Machine Learning Algorithms and Wireless Sensor Network Applied to Medical Diagnosis: A Systematic Review

In this research work systematic approach were used to conduct a survey on recent contributions of the authors that applied Machine learning algorithms or computational intelligence, Artificial intelligence and soft computing techniques such as Artificial Neural Network, Fuzzy logic Genetic algorithm, Artificial Immune System Swarm intelligence among others or any combination of these techniques Neuro-fuzzy, Adaptive Neuro-fuzzy, Neuro-genetic, fuzzy-genetic, and so on that is soft computing to medical diagnosis and also a systematic review on the application of wireless sensor network, wireless sensor is a veritable embedded system with a wireless communication function, and that is capable to: Collect physical quantities such as heat, humidity, temperature, vibration, radiation, sound, light, movement, etc. Convert them into digital values which are sent as sensed data to a remote processing station or base station (WSN) to medical and health care delivery. However the survey believed that combining wireless sensor network with soft computing techniques, artificial intelligence techniques can perform well in providing health care services compare to one technique because any of these techniques has certain limitations but together perhaps they two or three techniques connected together would reduce error to a minimum level.


Introduction
Machine Learning algorithms are also known as Computational Intelligence such as Artificial Neural Network (ANN), Fuzzy logic (FL), Genetic algorithm (GA), Swarm Intelligence (SI), Artificial Immune System (AIS), Simulated Annealing (SA), Genetic Programming (GP), Evolutionary Strategy (ES), Evolutionary Programming (EP), Multi Agent System (MAS) Ant Colony Optimization (ACO) and Particle Swarm Optimization (PSO) Probabilistic Reasoning (PB) and the combination of two or three of these techniques resulted in hybrid system technically known as soft computing which includes Neuro-fuzzy, is a combination of Neural network and fuzzy logic, Neuro-Genetic which is a combination of Neural Network and Genetic algorithm, Fuzzy-Genetic which is a combination of Fuzzy logic and Genetic algorithm, are widely used in modeling intelligent systems for medical diagnosis and prognosis.

Neuroscience and Neurocomputing
Neurocomputing also known as artificial neural network or connectionist is a mathematical representation of the human neural architecture, reflecting its "learning" and "generalization" abilities. For this reason, ANNs belong to the area of computational intelligence/artificial intelligences. It also belong to a machine learning algorithms. Neural network are widely applied in model highly non-linear systems in which the relationship among the variables is unknown or very complex [1].
Neuroscience also known as the science of brain. Human brain is a source of natural intelligence and a truly remarkable parallel computational system. Brain cells function about 10 6 times slower than electronic circuit gates, but human brains process visual, sense of touching and auditory information much faster than any modern computer could do in millions of years. The method behind this speed is the number of the neurons which is around 10 10 with the number of connections. As shown in figure 1.  [36].
Most of the computational intelligence techniques that are applied to medical diagnosis consider symptoms, biochemical analysis in labs and other feature like age, gender, as an inputs parameters to the systems, these would be further define with some linguistic variables like mild, moderate, severe, very severe, old, young, and so on, numeric values are assign to this variables which would form a datasets for training and testing of the system as illustrated in figure 1 below; where Artificial Neural Network are applied to medical diagnosis. Artificial Neural Network is machine learning algorithms that require datasets for training and testing. It is made up of inputs layers, hidden layers and the output layers. In inputs layers every input has some weight which are subject to adjustment.

Fuzzy Logic
The mathematical foundation of fuzzy logic rest on the shoulder of fuzzy sets. Fuzzy set was first introduced by Lofti Zadeh in which it derived from classical set theory. Fuzzy logic consist of inputs parameters knowledge based and outputs the number of inputs and output are strongly depend on the complexity of the system build. The inputs parameter has some linguistic variables which would be used to build the rules for the system. In fuzzy logic membership functions are also required such as trapezoidal, Gaussian, Singleton, Triangular etc. Fuzzy logic are also applied in medical diagnosis using either of its two methodologies that is Sugeno and Mamdani. The figure 2 below illustrates how Sugeno model used in medical diagnosis;

Membership Functions
In fuzzy logic there many membership functions applied to different research area such as Gaussian, triangular, trapezoidal, singleton among others which are all presented mathematically and graphically.

Triangular Membership Function
Triangular is a linear membership function illustrated in the figure below; where a, b and c represent the x coordinates of the three vertices of µA (x) in the fuzzy set A. The coordinate "a" is defined as the lower boundary in set A whose degree of membership is zero. The coordinate "c" is defined as the upper boundary whose degree of membership is also zero. Finally, coordinate "b" is the third apex of the triangle whose degree of membership is one. [66] As shown in the figure 4 below; Mathematically can be represented as;

Gaussian Membership Function
The Figure 5 below; illustrates a typical Gaussian membership function where x is the input variable, b is the center of the membership function and σ is the constant that represents the width of the membership function [66].

Bell-Shaped Membership Function [67]
A bell-shaped MF has symmetric shape and is given by the expression

A Trapezoidal Membership Function [67]
A trapezoidal MF is described by four parameters; a, b, c, and d and it is given by the expression

Soft Computing
In soft computing that is when two computational intelligence techniques are applied to medical diagnosis for example Neuro-fuzzy are applied in diagnosis of tuberculosis as shown in figure 3 below; Figure 6. Neuro-fuzzy models for TB diagnosis [27].

Machine Learning Algorithms Applied to Medical Diagnosis
In [38] Developed knowledge-based Expert system for diagnosing tropical infectious diseases on web based platform to receive input in the form of physical symptoms and results of complete blood examination in the laboratory. [16] Developed a Fuzzy Inference System model for medical diagnosis in the area of psychology and psychiatry. [17] Presented a kidney diseases Diagnosis system using fuzzy logic the results of the research indicated that, in the diagnosis of eight cases of kidney diseases through the examination of 21 indicators using fuzzy logic. [9] Developed a decision support system which predicts the possibility of heart disease risk of patients for the next tenyears using fuzzy logic and decision tree. [41] Used artificial neural networks with six factors (smoking, age, weight, height, sex and High blood lipid) for high blood pressure. [2] Developed an adaptive Neuro-Fuzzy Inference System for Diagnosis of Ebola Hemorrhagic Fever.
In the work of [3] developed a medical diagnostic system using Visual Prolog Programming language. [54] Fuzzy Abdominal Obesity models were developed on the basis of Body Mass Index, Waist Circumference and Weight to Height ratio measurements. [5] Developed a Fuzzy Expert system with 8 fuzzy rules for the diagnosis of hypertension. Fuzzy logic was applied in the work of [64] to health monitoring system using wireless sensor network for deciding final observation of patient health status based on bio-signal reading. [65] Have determined the relationship between risk factors (insulin, leptin, homocysteine, and urinary albumin excretion) and circadian Blood Pressure variations in essential hypertensive subjects. In [47] Artificial Immune Recognition System (AIRS) with fuzzy labeling was used to form an unsupervised learning AIRS approach and to substantiate tuberculosis diagnosis identification by using clinical tuberculosis diagnostic data sets. [15] Developed an expert system for human disease diagnosis it works on the patient data by combining production rules and a neural network. [19] investigated the risk factors for the development of diabetes mellitus, the hypertension, and the dyslipidemia simultaneously in a community-based observational cohort study using sex, age, Body Mass Index, smoking, alcohol and exercise parameters. [59] Proposed a Hybrid system for diagnosis chest diseases using Artificial Neural Network (ANN) and Genetic Algorithm (GA). In the research of [61] they have developed a Web based Fuzzy inference tool that evaluates the risk of exacerbating cardiovascular disease.
In [4] Develop an expert system for diagnosis of diseases thyroid cancer by combining the methods of neural network and certainty factor, In the research of [18] they have developed expert systems for diagnosis of asthma disease by the method of Fuzzy Logic. [32] Developed expert system for diagnosis of heart disease by the method of Fuzzy Logic. [60] In his work titled "Fuzzy Expert System for Medical Diagnosis" [14] developed a web-based Fuzzy Expert System for the public to diagnose the hypertension risk. [63] tested a logistic regression model to examine the independent influence of changes in body mass index, health related behaviors and social risk factors on changes in self-reported diagnosis of hypertension with using body mass index, physical activity, smoking, alcohol, acute condition, memory status, gender, residence, education, marital status, and income parameters.

Systematic Literature Review
Machine learning algorithms or Computational intelligence techniques are recently applied for classification, prediction of the presence of microorganism, biomedical medical image extraction, selection, and Medical diagnosis, optimization of medical system, estimation, pattern recognition, testing, data mining, and determination of severity level of disease among others. The contributions of authors that applied computational intelligence and/or Artificial intelligence techniques and soft computing to medical diagnosis are systematically summarize in the table 1 below;  [7] Medical image classification using ANFIS and Neuro-fuzzy 28 Hoang, et al., (2013) [21] Analysis on Heart beat Interval using Neuro-fuzzy 29 Umoh & Ntekop, (2013) [57] Presented a framework for diagnosing and monitoring Cholera using Fuzzy logic 30 Awotunde et al, (2014) [3] Fuzzy expert system for medical diagnosis 31 Bapita & Heena, (2014) [8] Classification of EMG based Diseases using Fuzzy Logic 32 Cinetha & Uma, (2014) [9] Fuzzy logic applied to Coronary Heart Disease 33 Neeraj, Anjali & Praveen (2014) [35] Early detection of Diabetes using soft computing 34 Shahaboddin et al., (2014) [47] Artificial Immune Recognition System applied in diagnosing TB Techniques used  35 Annie et al, (2015) [2] Diagnosis of Ebola using Adaptive Neuro-fuzzy inference system 36 Adlassnig, (2015) [6] Medical diagnosis using fuzzy sets theory 37 Filipe & Henrique, (2015) [20] Artificial Neural network applied to diabetes control 338

Wireless Sensor Network
The recent development of high performance microprocessors and novel sensing materials has stimulated great interest in the development of smart sensors physical, chemical, or biological sensors combined with integrated circuits. It is not uncommon to place multiple sensors on a single chip, with the integrated circuitry of the chip controlling all these sensors [68]. An emerging application for wireless sensor networks involves their use in medical care. In a hospital or clinic, outfitting every patient with tiny, wearable wireless vital sensors would allow doctors, nurses and other caregivers to continuously monitor the status of their patients. In an emergency or disaster scenario, the same technology would enable medics to more effectively care for large numbers of casualties. First responders could receive immediate notifications on any changes in patient status, such as respiratory failure or cardiac arrest. Wireless sensors could replace existing wired telemetry systems for many specific clinical applications, such as physical rehabilitation or long-term ambulatory monitoring [69]. [70]  cluster is designated as group leader called cluster head that is responsible of coordination of the sensors under its responsibility and act as a gateway to another cluster [71].

Application of Wireless Sensor Network to Health Care Systems
In the research of [72] Improving Health Care through Mobile Medical Devices and Sensors recommends that Mobile technology offers interesting ways to help with health care access, affordability and service delivery. [73] In their research paper they have identified challenges those developing countries particularly Tanzania will face while adopting e-healthcare. [73] agree that apart from the identified challenges; privacy, security and confidentiality are among the common challenges to both developed and developing countries. [74] They conduct a surveyed on the existing middleware design approaches for wireless sensor networks. [75] In their research paper present a monitoring system that has the capability to monitor physiological parameters from multiple patient bodies. [76] Developed a robust, scalable software platform for medical sensor networks, called Code Blue. [77] Believed that with all new technologies we face with problem of failure or success. Despite the promising features, risks of failure are always present. [78] In their paper, medical sensors were used to collect physiological data from patients and transmit it to Intelligent Personal digital Assistant (IPDA) using ZigBee/IEEE802.15.4 standard and to medical server using 3G communications. [79] Believed that the application of the Wireless Sensor Networks in healthcare systems is divided into three categories: monitoring of patients in clinical settings, home and elderly care center monitoring for chronic and elderly patients, and collection of long-term databases of clinical data.
[80] In their research paper; they describe the potential of biomedical smart sensors they also explain the challenges for wireless networking of human embedded smart sensor arrays and the preliminary approach for wireless networking of retina prosthesis.
[81] Their research paper evaluate the integration of Cloud computing, wireless sensor technology and Internet of Things in a healthcare environment.

Conclusion
Machine learning algorithms or Computational intelligence techniques are recently applied for classification, prediction of the presence of microorganism, biomedical medical image extraction, selection, and Medical diagnosis, optimization of medical system, estimation, pattern recognition, testing, data mining, and determination of severity level of disease among others. The contributions of authors that applied computational intelligence and/or Artificial intelligence techniques and soft computing to medical diagnosis are systematically. However, in this survey a system review were also conducted on application of wireless sensor network in health care services then we proposed that combining machine learning algorithms and wireless sensor network would improve the performance of health care system.