random Forest model, R2 (the coefficient of determination) value was 82.16%, E (efficiency) value was 0.6268, Slope value was 2.3362, and MAPE value was 5.4132%. The Random Forest model has been observed to give excellent results for the three-year forecast of dynamics of the HIV epidemic in Turkey.

Artificial intelligence (AI) and Deep Learning (DL) started to play an active role in real-life problem solutions, and they have a rising trend across all medical fields, including dentistry. Since there are advanced improvements in image recognition techniques, a better radiological diagnose, prediction of the prognosis, and clinical decision making with a reduced workload are becoming possible for dentomaxillofacial radiologists. Promising results were obtained regarding dental caries detection, periapical/periodontal lesion detection, anatomical landmark localization, osteoporosis diagnosis, and implant dentistry; nonetheless, AI models do not substitute for most of the conventional processes yet. Further studies should be done to verify the feasibility and reliability of AI and DL applications in clinical practice. This chapter focuses on artificial intelligence and machine learning applications in dentomaxillofacial radiology.

The pandemic, which frightened the whole world, was reported in December 2019 as mass pneumonia cases in Wuhan city of China. The fact that the deadly new type of coronavirus can be transmitted extremely easily from person to person has also increased the spread of the disease. This spread negatively affects social, economic, and demographic life all over the world. This study aimed to identify which chronic and other diseases in combination with COVID-19 caused mortalities around the globe. As a result of the analysis, the appropriateness of the random effect unbalanced panel data model to the research purpose was determined. Coronavirus deaths related to the results of the Wald test used in the Generalized Least Squares (GLS) Technique, cardiovascular, diabetes, hypertension, respiratory disease, cancer, and other diseases are significant. In addition, the hierarchical clustering technique was applied to the meaningful model. According to the Ward Technique results, countries with similar chronic and other diseases for Coronavirus-related deaths were included in the same cluster.

On the other hand, the multi-layered Perceptron (MLP) model, one of the Artificial Neural Network (ANN) methods, was applied to the same model. The aim is to determine which chronic disease has a more significant effect on the Coronavirusrelated death factor. Literature research shows that hypertension disease ranks first in Corona-related deaths worldwide. The analysis of the MLP model made for this purpose determined that hypertension disease was in the first place in pandemic deaths.

We consider two strain influenza model with two vaccination in which strain 2 is the mutation of strain 1. A mutation is a significant change in the genetic makeup that arises due to errors during DNA replication or environmental factors. In this case, strain 2 was assumed to have developed due to alterations in the proteins that comprised strain 1. Proper vaccine administration is a critical component of a successful influenza control program. It is a key part of ensuring that vaccination is as safe and effective as possible. Unfortunately, it is easy to make vaccine administration errors. While certain vaccines administered incorrectly may be valid, such errors can leave patients vulnerable to disease. The aim of this chapter is to study the effect of administering strain 1 (V1) vaccine against strain 2 and strain 2 (V2) vaccine against strain.

In the previous chapter, two different types of vaccine were used for the two strain epidemic model; the results have demonstrated the significance of choosing the right vaccination. This chapter added delay to model (1), which is given in chapter 9. Here delay describes the incubation period. The model consists of four equilibrium points; disease-free equilibrium, endemic based to strain1, endemic with respect to strain2, and endemic with respect to both strains.

The global stability analysis of the equilibrium points was carried out through the use of Lyapunov functions. Two basic reproduction ratios r1and r2are found, and we have shown that if both are less than one, the disease dies out. If one of the ratios is less than one, an epidemic occurs with respect to the other. It was also shown that any strain with the highest basic reproduction ratio would automatically outperform the other strain, thereby eliminating it. Condition for the existence of endemic equilibria was also given.

Numerical simulations were carried out to support the analytic results and show the vaccine's effect for strain1 against strain 2 and the vaccine for strain 2 against strain 1. It is found that the population for infectives to strain 2 increases when the vaccine for strain1 is absent and vice versa.

Infertility is a major concern in health sciences. Many treatment strategies are being developed to overcome this problem. Assisted reproductive technologies (ART) are used in the treatment of infertility. Depending on the cause of infertility, different approaches can be applied in the field of ART treatments. However, since infertility is complex, it may be difficult to select the best treatment strategy for each patient. Mathematical modeling is used to understand better and sometimes even predict a pattern and outcome of biological processes. Thus, developing models help scientists and medical doctors select the best way of treatment and improve pregnancy rates. To date, a number of mathematical model systems have been tested to classify different parameters in infertile patients to develop a model that can predict the chances of becoming pregnant by identifying the behavioral design. In this chapter, several mathematical models are reviewed that corroborate the data obtained from infertile patients and predict the outcome depending on different parameters, including female age, follicle size, and hormonal levels.

Retinoblastoma is the most common primary intraocular malignancy in children (1 in 15.000-20.000 live birth). It involves an uncontrollable growth and division of cells in the retina (neurosensory layer of nerve cells lining the back of the eye). The most common first sign of retinoblastoma is a visible whiteness in the pupil called “cat's eye reflex” or leukocoria. Another sign is strabismus; eyes do not point in the same direction. Children with retinoblastoma may have inherited a gene from their parents that causes this disease. Retinoblastoma is a curable disease with a very low mortality rate; early diagnosis can result in a 95% chance of treatment success and survival of the patient, with a likelihood of saving 70 to 80% of the vision in the affected eye(s). This study aims to shed more light on the parameters that affect the different treatment alternatives of retinoblastoma and how these parameters affect the preference ranking of each technique. In this study, we analyzed and ranked the most common treatment techniques of retinoblastoma using fuzzy PROMETHEE (Preference ranking organization method for enrichment evaluations), a multi-criteria decision-making tool using some parameters. The analysis results based on the parameters, criteria, and weights used suggest that cryotherapy is the most favorable treatment technique for treating retinoblastoma, followed by thermotherapy, chemotherapy, photocoagulation, enucleation, and, finally, radiation therapy. Using fuzzy PROMETHEE allows the decision-maker to change the parameters, criteria, and weights according to the situation and the desired outcome. However, fuzzy PROMETHEE for this application is to aid the decision-maker in arriving at a decision and not be followed blindly without an expert opinion.

Hemorrhoids (Piles) are widespread diseases, and they appear in the form of itching, bleeding, and discomfort in the anal area, which are swelling of the veins of the anus and lower part of the rectum. Hemorrhoids may be internal to the rectum or external as they occur under the skin around the anus. It occurs due to increased pressure inside the abdominal cavity and pelvis due to constipation or pressure on veins during pregnancy. Many effective treatment options are available, and most sufferers can relieve symptoms by using home remedies, lifestyle changes after excluding other causes of anal bleeding. Most people over 30 years old have suffered from piles or some symptoms. Many patients do not like to explain that they have an infection in the anus or rectum region. As a result, such patients do internet searches for medications and remedies for this type of condition, which produces a plethora of misleading and superfluous material, as well as wrong information. Also, doctors sometimes cannot decide how to operate when dealing with hemorrhoid removal. For these reasons, we conducted this study to give information about all of the available treatment techniques and compare them using the multi-criteria decision making (MCDM) or multi-criteria decision analysis method fuzzy TOPSIS analysis technique. The treatment techniques discussed in this study are seven nonsurgical techniques (Fiber-diet with water, the enema, ice usage, creams, IR coagulation, rubber band ligation, and sclerotherapy) and three surgical techniques namely stapled hemorrhoidectomy, laser hemorrhoid surgery, open or closed hemorrhoidectomy. Also, the parameters used are (total cost, efficiency, recovery period, survival rate, practicality, comfortability, hospitalization time, and procedure time, all gathered from previously recorded information of hospitals, doctors, or patients. This information for each technique is analyzed, ranked, and compared using the previously mentioned TOPSIS MCDM technique.

Highly mutable organisms often challenge primer design for diagnostic PCR kit manufacturers due to new mutations occurring in hybridization sites. Novel variants may require reconsideration of the existing PCR primers and even result in misdiagnosis. While conserved sequences are often the main target of primer design algorithms, they often do not consider possible new mutants. We represent a generalizable algorithm for filtration of the sequence to identify conserved sequences and the less likely regions to mutate. Primers selected from the filtered sequences are expected to target regions with lower mutation rates and consecutively act indifferent to more variants of a target pathogen, providing long-lasting primers and less frequent primer redesign.

Leishmaniasis is a disease that affects humans, canids, rodents, and vertebrates. It is caused by different species of Leishmania protozoa, with tiny sand-fly insects as the incriminated vectors of the disease. Some cases of the diseases require treatment, while in other cases if left untreated, it can lead to the death of the affected victim. The factors that affect the distribution and abundance of these vectors are still not completely known. Therefore, it is of utmost importance to fill in the gaps necessary to fully understand the contributing factors to the abundance and distribution of the vectors and how these factors affect their behavior. Artificial Neural Networks is quite an accomplished problem-solving approach in computing and information technology-oriented devices. In this paper, backpropagation neural networks are applied in recognizing some vectors of leishmaniasis using some environmental factors and how these environmental factors contribute to the distribution and abundance of the vectors analyzed in this study.