In the modern world, medicine plays an important role in the life of mankind. Thanks to advances in the field of medicine, you can achieve an unprecedented life expectancy and improve its quality. However, a sharp increase in the amount of medical information has led to the need for high-quality and fast processing, since people do not always cope with this work, the «human factor» does not allow processing information with 100% accuracy. A similar situation is observed in the daily work of a doctor: there is a medical error, the frequency of which the scientific community is trying to reduce. Artificial intelligence and machine learning systems should help in this issue. Another factor that makes artificial intelligence systems extremely promising is the relative cost-effectiveness and benefits of using these systems. By 2024, the market for artificial intelligence for medicine is projected to grow to $10 billion.

Also, the introduction of artificial intelligence systems will reduce costs in the key areas of development of the AI market for medicine. Investments in AI software platforms that provide tools, technologies, and services based on structured and unstructured information will be measured at $ 2.5 billion per year.

Currently, AI is defined as various software systems and the methods and algorithms used in them, the main feature of which is the solution of problems, like a person.

Artificial intelligence (AI) allows computers to learn from their own experience, adapt to set parameters, and perform tasks that were previously only humanly possible. In most AI implementations, from computer chess players to driverless cars, the ability to learn deep and process natural language is crucial. Thanks to these technologies, computers can be «taught» to perform certain tasks by processing a large amount of data and identifying patterns in them.

Expert systems are applied AI systems in which the knowledge base is a formalized empirical knowledge of highly qualified specialists (experts) in a narrow subject area. Expert systems are designed to replace experts when solving problems due to their insufficient number, insufficient efficiency in solving the problem or in dangerous (harmful) conditions. Usually expert systems are considered from the point of view of their application in two aspects: for what tasks they can be used and in what field of activity. These two aspects leave their mark on the architecture of the expert system being developed. The following main classes of tasks that can be solved by expert systems can be distinguished:

• diagnostics;
• prediction;
• identification;
• management;
• design (configuration);

The development of an ES is possible only if there are experts in the field, and the experts must agree in their assessment of the proposed solution; the problem must belong to a sufficiently structured area; the solution must not use much common sense (i.e., a wide range of General information about the world and how it works), but must be based on some knowledge in order to derive objective knowledge.

Human competence weakens with time, and a break in activity can affect professional qualities. The transfer of knowledge from one expert person to another is difficult, unlike the transfer of information between ES. This is a simple process of copying data from one system to another, without the need to re-lay the data and long-term training, as in humans. Also affects the so-called «human factor», because of which the decision-making by a person can be difficult, which can lead to critical situations, especially in medicine.

Neural networks are based on modeling processes that occur in the human brain. Artificial neurons are combined in networks, connecting the outputs of some neurons with the inputs of others. In simplified terms, a neural network is simply a program that receives data at the input and gives answers at the output. Being built from a very large number of simple elements, the neural network is able to solve extremely complex problems.

There are also more complex models in which the output of one network is directed to the input of another. These models create cascades of neural networks, so-called multi-layer neural networks.

Another interesting type of neural network is a neural network with feedback (RNN, recurrent neural network), when the output from the network layer is fed back to one of the inputs. Such platforms have a «memory effect», meaning that information is not lost when moving from one» neuron » to another, which makes such systems incredibly efficient. Such systems can be used to predict the behavior of a living object, which is actively used in medicine.

Deep (machine) learning is a set of algorithms based on neural networks that attempt to model high-level abstractions in data using architectures consisting of many nonlinear transformations. this is a section of artificial intelligence theory that focuses on finding methods for solving problems by learning to solve similar problems. To build such methods, we use the tools of algebra, mathematical statistics, discrete mathematics, optimization theory, numerical methods, and other branches of mathematics.

The main objective of machine learning in medicine is to reduce the burden on medical personnel due to the need to decrypt analysis data, computed tomography (CT) and magnetic resonance imaging (MRI).