Abstract

The healthcare industries need to deal with the uncertainty in the patient data, random reports, vagueness in medicines and prescriptions, a patient’s mental and physical condition during testing, etc. The Fuzzy Logic (FL) can resolve these issues as it works similar to human minds based on approximate data. The algorithm removes redundancy from a massive database, summarizes, and develops crisp solutions. Its application in the healthcare decision-making process is fast developing due to the difficulties involved in designing and the development of a precise biological system model. The FL enforces researchers, medicos, and technology producers to explore smooth and creative solutions. It can be used as a versatile tool to model, manage, and control as our knowledge, and experience cannot be defined using explicit mathematical models. This motivates the authors to investigate the basic FL framework and analyze its application domains in this field. A descriptive study has been conducted to examine and focus on the healthcare decision-making process based on ranking, classification, data mining, feature selection, pattern recognition and optimization. It has also been addressed how FL-based strategies can be utilized to decrease uncertainty when working with simple and discriminative algorithms.

Keywords: Fuzzy logic, healthcare, crisp solution, decision making, fuzzifier, inference engine

3.1 Introduction

There are several context-depending scenarios where hard computing finds it difficult to provide an adequate model that can cater to all the requirements. For example, in a real-world environment such as healthcare, it is quite ambiguous to provide precise definition or meaning, or descriptions of many medical terms or relationships among important concepts accurately. The terms such as old, severe, long-time, certain point, etc. make the boundaries vague, unclear, and fuzzy. Similarly, many terms such as large, very large, extremely large, small, medium, very small, etc are relative and it is impossible to express them numerically [1, 2]. These factors make fuzzy logic an attractive alternative in the field of healthcare. Further, imperfect knowledge and the unpredictable nature of the available data lead to uncertainties in medical decision-making processes [3]. Some of the issues that make the development of effective healthcare identification system models can be viewed as below.

• Difficulty in understanding our biological mechanisms completely.
• The non-availability or ever-developing nature of test equipment.
• Difficulty in measuring the patient readings with precision.
• Ambiguous ranges of the standard value and test results.
• Presence of several conditions simultaneously.
• Missing of desired information and the presence of redundant data in many cases.

These factors make the traditional quantitative mechanism inadequate and complex. There always exists an inevitable and substantial amount of fuzziness while representing the biological system behaviour and its characteristics. This is due to the unavailability of an inappropriate mathematical framework to handle these entities involving a huge number of interrelated elements. Further, the presence of several variables makes these systems vague and fuzzy. Fuzzy sets can initiate notions of continuity into deductive thinking. In this way, these sets explore conventional symbolic systems in continuous form based on tabulated rules. As medicine exists in a continuous domain, the concept of continuous subset features such as fuzzy scores, fuzzy alarms, conventional scoring systems, etc. can accurately represent them. Similarly, fuzzy control is one such developed approach that utilizes a rule base to map the input into the output domain in continuous form. These logics can reason and quantify several ambiguous linguistic expressions in an appropriate form to alleviate imprecision and uncertainty.

In medicine, to address issues such as incompleteness, inconsistency, uncertainty, etc. it is not always essential to tackle micro-objectives and micro-phenomena. Due to a little bit of knowledge and it’s inexplicit as well as contrasting behavior much is happening in medicine than in physical sciences [4]. This kind of problem may add value to healthcare industries. These can assist strategic decision-making occupying complex consequences requiring immediate medical attention. This kind of inaccuracy or incompleteness can be envisaged in several ways such as:

• Information and knowledge about patients with varying categories, and attributes which is bound to be unpredictable.
• A patient’s case study and medical history described by him or her which may be highly subjective, exaggerated, simulated, or understated. Further, the failure to mention past operations, ignorance, or recollections of earlier symptoms or diseases often creates confusion or dents in successful diagnosis or treatment.
• Physical observation, in this physician, collect the data. They are doing errors and neglecting precious instructions, or also forgot to observe the whole test. There are some limitations between the regular and laboratory statement.
• Sometimes measurement errors in pathological tests, imperfect calculations, failure to focus on vital indications, overlooking minute details, and incomplete tests often mislead the medical practitioner. Similarly, faulty equipment, poorly managed and developed labs, inadequate knowledge, failure to interpret the range of standard and pathological status aggravate uncertainty.
• The accuracy of the ultrasonic, x-ray, and others laboratory examinations is highly dependent on the efficiency and knowledge of the lab technicians, reliable interpretation by the medical examiner, the behavior of the affected person before the samples are taken, optimal functioning of the testing equipment, etc.

Some of the most essential information required to evaluate a patient’s health situation lies in how accurately an examiner interprets the laboratory test results, medical history, and a patient’s behavior. It is highly unpredictable to conclude during an initial examination either physically or analytically, thus warranting further study on patients’ condition. There are many variations in reports, categorization of testing techniques, interpreted data, etc. makes the diagnosis vague and leads to uncertainty. Thus, the application of FL techniques has been widely explored in anaesthesia, internal medicine, electrophysiology, radiology, pharmacokinetics, etc. to model and control data. Further, it can help in diagnosis and prediction of disease based on data on asthma, cardiac heart, Alzheimer’s, diabetes, etc. using remote monitoring systems to help patients in distress. The sensor data on a patient’s health condition during standing, walking, jogging or running, sleeping or laying, and other activities can provide vital inputs to develop effective models using FL. Such analysis can accommodate patient care with comfort, safety, and convenience by the use of labelled data on a numerical scale. The objective of the fuzzy models is to develop effective recognition health monitoring models at lower cost and precision. FL decision-making healthcare models can determine the type and level of affected disease using data mining algorithms to patients irrespective of their age or gender. The latest updates of the patient’s condition, its severity, and the type of treatment to be prescribed based on the age can be estimated using the Fuzzy IF-THEN rules. There have been many soft computing approaches such as the Artificial Neural Network (ANN), Bayesian Network (BN), Fuzzy Inference System, Swarm Intelligence, Genetic Algorithms (GAs), and Fuzzy Cognitive Maps (FCMs) explored effectively to remove uncertainty and vagueness in this field [1]. Further, the application of FL can be extended to many service sectors to categorize analytic Hierarchy Processes that are reliable, tangible, responsive, and can assure empathy among medicos. Finally, the combination of Engineering and FL-based techniques can be applied to identify risks due to cancer, heart failure, kidney malfunction, diabetics, birth abnormalities, surgical complicacy, etc. in near future.

This piece of work analyses the application of Fuzzy Logic (FL) in healthcare industries based on earlier researches carried out in this field. The objective is to investigate and develop a concrete roadmap to forecast further developments in this field using the concept of fuzzy technology. Such analysis can provide vital clues on the application of FL techniques in healthcare industries to assist the decision-making process and remove uncertainty [5, 6]. Its capability to define relevant relationships set in a data system can assist the diagnosis, prediction, and treatment of the findings in several clinical outcomes [3]. The technology can be effectively applied to many vivid medical application domains which may be invasive, conservative, regional, neurological, basic sciences, signal and image processing, oriental, laboratory, nursing, etc. The focusing area mostly lies on the ability of the FL technology in handling uncertainty and improving decision-making in the field of medical industries.

3.2 Background

During the last few decades, the healthcare industries have grown leap and bound as compared to other computer-aided technologies. The exponential rise in medical industries and its vagueness has warranted the application of advanced technologies and creative strategies including FL modelling and artificial intelligence. The demand for FL techniques rises due to its extended application of the classical Boolean algebra in computer-aided healthcare applications. The technique connects the concepts to symbols, deals with the semantics of a corresponding domain, compares the associated constraints, and extends the particularizations similar to human reasoning [4]. The theories associated with Fuzzy set and fuzzy logic help to characterize uncertainty often encountered in the formation of medical concepts, interpretation of a patient condition, diagnosis of complex disease, and making therapeutic decisions. These Fuzzy sets are capable of defining the medical symptoms, test results, signs, diseases, diagnoses, prognostic, and, therapeutic information at ease and accurately, by conserving the obvious vagueness. Further, these logics can draw strict and approximate inferences reasonably. The technics suits a broad range of healthcare domain as mostly medical concepts requires causal, definitional, and heuristic, and statistical, knowledge. Sometimes, practical medicine has to accommodate incomplete, uncertain, vague empirical medical concepts to decide several medical procedures. Similarly, fuzzy automata can be utilized in real-time to deal with patient monitoring devices by accessing real-time medical information systems. It can be inferred that the use of FL and Fuzzy set theory can be further extended to develop knowledge-based healthcare systems such as interpreting the medical findings, disease diagnosis, prescribing optimal medical treatments, and monitoring the real-time patient data. A few applications of FL technologies have been explained below.

In Vienna General Hospital ICUs, the FuzzyKBWean has been used to control and optimize the ventilation and weaning process of patients during cardiac surgery. This is an open-loop fuzzy control system that accessed the Patient computer data management system fixed near the patient’s bed [5]. The FuzzyKBWean rules consider the patient’s physiological parameters as well as the actual ventilator settings using fuzzy antecedents and propose an appropriate new ventilator setting in time based on the rules. Correspondingly, the cooperating physicians can directly implement knowledge base changes in the ICU using the FuzzyKBWEdit rule editor [6].

To monitor patients’ data in the ICU intelligently, the FuzzyARDS/MONITOR has been used successfully [7]. It detects the acute respiratory distress syndrome (ARDS) of patients timely and accurately to provide adequate therapeutical advice. The fuzzy automata remain a suitable approach in this case since the ARDS is a medical entity very ill-defined. In this technique, the ARDS states consider as pathophysiological entries corresponding to several states of applicable therapies. Thus, the affected individuals can be assigned to one or more domains partially. The fuzzy states are assumed to be false or true or partially true while transiting from one state to another. These fuzzy states are used to represent several high-level medical entities which may include trends of linguistical expressions. These can be evaluated permanently using the steps of data-to-symbol conversion with an adjustable time granularity. Similarly, the FuzzyARDS/STUDY is a web-based study system that facilitates the entry of data at the different study centres which may include the fuzzy criteria, fuzzy scores, and evaluation of patient data using interval techniques to determine missing variables in the available repository.

In this regard, The CADIAG project helps several diagnostic processes of internal medicine using computer-aided technologies. It deals with many clinical issues such as identifying different diseases based on the pathological findings, further prescribing essential examinations to gain extended diagnostic hypotheses and to validate the supported findings, and searching for unaccounted patient’s pathological findings [8]. The CADIAG in Vienna General Hospital is fully integrated into its WAMIS medical information system. Its successor is named as MedFrame/CADIAG-IV which offers the positive and negative diagnostic hypotheses as well as several therapeutical proposals. The core of this technique include an integrated system that include the patient information and knowledge base system, editor modules, differential therapeutical and diagnostic modules as well as modules of immediate case evaluation.

3.3 Fuzzy Logic

The basic components of FL have been shown in Figure 3.1.

Fuzzy Logic has been applied in several home automation systems such as razors, Elevators, medical devices, subways, risk analysis in bank credit transactions, bond ratings, etc. It considers the approximate rather than the crisp values or values under certainty. In the input space, the logic assigns a number or a membership function to every element . The membership function (MF) aims to map each input value to a corresponding membership value which indicates its membership degree in a fuzzy set. The MFs assign the values in an interval of 0 and 1 [9]. Among several shapes of MFs, linear, trapezoidal, triangular, and Gaussian are most popular. However, the triangular MF is frequently chosen due to its ease of use and practicality [10].

The fuzzifier in FL does the fuzzification a process that converts a crisp entity in a fuzzy set to a grade of MFs representing the linguistic variables [10, 11]. The obtained fuzzy sets on fuzzification are then processed by the Inference Engine (IE) a processing unit based on the rule base. The IE is an important module of the FL system and the knowledge processor which draws conclusion or inferences based on the existing data and the expert’s reasoning. It maps a chosen input to the desired output using fuzzy sets. The knowledge base comprises the rule base and the database to determine the performance of an FL system [11]. The knowledge base contains several rules and uses the combination theory to choose the valid rules for the domain experts. A rule is considered to fire provided any of the parameters such as mild, moderate, severe, etc. is true else it does not fire.

The defuzzifier is used to translate the IE output into the crisp output. While the defuzzification input remains fuzzy the defuzzification output is crisp (a single number). The reason is usually the IE output is a fuzzy set, whereas while crisp values are mostly desired in a real-life scenario. The commonly used defuzzification mechanisms are

• max criterion
• centre-of-gravity
• the mean of maxima.

Figure 3.2 provides the advantages of the FL system while the disadvantages are provided in Figure 3.3.

3.4 Fuzzy Logic in Healthcare

The FL has been found applied in the field of healthcare either in single mode or hybrid mode. The hybrid mode is known as the neuro-fuzzy system which combines NN and FL or fuzzy plus Bayesian algorithm. The various ways in which the application of FL has been made are shown in Figure 3.4.

Clustering splits the data into clusters or groups and is a technique of data mining. It distinguishes the desired patterns by clustering them wherein similar features are kept in its corresponding cluster [10]. FL using the clustering techniques found its application in the field of images, cells, cancers, genes, etc. [11–13]. The Ranking is another FL mechanism to rank the risk factors, tests, suppliers, medicines, performance which assist in several strategic decision-making processes [14]. FL uses the classification mechanism group memberships for the predefined and known classes. This data mining FL technique has been found in many healthcare domains by several researchers [15–17]. In FL, the feature selection techniques aim to remove the redundant data and provide the relevant samples. This way, it reduces the memory space, increases the system response time, and releases the burden of the classifiers by preventing excessive training. The technique has been found to spot affected patients to get rid of unhealthy images, tissues, or cells [18, 19]. Pattern recognition is another FL application that deciphers the desired patterns in the corresponding subject. The authors used the time series analysis to determine the desired patterns based on time-dependent rules to identify medical images [20, 21]. In general, the medical data in the repository remains crisp. To extract vague information from the repository containing crisp requires advanced techniques such as Fuzzy query building, employing fuzzy ontology, Constructing fuzzy relation, etc. In Fuzzy query building, it is possible to fulfil the individual information needs, crisp relational query languages’ have been used though it remains inadequate [22]. For example, to develop a query like “exhibit the values of low liver enzyme for a child affected with celiac disease” we need to define the linguistic concepts in the language such as child or low. The objective of developing the queries to deal with complexity and impreciseness. Fuzzy SQL is an ORACLE-based fuzzy query language used to write flexible queries with the help of linguistic labels and fuzzy attributes in a crisp table.

Similarly, ontology can be used to represent a formal vocabulary specifying the domain knowledge as hierarchical concepts of structure, relations, and their attributes which can be read and understood by both man and machines. Thus, the medical ontology can conceptually represent the medical attributes in a symbolic way [23]. Nevertheless, these are based on the crisp concept, hence lack the desired elasticity or flexibility. It requires the application of FL in which the relationship among entities can be drawn based on the membership degree. The addendum of Database Management System (DBM) in FL is called fuzzy Relational DBMS in which the data in the database is represented using linguistic expressions. For example, the diabetics level in linguistic form can be represented as very low, or low, or high, or very high, etc.

The FL techniques have been widely explored in medical data mining applications. It uses data mining techniques to fetch both knowledge and information from the healthcare data repository for improving the quality and reducing the costs [24]. It is capable to deal with heterogeneous data involving symbolic, numerical, precise, approximate, ambiguous, and complexity. The FL techniques have been quite successful to describe these types of data with robustness being adaptable to the user’s environment and variations in different parameters [24]. Several data mining mechanisms such as machine learning, decision trees, comparison measures, clustering, association rules, etc. include the FL concepts [25]. The application of FL for the diagnosis of diseases using the data mining techniques such as the ANNs, DTs, and Bayesian network has been attempted by earlier researchers [26]. The algorithm has been used for the diagnosis of cancer and heart diseases in Bayesian and ANN platforms [25].

The FL Cognitive Maps (CMs) must support decision-making in complex healthcare industries The system must fetch the desired causal knowledge and information from the medical data repository, develop the knowledge base, and must make suitable inferences [2]. The CMs are the knowledge description schemes that link the symbolic ontology approaches to the raw patient healthcare data [23]. They provide the relationships between cause and effect with ease and interpretable forms. The Fuzzy CMs are the dynamic Fuzzy models to describe the relationships among different medical concepts [23]. These Fuzzy CMs are dynamic, tunable, and combinable to experience the relationships among the medical concepts which are often vague, hidden, imprecise, and uncertain [27]. However, the application of Fuzzy CMs in healthcare industries requires defining the degree of influence of one concept on the other and vice versa. It requires historical data to observe the influences and their level of weights. The FL techniques have been explored often in internal medicine such as gastroenterology, rheumatology, pulmonology, and hepatology [27, 28]. In these studies, Fuzzy labels are derived using soft matching techniques and are used to provide the relationships among the diagnoses and signs of human health status. It has been applied in expert systems such as the CADIAG in which the medical data available in the information system has been used to provide the desired medical information. This information is then fuzzi-fied based on the fuzzy rule base. The advanced CADIAG-2 has been found applied in the medical domains such as gastroenterology, rheumatology, and hepatology [28]. There are other FL-based expert systems found to be successfully explored are the RENOIR, SPHINX, and CLINAID. Figure 3.5 lists some of the application of FL in healthcare industries.

There have been considerable researches carried out to apply the FL in the evaluation and investigation of cardiovascular diseases, cardiac functions, cerebrovascular disease, ECG, cardiovascular medicine, artificial heart control, etc. [29–31]. TOTOMES is another FL application area designed and developed to determine the dynamics of cardiovascular disease during ventricular assistance. The technique interprets in multiple platforms to identify and detects different organ systems dynamically using fuzzy reasoning to estimate a patient’s health state. It can diagnose and detect the malfunctioning of a particular organ based on the fuzzy rule base. For example, tomography images have been associated with the FL techniques to classify the patients’ heart function with 94% accuracy [32]. The analysis and investigation of coronary artery disease, ECG classification, cardiac disease identification are other several applications discussed in the literature [28]. In pediatric ICU, the FL expert system has been used as a warning mechanism by regularly assessing the EEG abnormality level [33]. These automated systems can distinguish the critical condition of the patients and their artifacts. The burdens on the diagnosis of cardiopulmonary emergencies in ICUs can be tackled using improved patient monitoring systems in real time. One such FL expert system is the FLORIDA which uses the knowledge bases to determine a patient’s physiological condition in ICU. Similarly, the systems can be further extended to other healthcare domains such as breath detection and non-invasive sensor fusion [34]. It has been utilized to investigate the interpolation and the patterns of unborn babies based on the cardiogram reports which can prevent fatal injury or unnecessary medical complications [35]. Although, little work can be found on the application of FL techniques in Endocrinology or the determination of thyroid, the knowledge-based system has been widely used to monitor diabetics with hierarchical neural networks [36]. In oncology, the FL is concerned to classify discriminating cancer tissues, or breast tumours from the normal tissue with the help of 3-D ultrasonic echographic images for therapeutic advice [37].

The FL has been used to cluster Gerontology and in the veterinary expert system using several approaches corresponding to different subjects in a flexible way. The FL systems have been applied in determining the orthopedic symptoms for gait event identification of paraplegic subjects using electrically stimulated walking. There have been many reported pieces of literature in the application of FL techniques in anaesthesia to control the drug infusion, maintain adequate levels of anaesthesia, relaxation of muscles, patient alarm, and monitoring using virtual reality simulation and blood pressure [38]. The vague linguistic terms such as too long, or too close or too short, or too far can be used to measure the performance of the virtual reality surgical simulator. In this regard, the real-time expert system for decision making and control has been proposed to indicate the concentration of inhaled volatile anaesthetics [39]. The system has combined both the clinical data and online measurements in a Bayesian and fuzzy platform. Anaesthetists were confident enough to follow the dosage advice given by RESAC in most of the patients.

An application of this technology in plastic surgery to manipulate tools using an unsupervised segmentation method to create a new image of the patient has been another area worth noting in this direction [39]. Similarly, the technique helps to adapt and control the pharmacodynamic and phar-macokinetic parameter changes of the patient to prescribe treatments for muscle relaxation.

3.5 Conclusions

In this chapter, the application of FL in several fields of healthcare has been investigated. The natural disagreement among the medical staff, vagueness in medical concepts, uncertainty, and imprecise medical terms can be generalized using the fuzzy concept as revealed from our investigations. The techniques assist the computer and medical personal to learn and develop the desired concept with elastic modeling methods to accommodate themselves with smooth borders rather than approaching the decision models with binary approaches. As healthcare decision-making becomes more complex and tough, the exploration of FL techniques introduces creativeness to deal with the increased pace of developments. Different variations in the FL applications, such as ranking, clustering, data mining, classification, pattern identification, feature selection, and optimization have been discussed. The study reveals that the techniques can be applied in every strategic decision making concerning the healthcare domain. These areas involve database management, supply chain, diagnosis, health data mining, information retrieval, etc. Although Fuzzy nature makes the conventional methods in medical decision-making suffer from elasticity, however, the following advantages make it more promising.

• Flexibility: FL considers every possible value be it deterministic or blurred nature.
• Robustness: The models developed using the techniques robust as compared to the conventional methods of system modeling.
• Efficiency: It uses the available data adequately and represents them using crisp values and linguistic representation, hence remains efficient.

There are a few limitations that still haunt the FL technologies and require further research. For example, it requires considerable time and effort to design and develop Fuzzy systems. This results in a higher computation time to obtain the desired output models. However, the healthcare domain is rapidly progressing with the advent of new medical technologies, machines, medicines, equipment, and findings. There is an exponential rise is in supporting and allies’ branches that contribute to this field enormously with medical informatics. The funds and donations to healthcare industries also increase each day to support decision-making processes that require further studies in this area. Thus, the field cannot deny the benefits of FL in solving healthcare issues on a large scale. The integration of FL with artificial intelligence, robotics, image, and speech processing are a few trending topics the future researchers need to consider. The advancement in new, expensive, sophisticated image and voice processing techniques provides better quality and minute details of images concerning EEG, ECG, MRI, CT scan, etc., which increases the applicability of FL. Medical imaging remains a vast industry, wherein a huge amount of money is being spent, which creates the space for FL. There is every possibility of an increase in research funding in solving uncertainty and vagueness using the Fuzzy techniques in the medical domain. The Fuzzy imaging techniques and FSQL with ORACLE compatible make future research lucrative and promising. Due to competence, fuzzy database management is going to further escalate in the future. Similarly, the hybrid techniques involving ANN, FL, machine learning, and artificial intelligence are going to play an important role in determining future decision making in healthcare sectors.

References