Artificial Intelligence Algorithms for Prediction and Diagnosis of Air Pollution Affecting Human Health

The complexity and growth of data in healthcare means that artificial intelligence (AI) will be increasingly applied in this area. This article (study) de-scribes the types of AI used by care providers and life and biophysical sciences companies. This paper describes the structure of the methodology of artificial intelligence (AI) algorithms and its machine learning (ML) subsystem with respect to the prediction of environmental pollution and its negative impact on humans. Key categories of AI applications focus on diagnosis and treatment or referral, patient engagement and adherence, and administrative activities. Diagnosis of diseases is a crucial and very important task of the physician in planning the right treatment and ensuring the health status of patients. The application of artificial intelligence (AI) can improve the level of diagnostic accuracy and efficiency. Indoor air quality (IAQ) is an important issue for well-being and good health, as most people spend almost all of their time in different types of buildings. Given the increasing availability of data and the rapid expansion of AI techniques, it is essential to explore the development of indoor and outdoor air quality predictions using AI techniques to improve and maintain IAQ.


Introduction
As a complex phenomenon, air pollution is a major problem worldwide because of its impact on the environment and the indoor environment of buildings, affecting the health of human beings.In this paper, we describe the impact of negative effects of environmental pollution on humans and its transformation into an artificial intelligence methodology and its machine learning (ML) subsystem in order to predict the negative impact of the environment with an early warning to intervene in human health.

Artificial intelligence, environment and human health
The degradation, decline and deterioration of environmental quality due to urbanization can be significant.It can be a major contributor to pollution and other problems related to sanitation, general waste management and provision of fresh drinking water.Today, the global environmental concern is the development of industry and transport, which is strongly influenced by the aforementioned urbanization, the development of smart cities, buildings, especially smart buildings, and the associated human population [1].Exhaust fumes from car traffic are a problem in large cities, while suburbs and smaller towns are polluted by airborne microparticles from local heating plants.Depending on their size, dust microparticles are trapped in different parts of the respiratory system.Larger particles attach themselves to the nasal mucosa, while smaller particles damage the lower respiratory tract and can reach the lungs, through which the ultra-fine particles travel on to the blood, which distributes them throughout the body.In the lungs, but also elsewhere in the respiratory system, the carcinogenicity of airborne dust particles can lead to the development of serious diseases.Tumors can develop from benzo(a)pyrene and carcinogenic metals.Other effects of particles can lead to an increase in cardiovascular and respiratory diseases, fetal damage in pregnant mothers, and links have been found with neurological diseases, diabetes, etc.The elderly, people with respiratory or heart disease, pregnant women and young children are most likely to experience these problems.
On the agenda is information from the World Health Organization (WHO), which has issued a report that has changed its approach to dealing with environmental issues.The report was that the presence of polluting small parts is significantly more dangerous to the climate than experts had assumed in their research solutions [3].
Air quality is assessed by an 'index' and is divided into several categories -very good, good, fair, fair, poor and very poor.Figure 1 shows that the effect of anthropogenic pollution has a significant impact on the environment, especially on humans in terms of quality of life and impact on health problems as presented in the previous section.Polluted air can shorten a person's life by up to six years.The biggest threat among the various factors is fossil fuels, especially coal.Air pollution has the worst effects in India, where people die on average six years earlier.People pollute the air, land and sea by burning fossil fuels, overusing chemicals and pesticides, and creating wastewater.The health consequences of this pollution are quite clear: air pollution causes about 8.8 million deaths a year worldwide.

Figure 1. The relationship between artificial pollutants with their impact on health and the environment
Climate change has been an increasingly debated issue recently.Currently, the main phenomenon to manage and prevent climate change is to reduce the amount of carbon dioxide released into the atmosphere.The way to achieve this is to use renewable energy sources such as solar, wind and hydro power.By using these energy sources, we can reduce the need for fossil fuels and the resulting carbon dioxide emissions.
Another way to manage and prevent climate change is to reduce the amount of deforestation.Deforestation contributes to higher levels of carbon dioxide in the atmosphere and reducing deforestation can help reduce the amount of carbon dioxide released.Planting more trees and protecting natural forests can help reduce the amount of carbon dioxide in the atmosphere.
Heat waves (HW) will become stronger and more frequent.This leads to events that can cause rapid changes in biodiversity patterns and in the structure and functioning of ecosystems due to humaninduced climate change [9].To overcome these problems, there is a need to focus on predicting air pollution, gaining knowledge about weather conditions and climate change and on this basis, implement an early warning system.
The concept of biotics in the field of biology is a rather complex process.It is generally referred to as biotic factors, which are living organisms, be they animals, plants and microorganisms.Biotic components of an ecosystem are all the living things that make up the ecosystem.In this case we are talking about animals, plants and microorganisms with reference to natural ecosystems.The easiest way to understand whether something is biotic is to ask whether it is a living element.Living components of the environment that influence the expression of a genetic factor are also considered biotic factors.These living things include mammals, birds, insects, arachnids, mollusks, and plants.Microorganisms are those included in the group of fungi, bacteria, viruses and nematodes.This is a fundamental reason why we must address this problem as a matter of priority.Every organism is shaped in its life and reproduction and subsequently shaped by its environment.Interactions between organisms and the environment are now being studied across ecosystems of all sizes, from microbial communities to the Earth as a whole.Environmental biophysical research in general today is geared towards making the subject matter understandable.This intelligibility is warranted: a) microclimate, b) how the organism functions in its microenvironment, c) how the organism responds to the influences of its microenvironment in a way that is naturally determined by nature or anthropogenic processes.
If we look at the need to combine physical principles with biological understanding, in the sense of humans better managing the planet in the context of quality of life, systems such as artificial intelligence algorithms need to be involved.Biophysics today involves a number of specific biological studies.It is a consequence of the relatively recent emergence of biophysics as a discipline.Biophysics ranges from sequence comparison to neural networks.In the recent past, biophysics has included the creation of mechanical limbs and nanomachines to regulate biological functions.Today they are referred to as fields belonging to bioengineering or nanotechnology.These definitions can be expected to be further refined.Biophysics is an interdisciplinary science that applies the theories and methods of the physical sciences to questions in biology and focuses on disciplines including: bioenergetics, cellular biophysics, channels, receptors and transporters, electrophysiology, membranes, muscle and contractility, nucleic acids, photobiophysics, proteins, supramolecular organization, spectroscopy, and many other areas.Environmental biophysicists are interested in the quantitative characterization of the physical environment in which biological organisms reside, including interactions between organisms and the environment.
The environment is affected in different ways and by different factors.Environmental elements that affect living organisms in any form are called environmental factors.Environmental factors can be divided into two main types: 1. Phytogenic factor.It is characterized by the suppression of the growth of certain plants or the development of animals by others under the influence of root secretions, the growth of root systems, and so on.2. Zoogenic environmental factors.This is the effect of animals on themselves or on representatives of the plant world.This factor is called the "group effect".
Areas of environmental biophysics include: energy exchange, flow of matter and momentum, conservation of energy and matter, temperature, water and water vapour, wind, properties of gases and liquids, conduction of heat and matter, heat flow of soil, radiation and other focal points.In assessing these influences, we can assume that the difference between biophysics and environmental biophysics is primarily a matter of scale.
Biophysics is an all-encompassing term used to describe the application of physical principles to biological phenomena.The potential contribution of biophysics to the field of medicine was not realized until about the 1950s, after significant advances in physical measurement techniques.Early efforts sought to understand and describe biological phenomena such as hemoglobin dissociation and cell-cell interactions and structure.The field then progressed to more complex tasks such as computer simulation of blood flow and tissue perfusion using models of bio-logical compartments dependent on vascular parameters such as vascular density, perfusion rate, and permeability.More recently, the combination of artificial intelligence methods with biophysical principles has resulted in paradigm shifts in areas such as the study of protein folding and structure and promises to modify drug discovery processes.
Few industries have more benefits and opportunities from the Internet of Things (IoT) than "healthcare" itself: sensorics offers huge benefits at every level, from inpatient treatment and community healthcare to social services and self-care.It provides new opportunities for e.g.disease prevention through screening and early detection.IoT technologies that are capable of automatically collecting and processing patient data allow us to capture early indicators of disease in the early stages of disease, when progression and adverse consequences can still be prevented.With the availability of a wide range of telemedicine solutions, the IoT has the potential to break down barriers of distance in healthcare, which are particularly acute in the regions.Some policy initiatives have accelerated the introduction of these remote patient monitoring systems in Central and Eastern Europe -for example, in the Czech Republic, where the General Health Insurance Fund (VZP) has started to cover remote monitoring for certain categories of patients.Thanks to the IoT, a wealth of health-relevant data will be available to inform healthcare decision-making.
Algorithms and modelling are taking healthcare to a new level: human empathy and expertise combined with the knowledge and precision of processors means better healthcare, less pressure on doctors and nurses, and more accurate diagnoses of diseases.In fact, healthcare, and healthcare in general, is where AI in both practical and experimental form is perhaps most prevalent.The implementation of AI in practice is so broad that it is difficult to define it into any groups or components; from booking a doctor's visit and consultation, to diagnosis, research and decision making, to treatment and care of the patient, to psychological support after treatment or palliative care.
An example might be how much the current healthcare system relies on technology other than covid.The discovery of the virus, the preparation, the controls, the diagnostics -it is all technological.Prevention, treatment, care -technology.Research, vaccine -technology.
Artificial intelligence, as we have already described, can be compared to human reasoning abilities.Figure 2 shows the relationship between artificial intelligence, ML algorithms and deep learning.Since AI is capable of managing complex, especially non-linear interactions, this feature makes it the best mathematical method for disease prediction and also for air pollution prediction.[11] Another indispensable function of AI is its ability to handle big data management and any complex problems of life in the world.The form of weather forecast modeling helps the designers to define algorithms and thus help other components in projection capabilities in the matter of efficiency and logic [12].
It also offers an outlook describing the impact of IoT technologies on various aspects of healthcare, including quality of care, physical access to healthcare services, patient quality of life and satisfaction.A phenomenon that will bring about an active role of the patient in healthcare then healthcare consumers will gradually become able to assess and collect more and more relevant data related to their own health, enabling the birth of a new type of BYOD -bring your own data.

Deep Learning
-Deep learning uses multiple-layered neural networks to build algorithms that find the best way to perform tasks on their own, based on vast sets of data.
-Subset of ML which make the computation of multilayer neural network feasible.

Machine Learning
-Machine learning employs algorithms that learn from data to make predictions and performance improves when exposed to more data over time.
-Subset of AI techniques which use statistical methods to enable machines to improve with experiences.

Artificial Intelligence
-Al is an umbrella term for machines capable of perception, logic and learning.
-Any technique which enables computers to mimic human behavior.
Let us return to the issue of the effect of air on the human body.The current era is characterized by an increased percentage of emissions.Atmospheric air has become more toxic and contaminated.The main air pollutants include CO2, NO2, SO2, CO and particulate matter, all of which have adverse effects on the ecosystem and human health.In Figure 3 we can see the main air pollutants and their causes on the environment and human health.These pollutants cause greenhouse effect, ozone depletion and photochemical smog in the atmosphere, which deteriorate the environment and cause major environmental disasters [16], which are responsible for several serious health problems (i.e., chronic diseases) [17].For example, the main cause of sulfur dioxide (SO2) is volcanic eruption and industrial growth.In addition, the presence of SO2 in the atmosphere affects the acidity of rainfall which damages the ecosystem and also leads to the deterioration of the human respiratory system [18].

Figure 3. Main air pollutants and their sources and effects
Poor air quality promotes a strong negative effect on a person's energy and thus also affects mental illness [17,18].Air quality monitoring is in place to address this negative influence.Here, the application of artificial intelligence can work well especially in designing energy sustainable smart cities. Fig. 4 shows a model for predicting the polluted environment.Environmental care is a topical issue today and takes three forms: • Negative impacts of human activities.
• Negative human intervention in the landscape and construction of buildings with inappropriate architecture.• Urban planning must be taken seriously and soberly.

Major Air Pollutants
• Particulate matter is made up of several distinct components.Some are directly released by industry, while others are produced by atmospheric processes.They produce haze and, if breathed, can cause respiratory issues.

PM
• In road transport, burning is the primary source of pollution.Nitrous oxide is a significant contribution to global warming.Acid rain and ground-level ozone production are mostly caused by nitrogen dioxide.It has a high prevalence of lung disease.
NO 2 • The principal source of sulphur dioxide is the combustion of fossil fuels for power generation.It can contribute to pollution; generate acid rain when combined with water, and induce wheezing and asthma.
SO 2 • Although the ozone layer protects humans from Ultra-violet radiation, ground-level ozone is a serious polluter.In the presence of sunlight, it is created from other contaminants.Ozone is a main component of smog and can be harmful to one's health.
O 3 • A gas produced by the combustion of fossil fuels in the generation of electricity.Natural processes also release it.Human emissions are connected to growing CO2 levels in the atmosphere and anthropogenic global warming.
CO 2 • A gas produced by the incomplete combustion of fuels, most notably from automobile travel.It has an impact on human health because it decreases the oxygen-carrying capacity of the blood.It also produces ozone when it combines with other atmospheric gases.

Figure 4. Air pollution prediction models
The updating and application of essential hygiene requirements in environmental care is based mainly on knowledge of the environment and weather conditions that adversely affect human heal Artificial intelligence has a significant influence in this concept in the context of applied methods for predicting environmental pollution.The input parameters for environmental prediction are the above two forms of environmental care.Here, then, the application of the artificial neural network method is offered.
A number of harmful substances, such as moisture, carbon dioxide or various odors and smells, are generated by human residence and activity in indoor spaces.Studies show that when the indoor environment is improved, performance and productivity at work also improve.
The environment has a significant impact on a person's lifestyle, and especially on their health.It is very difficult to determine the degree of environmental influence on a person, especially based on the state under normal conditions.This process is dependent on somatization, i.e., it is necessary to understand the state of the human body in contrast to its soul.That is to say, the human brain is so arranged that its mental processes are materially manifested in the form of motor autonomic reactions.

Artificial neural network as a model for air pollution prediction in relation to the diagnosis of the resulting disease
An artificial neural network is a concept that is inherent in a certain kind of computer technology that attempts to mimic the human brain.An artificial neural network, or ANN, involves simulated neurons and stimuli to attempt to reproduce brain functions.This wide range of software and equipment uses If you want to model non-linear dependencies in the case we have already described, then Artificial Neural Network (ANN) is a more common and evolving option for air pollution prediction.ANN requires much less stringency in statistical constraints, so it is necessary to model complicated nonlinear interactions and train many algorithms in this case.For example, a hybrid ANN model has a wider range of possibilities by integrating many methodologies within a single computational model than a single ANN model.
Artificial intelligence in an environmental quality assessment system can more easily diagnose specific patient diseases based on its predictions.At the end of this treatise, we can conclude that the future of "conventional medicine" is closer than we realize, with patients seeing the computer first and the doctor second.This feature allows for more time to be spent with the patient and thus also reduces fatigue.
Artificial neural network is one of the computational models used in artificial intelligence.Its model is the behavior of the corresponding biological structures.An artificial neural network is a structure designed for distributed parallel data processing.
Reinforcement learning means that a system of rewards and punishments is used to train the algorithm.That is, the so-called minimization of the function in the presence of a penalty is introduced.Artificial intelligence, unlike human intelligence, is machine intelligence [19].
Artificial intelligence is an umbrella field of machine learning where machines simulate the human mind [20].Artificial intelligence is the interplay between the interplay between software and hardware.AI is the creation of an algorithm and then a module, software, is created.If we want to define a conceptual framework that implements AI algorithms, it is expressed with the help of artificial neural network [21].Such an artificial neural network mimics the behavior of the human brain, which is characterized by interconnected neurons that undergo weighted communication between channels [22].
The main goal of health-related AI applications is to analyze the relationships between treatment or prevention techniques and patient outcomes [23].Artificial intelligence is mostly used in procedures such as diagnostic processes, development of treatment protocols, drug development, personalized medicine, and patient monitoring and care.
In the case of healthcare, artificial intelligence is constantly evolving.Currently, it is applied in the frequency of several diseases, and these are: • cancer level assessment [24], • diseases of the nervous system [25], • heart disease and cardiac disease [26].
For example, consider a vector machine that is used to classify any particular object classified into two groups.We have a result Yi and we have a classifier Yi-1 or 1, which represents whether the i-th patient is in group 1 or 2, in a predetermined order.We can then define that the subjects can be divided into two groups by introducing a decision boundary defined on the features (neuron inputs) xij, which can then be written (see Figure 5): where wj is the synaptic weight placed on the j-th feature to manifest and influence the outcome of the others.An important feature of SVM is the determination of the model parameters.The decision rule then implies that if yi>0, the i-th patient is assigned to group 1, i.e., label Yi=-1; if yi<0, the patient is assigned to group 2, i.e., label Yi=1.The membership class is indeterminate for yi=0 points.Subsequently, b is the threshold of the neuron, indicating the threshold of neuron activation.That is, if is less than the threshold, the neuron is passive (inhibited) and if ∑ (    )  =1 is greater than the threshold, the neuron is active (excited).An artificial neural network consists of artificial (formal) neurons, whose precursor is a biological neuron.The neurons are interconnected by synaptic links and transmit signals to each other and transform them by means of activating transfer functions.A neuron has any number of inputs but only one output.

Figure 5. Artificial neuron model, neural network
The activation (transfer) function of a neuron in artificial neural networks defines the output of the neuron when given a set of neuron inputs.Nonlinear activation functions allow neural networks to solve non-trivial problems with a small number of neurons.A sigmoid with parameters of steepness (defining the bandwidth of the neuron's sensitivity to its activation potential) and threshold (defining the offset of the origin of the function) is often used, along with its limiting forms such as linearity for steepness approaching infinity and sharp nonlinearity for steepness approaching zero: The goal of neural network learning is to set up the network to give the most ac-curate results.In biological networks, experience is stored in synapses.In artificial neural networks, experiences are stored in their mathematical equivalent -weights.We distinguish neural network learning into learning with a teacher and learning without a teacher.The learning phase of a neural network is called adaptive, and the equipping phase after the neural network has been learned is called active.

The error back propagation algorithm.
The error back propagation algorithm is a typical learning method of artificial neural networks.It is used in multilayer networks for learning with a tutor, i.e., when the desired outcome is always known on the set of examples used for learning.Back propagation of error is based on the gradient descent method.The learning quality of an artificial neural network is described by an error function, most often a quadratic error: Where E is the error, N is the number of samples presented to the network, n is the number of neurons in the output layer, zij is the desired output of a given neuron and a given sample, yij is the computed output of a given neuron and a given sample, ∆w is the vector of increments of the gradient step weights, and α is the magnitude of the gradient step.
The goal of learning is to minimize the error function depending on the input weights of the neurons, and the gradient descent will generally only find a local minimum, so some inertia is introduced into the gradient descent by respecting the direction of its past descent step, i.e., the past gradient is added to the current gradient and the current descent step is performed in the direction of their sum, this deformation of the gradient descent then allows the gradient descent to slip out of a shallow local minimum.Learning a neural network involves changing the weights of the neuron inputs.The error back-propagation algorithm proceeds in the following three stages at each step: • Samples are applied and the outputs are counted for each sample in a forward direction (the input signal propagates forward through the network).• The counted outputs are compared with the desired outputs, i.e. the error is calculated as described above.• Based on the computed error, the values of the adaptation functions are computed in the direction from the last layer to the first layer (to compute the values of the adaptation functions of the subordinate layer, the values of the adaptation functions of the superordinate layer must 11 already be computed), i.e., the gradient of the error function is computed, based on which a gradient descent step is performed, i.e., the input weights of the neurons are adjusted so that the error value decreases.Thus, the computation proceeds backwards from the output layer to the input layer (hence the backward propagation of the error), the weights changing according to their influence on the error.

Disease diagnosis solved by artificial neural network
Deep learning is a set of machine learning methods inspired by information pro-cessing and distributed communication in a network of biological neurons.Artificial neural networks should be trained in deep learning.ANNs are artificial neural networks.
Artificial intelligence (AI) is essentially a simulation of human intelligence.It is currently used in many applications, be it cloud, enterprise, healthcare or consumer applications, and even in embedded firmware.Artificial intelligence is designed to perform tasks from the simplest to the most complex.The goal is to mimic human cognitive activities.
AI can be applied in the healthcare industry in various healthcare careers.Data in healthcare can be structured but also unstructured.AI machines and applicators include machine learning methods especially for structured data and neural networks and advanced deep learning for unstructured data.The main areas in healthcare and human disease in general, where AI tools are currently being used include cancer, neurology and cardiology.
In general, AI and its applications, especially in healthcare, make us wonder whether its tools can replace human phenomena such as the doctor himself.In the practical case, AI tools cannot replace doctors, but they can help doctors achieve better results in precision work in defined medical areas.
A prerequisite for this definition is the need to work from a range of data collected in the process of treatment in a healthcare facility.Artificial intelligence, as we have already mentioned is a technology that is based on a number of defined technologies from the platform of applied informatics.Also, we have already mentioned that the most widely used technology in healthcare is machine learning.Machine learning is mostly used in so-called precision medicine [27].This type of determination will require training the model using datasets, then this approach is called supervised (learner) learning.
If it's a diagnosis of a given disease, then it determines the specific medical condition, the nature of the disease or the health problems that the patient may have.Diagnosing a disease can sometimes be a very easy task, but on the other hand it can be a more or less difficult task.There are large data sets available, but on the other hand there are limitations in the tools available that can accurately identify patterns and their consequences or predictions.
The methods used simultaneously in the diagnosis of a given patient's disease are carried out intuitively based on the experience and knowledge of the doctor.Therefore, such a diagnosis is prone to error.If we use some of the artificial intelligence (AI) methods in diagnosing a patient, such methods are definitely more accurate than manual experience.Therefore, we can conclude that the so-called automatic diagnosis of a patient reduces its error rate significantly compared to the current state of diagnosis.This eliminates the errors caused by the human factor even if it has high expertise.This is moving medical science towards more accurate assessment of patient's health status and consequently also towards very significant cost saving measures on a comprehensive scale such as spending a lot of energy, time, speed and more accurate treatment of patients, faster cure of patients and hence their participation in the workforce and faster recovery etc.
Experimental and statistical research of a disease is characterized by the fact that some tissue is damaged or destroyed, causing, for example, various cytological reactions, and this is technically called a pathological process [28].The typical manifestation is a symptom of pain or redness of the tissue with swelling.Such a disease has its own symptoms, which are then defined by a specialist (physician) in the clinic [29][30][31].The diagnosis is then determined based on the medical condition that results from the disease accompanying and producing evidence of it.

Figure 6. Diagram of the diagnostic process
The basic importance of pathology therefore lies in the diagnosis and treatment of living patients and not, as the general public mistakenly believes, in autopsies of the deceased.Autopsies, which are primarily used to determine the cause of death, continue to be part of the field, although to a much lesser extent than they used to be and form only a small part of the pathologist's work.This is mainly due to the development of new diagnostic methods and their improvement. 2 We can practically state that at least one diagnostic procedure (test) is performed during this procedure.Based on experience, we can state that in such a procedure, practically one diagnostic procedure or test is performed.In the wake of this statement, the physician's work follows in the diagnosis procedure, which involves a logical procedure, based mainly on experience, that allows the physician to gather information about the disease in question [33].The diagnosis of diseases is also the most challenging process, it is a very fundamental phenomenon for the medical professional (physician), as well as before reaching a conclusion to the disease.
Proper treatment of a patient may be delayed or missed due to serious health problems because of a mistake in the diagnostic process.Unfortunately, not all physicians have expertise in every area of the medical field.Doctors working in hospitals gain a great deal of experience because they are constantly in contact with patients and thus, on the other hand, gain the necessary experience both in terms of knowledge of the disease process and in the application of the electronic medical equipment installed.When it comes to the automation of diagnostic systems and medical technological equipment, physicians gain such experience that they can also design them, in particular, recommend the application of such functions and devices that can then be used in the design of automated medical devices at the level of "machine solutions", such as medical electronic equipment, medical instruments, measuring equipment, X-ray, etc [34].
On this basis, a prerequisite for the achievement of quality research on diagnostic activities in the health care facility was created, and at the same time, the opinion about the creation of a system of cost reduction, which is needed in the design of a decision support system.
To classify and diagnose a particular disease, it is necessary to master various parameters related to the patient's respective disease, which requires a deep knowledge of the doctor and his experience.In this case, we can subject it to artificial intelligence.In current medical practice, various medical devices, medical electronic devices, and also in some cases artificial intelligence-based devices and techniques are emerging.In this case, these are medical diagnostic methods dealing with accurate diagnostic results defining the most accurate assessment of a given disease.We must state at this point that artificial intelligence is an indispensable part of applied informatics.Applied informatics extends to practically all fields of human activity, from healthcare, culture, sports to environmental assessment, economics, technology and industrial practice.The techniques used in artificial intelligence are based on cluster analysis methods, neural networks, optimization methods, as well as deep learning and machine learning.Today, such systems are called or named as intelligent systems in the field.
Gradually, intelligent health systems are being replaced by automated AI-based techniques where human intervention is negligible [35], [36].As an example, an artificial neural network (ANN) is a set 2 https://www.euclaboratore.cz/onas/obory/patologie/?utm_source=seznam&utm_medium=cpc&utm_campaign=VISI_CZ_S_SE_Slu%c5%beby &utm_content=Patologie&utm_id=885980&sznclid=xayh-PT2_Pzy_Pb08PX38v3y8Pb99ve5sfj08_Py_fP09_z26_Tz9bmxoPj08_z08fLw8vDz6_Ty9bmm-PGG8fz09fSAg4H1gPOH9Pfw9vHy8_Dw9_3zh4bwhP39 Step 1: The Medical History Step of neural units that are designed similarly to the structure of biological neurons in the human brain as well as other living organisms.Practically, it is a kind of simulation of the human brain, whose function is modeled on some model device or digital twin in the form of the resulting intelligent device -system.If we look at the structure of the nervous system, we find that each neural unit is connected to many other neurons, which is the so-called bipartite graph [37].Such systems are learning systems, which are then trained completely automatically.We have already mentioned that a given disease in any individual is a complex and lengthy process in terms of defining it.We can shorten this lengthy process, and very significantly, by applying artificial intelligence.The application of intelligent systems in healthcare can collect an enormous amount of data that needs to be understood and also be able to detect even if only a small part of it, which can then be used as a basis for defining the prediction of morbidity and disease [38], [39].All this then leads to accurate diagnostic results and thus to more accurate decisions.When deep learning is applied in various areas of medical practice, we can better and accurately search for relevant drugs, better determine medical outcomes when imaging a case, and we can also detect Alzheimer's disease [40], etc.It has recently been published that the application of automated breast cancer detection has resulted in significantly higher breast cancer detection and even higher accuracy than a radiologist can detect alone.Another resource and innovative factor in healthcare is the Internet.We now know that the relevant software, which is based on an artificial intelligence platform, detects disease far earlier and well in advance of the disease occurring.It is that particular disease detection, applied in the most accurate way possible, that underpins the correct and precise treatment provided.
An expert system is a system that is based on the representation of expert knowledge that uses a computer system to solve given tasks.Or another definition is that an expert system is a computer system that searches for a solution to a problem within the scope of a certain set of propositions or a certain cluster of knowledge that have been formulated by experts for a specified domain.
Expert systems as described above are the basis for evaluating the overall future intelligent system in order to define a structural medical expert system.Such an expert system would be the basis for implementing a medical system with technology aimed at evaluating and defining diagnostic methods along with prescribing treatments with the most appropriate treatment and deployment of targeted drugs.
Fuzzy logic is generally used (as opposed to Aristotelian -two-state logic) to define states that do not have sharply defined boundaries.For a given input value, two or more states (heat is when ...) or interval states can be true to some extent.Fuzzy logic is a superset to conventional two-state logic, so two-state logic is a subset of fuzzy logic.The function set of fuzzy logic is extended to work with values between true and false, i.e., partial truth.Now let's introduce a process called fuzzy logic.It is a process that is used for recognition, classification and diagnosis.This logically inspires us in that we can easily apply fuzzy logic in the evaluation of disease diagnosis.For this purpose, we can divide fuzzy logic into several directions, see Figure 7: • Process Fuzzification: this process is described in such a way that, e.g., if changes in input values occur immediately without time delay, we understand this process at the level of a process value held on the set.• Knowledge base: It is part of fuzzy logic.Here knowledge base consists with given rules where its data is structured and unstructured.• Defuzzifier: it is the process of converting into crisp logic.

Figure 7. Fuzzy logic process diagram
Fuzzy Logic is expressed by creating fuzzy classes of some parameters.The rules and criteria are easy to understand.When it comes to data processing, these are represented using information in fuzzy logic.In other words, statistical models cannot detect the lack of missing values, on the other hand, these data values preserve a certain categorization [41].These reasons can be easily understood using machine learning (ML).ML is nowadays significant just in natural language processing, information mining, image detection and disease detection.Machine learning is a sub-form of artificial intelligence in which the machine itself learns and performs tasks through training.What is the function and application of ML in medicine?
• Inspires confidence in patients, • is applied especially in the early stages of disease, • ML algorithms are represented by decision trees, • the use of ML algorithms leads to rapid prediction of disease, even with high accuracy, • the initial learning phase starts with observation, • the algorithms used look for data patterns and predict better decisions on this basis, • disease prediction is the most important in this case, • different types of diseases can be predicted, • ML is described by an explanation process, based on the so-called step-by-step process, see Fig.In the last fifteen years, as advances in computer technology have made it possible to use the personal computer as a teaching tool, the field of practical application of models of physiological systems in medical education has developed, as well as the related technologies of creating educational simulators and multimedia educational applications using simulation games.
Effective creation of modern medical teaching programs with simulation games requires interdisciplinary cooperation of many professions: from system physiologists creating simulation models of physiological systems, educators responsible for the scenario of the teaching application, artists creating interactive animated images and designing the external graphic appearance of the whole application, to computer scientists and programmers integrating the whole work into the final form executable via the Internet on a computer.
Unlike the engineering sciences, in biology and medicine we do not often encounter mathematical representations of reality.It should be noted that the process of formalization, i.e. the conversion of a purely verbal description of the relevant network of relations into a description in the formalized language of mathematics, is delayed in the biological and medical sciences, compared to the technical sciences, physics or chemistry.If the process of formalization in physics began sometime in the seventeenth century, in the medical and biological sciences, because of the complexity and intricacy of biological systems, it comes only with cybernetics and computing.The methodological tools here are computer models created on the basis of mathematical descriptions of biological reality.
In addition to the above, personal computers and significant improvements in software commercialization suggest that simulation will gradually become a widely used method, both for technical and non-technical health analysts.In the healthcare industry, from a competitive perspective, simulation in particular may be a very special and valued tool for increasing efforts in continuous process improvement and redesign.It is the so-called common process problems and the entire expanding networks of healthcare facilities, including managed care networks, that have created opportunities for the reuse of simulation models.For example, healthcare providers generally suffer from unacceptable waiting times in waiting rooms.This is precisely the problem that could be solved by simulation.
Simulation is a process of imitation that involves modelling natural, human, and other systems.Since simulation mimics any logical system, in healthcare it can be the movement of patients through different clinical departments with different specialties.Further, it can be the process of the duration of each ongoing type of treatment, which ultimately can be a valuable tool for evaluating and comparing changes in treatment processes.Various measurements can be recorded during the simulation, which can then be subjected to statistical analysis.This aforementioned treatment simulation can be defined: a) directly in the clinic, where it is used primarily for the study of certain diseases, where the fundamental attitude to the matter is biological processes in the human body, b) directly in the operating theatre, where it is applied in order to capture, analyse and study medical procedures, including elective surgery, c) in the field of management, where it is applied as a tool for management and decision-making purposes, including strategic planning, d) in the pedagogical field, where it is applied in the teaching and learning process, in the application of virtual environments enriched with simulation experiments.The managerial and operational directions of the simulation are carried out simultaneously.Both of these directions form components for the control of selected processes in the healthcare sector.Such complex inter-departmental collaboration contributes to a complex intra-enterprise workflow and even to the creation of a complexly defined model.The duty of such a simulation model is to annotate this process using inter-organizational flows in a systematic and manageable way.
Models should be broad in their representation, but still communicative in terms of stakeholder understanding.An important advantage of this simulation in the process of modeling a healthcare facility is the ability to model even complex patient flows and to test "what if" scenarios.The success or failure of simulation studies depends on following a standard procedure on the healthcare platform.In order for the model to match reality, we must account for constraints in its design.
Mathematical modelling and simulation have been used for a long time, especially in laboratory measurements and medical biophysics.Among such AI tools are, for example, computational methods incorporating biomedical knowledge in relation to automating human reasoning processes.In the recent past, we have witnessed a resurgence in the modernization of artificial intelligence (AI) that is underscored by rapid advances due to the application of computing and the Internet.Another application in healthcare is the process of big data analytics and cloud computing [42]."Progress is impossible without change, and those who cannot change their minds cannot change anything," said Nobel laureate George Bernard Shaw [43].Artificial intelligence has made significant inroads in the field of medical biophysics such as cancer treatment, nanotechnology and drug delivery [44].Artificial intelligence is also the source of the way people think.Big data processing system is nowadays addressed by machine learning as a sub-form of AI, where the means of application are AI algorithms that are capable of handling large data set [45].Today, artificial intelligence uses the advances in big data cloud platform to do its work because of the huge amount of data collected for machine training [42].A promising idea is that of retaining the power of final decision making in the setting of AI as a digital assistant, where a competent researcher in a clinic uses AI to solve problems in medical biophysics even though it may be replaced by other means e.g.digital tools [46].We use simulation to predict disease based on a validated model.Some simulation methods, such as Monte Carlo [47,48], which is based on random selection, can easily predict the numerical outcome because other approaches to find a suitable outcome are difficult to find through traditional mathematical methods.Today, there is a wide range of medical biophysics applications that use modeling and AI resources.For example, let us mention the study of the COVID-19 pandemic, which is carried out by Monte Carlo simulations.

Conclusion. The future of artificial intelligence in healthcare
The application of artificial intelligence is advancing very fast nowadays and therefore has a future in healthcare.We can already say that it has its place in the diagnostic field, but also in the therapeutic field, because its manual expression is often difficult.At the same time, the imaging process within radiological images is scanned by artificial intelligence machines.Speech and text recognition including patient communication and clinical photography will continue to grow in popularity.However, it is important that AI in healthcare is available on an ongoing basis without limitations.Any AI programs must undergo an approval process and integrate with EHR systems and be standardized.In addition, there must be training for physicians, healthcare professionals in selected specialties and specialized programs that are directly related to AI practices.It is safe to conclude today that AI applications in clinical practice will become increasingly important over the next five years, with the irreplaceability of AI being a priority.

1 2 2 )𝑙𝑛0••
and for  < 0 resp.for  > 0 we get lim →∞ () = 0 resp.lim →∞ () = 1 (By choosing the activation function of the neurons of the input or output layer of the neural network, we can determine the method of transformation of the data fed to the network: • Sigmoid () = (1 +  −(− ) −1 where from ad 1) and ad 2) follows  = Gaussian curve: () =  −(−) 2 −  0,05 =  −6 2 follows  = − Ricker wavelet: ln() = − 2   () -different sensitivity bands correspond to this transformation, or its non-negative part.The parameters of these transformations have the following meaning: ϑ -mean value of data fed to the neuron from the training set σ -standard deviation of data fed to the neuron from the training set In addition to these activation functions, various modifications are used: Identity -linearity modified by shifting the centre of symmetry to the origin Hyperbolic tangent -extending the range of sigmoid values to the interval from -1 to +1 ReLU -composition of the constant (left of origin) with the identity (right of origin) Radial basis -Gaussian curve or Mexican hat 1.2.1.Network Learning.

Figure 8 .
Figure 8.The structure of machine learning