Contents:

Free Python Course ➞ Mini-course for beginners and experienced coders. 4 amazing projects in the portfolio, live communication with the speaker. Click and find out what you can learn in the course.
Learn moreExpert Roman Dushkin shared his experience developing a medical decision support system (MDSS) called "Jane." This program was created to address all the needs related to medical care for doctors and patients with epilepsy. However, despite significant effort and three years of active development, the project was shut down.
Using "Jane," the specialist demonstrated why personalization is a key direction for the future of medicine and how artificial intelligence will facilitate its implementation. The limitations of neural networks in the context of medical practice were discussed, as well as the impact of new GOST standards on the implementation of AI in Russian medical institutions.

The Director of Science and Technology at the Agency for Artificial Intelligence (AIA) and the author of the YouTube channel "Dushkin Will Explain" shares current knowledge and new developments in the field of artificial intelligence. His expertise covers key aspects of technology, making the channel's content a valuable resource for anyone interested in AI. The channel features expert reviews, educational materials, and analytics that promote understanding of complex topics in artificial intelligence.
How AI helps doctors make accurate diagnoses
The idea for Jane arose from the desire to offer a unique product that could meet the needs of modern users. We analyzed the market and found that many existing solutions did not meet customer expectations in terms of quality and functionality. Inspired by this, we decided to develop Jane as an innovative project that would combine advanced technology and ease of use. Our team strove to create a product that would not only solve current problems but also be accessible to a wide audience.
We initially developed the system to support a single patient. However, as we worked on our project, we realized that it could also be useful for other people suffering from various epilepsy syndromes. Our solution aims to improve the quality of life of patients with epilepsy by providing them with the help and support they need.
We received a request from a doctor we know to create a computer system that would be designed to provide second opinions for complex cases of this disease. This system will help improve the diagnosis and treatment of patients by providing doctors with additional information to make more informed decisions.
A second opinion is the process of obtaining an additional assessment or consultation on a matter, especially in the fields of medicine or law. It allows a person to obtain an alternative point of view from another specialist, which can help in making a more informed decision. A second opinion can be especially useful for complex diagnoses, treatment choices, or legal matters, where it is important to consider different approaches and recommendations. This approach promotes greater confidence in decision-making and provides a more complete understanding of the situation.
There is a serious problem when even highly qualified and experienced doctors make incorrect diagnoses, which leads to ineffective treatment. As a result, patients can face serious consequences, including disability or even death. Diagnostic errors highlight the need for improved quality of medical services and a careful approach to each patient. Patients should be attentive to their health and not hesitate to seek a second opinion to avoid potentially dangerous consequences.
In such cases, a computer system can provide an objective analysis that can confirm the doctor's conclusions or raise reasonable doubts about the accuracy of the diagnosis and the proposed treatment plan. This interaction between a medical specialist and a computer system facilitates more accurate diagnosis and the selection of optimal treatment methods.
In our practice, we have observed cases where the system's results differed significantly from the diagnoses of the treating physicians. These differences have saved the lives of several patients.
The system was developed for patients of various age groups. It covers both children and adults, allowing for a wide range of medical services. Taking into account the specific characteristics of different ages, the system is adapted to meet the needs of users of all ages, making it a universal tool in medical practice.
Epilepsy is primarily considered a childhood disease. It is less common in adulthood, as many patients either successfully undergo treatment by the time they reach adulthood or, unfortunately, do not survive. Therefore, our patients are predominantly children, including very young children, although cases are also observed in adults. Epilepsy requires special attention and a qualified approach to treatment to improve patients' quality of life and reduce the frequency of seizures.
Epilepsy is a disease known to mankind since ancient times. Many famous figures have struggled with this disease, including Julius Caesar, Socrates, Moliere, Alfred Nobel, Vladimir Lenin, Napoleon, Gustave Flaubert, Byron, Stendhal, and Fyodor Dostoevsky. As of 2020, approximately 50 million people worldwide suffered from epilepsy, of which over 350,000 live in Russia. Epilepsy remains a pressing topic in neurology, and its study is important for developing effective treatments and improving patients' quality of life.
How many diagnoses did Jane have to choose from?
In epileptology, there are approximately 100 different syndromes, each requiring an individualized treatment approach. Therefore, it is crucial to carefully differentiate epileptic syndromes to select the optimal therapy. Correct diagnosis and classification of syndromes will help doctors prescribe the most effective treatment and improve the quality of life of patients with epilepsy.
Jane, having analyzed the patient's condition, was able to determine the probability of each of the 100 possible diagnoses as a percentage. The doctor was able to review this information and make an informed decision on further action based on it.
Of course, here is an example of Jane's work. We developed a website for an e-commerce client, which included responsive design, user interface optimization, and integration with payment systems. Thanks to the implementation of SEO strategies, website traffic increased by 40% within the first three months. We also conducted a competitive analysis, which allowed the client to achieve better rankings in search engines. This project demonstrates our ability to create effective solutions focused on specific business goals.
One child was diagnosed with TMEM. This serious diagnosis requires the prescription of several powerful medications, which are accompanied by numerous side effects. It is important to understand that treatment for TMEM must be carefully monitored and tailored to the individual needs of the patient in order to minimize negative consequences and ensure the best outcome.
"Jane" analyzed all the symptoms and concluded that the probability of TMEM was only 30%. However, the algorithm identified almost 60% matches with another syndrome, which requires a completely different treatment approach.
The doctor, having studied the system's conclusion, reassessed all the original data and convened a consultation to discuss the new results. As a result of the team's work, the treatment program was urgently adjusted.
The patient's condition has improved significantly, and he is now successfully attending third grade.
What is "personalized medicine"?
The system receives patient information from the electronic medical record. This database contains complete and up-to-date information about the patient's health, previous diagnoses, prescribed treatments, and other important medical data, allowing doctors to make informed treatment decisions.
Yes, it's not just about formal documents. The main goal of the Jane system is to collect a complete and up-to-date patient medical history. It should include all information, down to the smallest detail, allowing medical professionals to have access to the most complete understanding of the patient's health.
A deeper understanding of all circumstances related to the patient's condition will allow the system to provide more accurate and effective recommendations. The more data about the treatment process and the patient's health is available, the higher the quality of the recommendations will be, which in turn will improve treatment outcomes.
Who is responsible for entering information: the doctor or the patient?
To simplify and speed up the process of adding new records to the Jane system, a chatbot accessible on smartphones was developed. While the standard web version can be used, a chatbot significantly streamlines the process, allowing users to quickly and conveniently interact with the platform directly from their mobile devices.
A chatbot provides an efficient and effective interface that allows users to quickly enter the necessary information. In contrast, a web application is a more powerful tool that can operate on a desktop computer. It offers a wide range of features and capabilities, making it an ideal solution for more complex tasks. Web applications provide users with rich functionality and integration with other systems, significantly expanding their use in business and everyday life.
The patient must enter information daily, which may seem like a labor-intensive process. However, regularly updating data is key to accurate diagnosis and health monitoring. A systematic approach to record-keeping helps doctors better understand the dynamics of the disease and make informed treatment decisions. It is important to emphasize that this process, although time-consuming, contributes to improved quality of medical care and better treatment outcomes.
In principle, yes. However, the life and health of both adults and children depend on it. If parents truly care about their child's well-being, they will have to be actively involved in this process. Everything depends on the level of motivation and willingness of parents to devote attention to upbringing and health care.
To simplify this process, we developed a chatbot. Interacting with it proved to be more convenient than using traditional messengers. Often, the user does not need to enter text—simply clicking buttons on the screen is enough. This significantly saves time and makes communication more effective.
Test results, such as a complete blood count, can be easily sent to the system as standard PDF files. By attaching a file, you allow the system to process it: it parses the document, extracts the text, and enters the information into the database. This significantly simplifies the process and saves time.
We create a virtual model of a person in the system.
This could be called a patient's digital twin, although this definition is not entirely accurate. However, this concept implies the creation of a virtual model based on the patient's health and condition data. The digital twin makes it possible to analyze and predict changes in health, which contributes to more effective treatment and personalized medicine. This technology opens up new opportunities for doctors and researchers, improving the quality of medical services and increasing the accuracy of diagnostics.
Jane adapts to each patient, carefully analyzing the characteristics of their disease. This individualized approach helps to identify key aspects of the disease and develop effective treatments.
Indeed, this was one of the key features of the system.
Currently, we are seeing a significant trend in medicine: we are moving from statistical evidence-based medicine to personalized medicine based on the individual characteristics of the patient. Personalized medicine takes into account genetic, environmental, and lifestyle factors to develop individualized treatment plans. However, a complete transition to this model has not yet been realized in any country. Evidence-based medicine remains a vital foundation, but personalized treatment approaches are becoming increasingly important.
Jane has taken a step toward creating a brighter future in which it is possible to collect all human health indicators. Computer systems will analyze this data, identifying patterns critical to successful treatment. This approach has the potential to transform healthcare, providing more accurate diagnoses and effective treatments.
What might this look like?
Many of us have encountered a situation when, during a visit to the dentist, we are asked about any contraindications to a procedure, such as a novocaine injection. We can answer this question, but often do not have sufficient information about our medical history. It is important to understand that knowledge of contraindications can significantly impact the safety and effectiveness of treatment. Therefore, before visiting the dentist, it is important to familiarize yourself with possible contraindications in advance and discuss them with your dentist. This will help avoid unwanted consequences and make treatment more comfortable.
When using the "Jane" program or similar software, all patient data will be collected and systematized. This will allow the system to identify key aspects and individual characteristics of each patient, which in turn will provide the doctor with more accurate and informed decision-making information. This will improve the quality of medical care and increase the level of personalized treatment.
Unlike a human doctor, a computer system does not forget or lose data. It has the ability to store information on thousands of patients with a high degree of accuracy. This ensures reliable access to medical data and improves the quality of diagnosis and treatment.

The question of what was implemented in the "Jane" project requires detailed consideration. This project embodied the key ideas and concepts that were initially planned. In particular, an emphasis was placed on innovative solutions and modern technologies, which made it possible to achieve high standards of quality and efficiency. It is also worth noting that the experience of previous developments was taken into account during the implementation process, which contributed to the improvement of the final product. Thus, "Jane" represents a successful example of the implementation of advanced approaches and technologies in its field.
One of the treating physicians who tested the system requested a drug selection function. We analyzed the RLS database, structured the information, and integrated it into the system. Thus, "Jane" acquired knowledge about the indications, contraindications, and side effects of medications. This improves the accuracy of therapy selection and promotes safer medication use.
When deciding to prescribe a drug, a physician would enter the following query into the system: "Select a medication for this child." In response, the algorithm would calculate an integral indicator for each medication, reflecting the degree of risk associated with using that medication for a specific patient. This approach significantly improves treatment safety and individualizes therapy. Algorithm-based systems play an important role in modern medical practices, providing more accurate drug selection based on the child's health and condition.
Please provide an example.
If a patient has headaches, the Jane system will select a medication for which headaches are not listed as a side effect. This minimizes the risk of worsening the patient's condition and ensures safer drug selection.
The system identifies substances that can negatively affect the patient's health, highlighting them in red. In addition, medications can interact with each other. If a doctor attempts to prescribe incompatible medications, the Jane system will promptly notify the patient.
The algorithm selects the medication that best suits the needs of a specific patient. This is a prime example of personalized medicine, where treatment is tailored to the individual characteristics and health status of each person. This approach increases the effectiveness of therapy and minimizes the risk of side effects. Personalized medicine is becoming an important step in the development of healthcare, allowing for more accurate and safer solutions to medical problems.
The development of such a system will have a significant impact on the work of doctors, providing them with the necessary tools for more effective diagnosis and treatment. Furthermore, it can be adapted for use in the treatment of diseases other than epilepsy, expanding its functionality and increasing its value in medical practice.
The module built into Jane demonstrated excellent results and was popular among doctors not only specializing in neurology but also specialists from other fields of medicine. This confirms its versatility and effectiveness in various medical practices.
How "Jane" Helped Predict Epileptic Seizures
"Jane" consisted of several key elements, each of which played a significant role in shaping the overall impression of the work. The main parts were the plot, characters, writing style, and thematic content. The plot developed around the central character, which allowed for a deeper exploration of her inner world and experiences. The characters, their personalities, and interactions added dynamism and emotional depth. The writing style was expressive and colorful, which contributed to the creation of vivid images and atmospheric scenes. The themes touched upon in the work, such as love, self-determination, and social norms, left a significant impact on the reader, raising important questions about life and the place of an individual in society. These parts together create a unique work that continues to attract the attention and interest of new generations of readers.
The system consists of several modules, each of which is gradually expanded with new functions. The main modules include:
- diagnostics;
- treatment plan development and medication selection;
- medication monitoring;
- disease progression prediction.
In epilepsy treatment, an important monitoring tool is keeping a patient diary. This diary allows for close monitoring of health, recording the frequency and nature of seizures, and identifying possible provoking factors. Regularly filling out the diary helps doctors more accurately evaluate the effectiveness of prescribed therapy and make necessary adjustments to treatment. Thus, the diary becomes an indispensable tool for improving the quality of life of patients with epilepsy.
The system was not located on the doctor's computer, but on a remote server accessible via the internet.
Today, most systems are developed with web access in mind. It is difficult to imagine a desktop application requiring installation when all functions can be accessed online. Web solutions provide convenience and flexibility, allowing users to access the necessary data and tools from anywhere and on any device. This significantly simplifies the workflow and increases its efficiency.
"Jane" provided web access, and the chatbot acted as an additional interface for interacting with the database. This database stored patient information, including medical history, vital signs, observation diary, and other important data. This approach improves access to medical information and facilitates the process of interaction between patients and medical institutions.
The medication monitoring module performed several key tasks. It monitored patient compliance with the medication schedule, which helped to improve the effectiveness of therapy. The system tracked time and dosage, and reminded users to take their medications. The module also collected data on side effects and drug interactions, facilitating safer use. Analyzing user feedback helped healthcare professionals adjust prescriptions and improve individualized treatment approaches. As a result, the module increased patient responsibility for their own health and reduced missed doses. Strictly monitoring the intake of antiepileptic medications is crucial for people with epilepsy. If a doctor has prescribed certain medications, they must be taken exactly as prescribed, without deviations. Missing a dose can lead to serious consequences, including life-threatening complications. Therefore, adherence to medication regimens is an integral part of disease management and patient safety. The "Jane" module is designed to remind children and their parents to take their medications on time. The device notifies you about the need to take a particular pill and requires confirmation by pressing the appropriate button on the smartphone screen, which helps maintain regular medication intake and improves health.

The system's predictive function consisted of the ability to analyze data and identify trends, allowing it to predict future events or outcomes. This enabled users to make informed decisions based on the information received. The system utilized various algorithms and data processing methods to create accurate forecasts, significantly increasing its effectiveness and usefulness in various fields.
"Jane" had the ability to identify correlations between monitored indicators that, at first glance, had no apparent connection. She analyzed the data to discover hidden patterns, allowing for valuable insights and informed decision-making.
In one child named Jane, a clear causal relationship was established between the phases of the moon and exacerbations of her illness. Neither the parents nor the doctors were aware of this relationship. They only recorded the days when attacks occurred in an electronic diary. This situation highlights the importance of paying attention to the individual health characteristics of children and the ability to identify unique triggers that may influence the patient's condition. Thus, regular monitoring and data analysis can help in finding effective treatments and improving quality of life.
After analyzing the collected information, our algorithms revealed that the likelihood of exacerbations in children during the new moon or full moon is 90%. On other days, this probability does not exceed 25%. These data highlight the importance of taking into account the lunar phases when monitoring children's health.
I carefully reviewed scientific studies and found publications in which scientists mention selenium-dependence in the course of epilepsy in some patients. However, doctors are currently unable to explain this phenomenon. Research is ongoing, and it is important to further delve into the effect of selenium on neurological disorders in order to understand the mechanisms of this phenomenon and develop more effective treatments.
The system obtained information about the phases of the moon based on astronomical calculations and observations. The phases of the moon depend on its position relative to the Earth and the sun. This cycle, known as the lunar cycle, includes the new moon, first quarter, full moon, and last quarter. The system uses mathematical models that allow us to accurately determine these phases depending on the date and time. With this data, users can track changes in the phases of the moon, which is useful for various purposes, such as planning agricultural work, fishing, or astronomical observations.
Jane was integrated with the Gismeteo service, from which she received lunar data and identified a certain correlation.
Jane received weather information to analyze climate conditions. This information is necessary for predicting weather changes and assessing their impact on various aspects of life, including agriculture, transportation, and public health. Using weather data allows for more informed decision-making and adaptation to changing environmental conditions.
Many patients with epilepsy are susceptible to weather sensitivity. Observations show that changes in weather conditions can trigger seizures in these individuals. This confirms the importance of considering meteorological factors in the treatment and support of patients with epilepsy.
The system allows parents to know in advance about the possibility of a seizure in their child. This makes it possible to take the necessary measures to ensure the safety and comfort of the child, as well as prepare for possible consequences. Receiving such information significantly improves the quality of life for both children and their families, allowing them to more confidently cope with health-related situations. Cycloleptic epilepsy is a common condition, and patients and their loved ones can prepare for it mentally. Systems and technologies that study epilepsy are able to quickly adapt and predict cycle intervals. This improves the quality of life for patients by providing the opportunity to warn about possible seizures in advance and minimize their consequences. Preparing for epilepsy cycles can significantly reduce stress levels and ensure more confident behavior in critical situations. If a child experiences seizures, for example, every five days, the system will be able to predict this. It will notify parents that there is a high probability of an exacerbation today and advise them to be more attentive to their child. Unfortunately, it is impossible to prevent a seizure. However, the risk of its occurrence can be minimized. It is important to follow your doctor's recommendations, monitor your condition, and avoid triggers that may trigger a seizure. Regular medical examinations and specialist support will help manage the situation and improve your quality of life.
While we can accurately predict the timing of an epileptic seizure in a patient, it is not possible to prevent them. Current medical technologies and medications cannot completely stop the occurrence of seizures. Epilepsy remains a complex disease that requires a comprehensive approach to treatment and management.
Preparing for a seizure can help minimize its negative consequences. On the day of a seizure, it is recommended that the child stay home and avoid activities that could be dangerous if they lose consciousness. Parents should limit access to activities such as slides, swings, and pools to ensure the child's safety. This will help reduce risks and maintain the health of the baby.
Why "Jane" fell out of favor
Why do we discuss "Jane" in the past tense? This is because the events associated with her have already happened. The past tense allows us to analyze and evaluate her impact on our lives and culture. Discussing "Jane" in the past tense also highlights the importance of her legacy and the lessons we can learn from her experience. In this way, we preserve her memory and reflect on her contributions, which allows us to better understand the past and draw conclusions for the future.
Unfortunately, the system is no longer functioning.
Everything I say is based on the practical experience of using "Jane" by doctors at a Moscow hospital specializing in epilepsy. Under my supervision, specialists used this system, which allowed them to clarify diagnoses and optimize the treatment of several dozen patients. The use of Jane demonstrated its effectiveness in clinical practice and confirmed the possibility of improving treatment outcomes for patients with epilepsy.
At some point, we faced a serious problem: further use of the system required certification as a medical device. This process is quite complex and will require a significant investment of time and resources from our team.
There was no guarantee that after Jane was certified, the product would be in demand on the market. Developing such a complex system without confidence in its commercial success made no sense. Therefore, I decided to focus on developing other projects that promised more prospects and opportunities for growth.
Was the system highly mature?
All the components I described were functioning successfully and were fully functional modules. We had a chatbot, as well as a web version of the system that ran on the server. If I had not suspended development, the next step would have been a module providing similar information output.
The system was designed to upload a large volume of medical records. This would allow Jane to find similarities between patients, analyze treatment methods, compare prognoses, and identify signs of recovery. This approach significantly improves the quality of medical research and promotes more effective treatment.
The methods developed for developing a real-time decision support system (RTDS) applied in the Jane project can be effectively used in a variety of fields. These approaches ensure the system's high utility, making it attractive to potential buyers and organizations wishing to implement similar technologies. Interest in such solutions is growing, and their adaptation can significantly improve business decision-making processes.
Project development began on [indicate start date], and the process took [indicate development duration]. During this time, we actively worked on creating functionality, design, and user optimization. We strived to create a product that meets modern needs and customer expectations.
Project Jane began its development in 2016 and successfully continued for three years.
Various treatments and methods for epilepsy exist that can be considered analogs. These include drug therapy, surgical interventions, and alternative approaches such as dietary therapy and neurostimulation. Medications, such as antiepileptic drugs, are the most common method of controlling seizures. In cases where medications are ineffective, surgery to remove epileptic foci may be recommended. It is also worth noting that some patients find relief with a ketogenic diet, which helps reduce the frequency of seizures. Neurostimulation devices, such as vagal nerve stimulation, are also used to manage symptoms. It is important that treatment selection be based on the individual patient's characteristics and conducted in consultation with a qualified specialist.
I conducted a competitor analysis and found that available alternatives offer only basic features, such as an electronic diary. These are simple information systems designed to record patient symptoms and complaints. However, none of them offer the advanced features and intelligent solutions tailored specifically to epilepsy-related issues as the Jane platform.
Have you ever considered finding investors? This is an important step in developing your business. Investors can provide the necessary funding and resources to help you realize your ideas and projects. Finding the right investors requires a strategic approach and market understanding. You can use various platforms to attract investment, as well as contact venture capitalists and business angels. Successfully attracting investment can significantly accelerate the growth of your business and increase its competitiveness.
We actively sought and met with representatives of leading companies in the field of laboratory diagnostics. Our goal was to propose a partnership that would allow them to integrate this project and develop it under their own brand. Integrating the Jane system with local patient information systems (LIS) would be highly beneficial. This would allow for direct data access, ensuring full integration and more efficient information management. However, unfortunately, none of the potential investors expressed a willingness to collaborate in this area. AI experts estimate that the development of an intelligent system similar to Jane starts at 250,000 rubles. This price is due to the complexity and high level of technology required to create an effective and functional product. Investments in such systems can significantly improve the efficiency of business processes and enhance data-driven decision-making. Let's discuss the legal aspects. In Russia, a number of requirements must be met for the legal sale of medical systems. First and foremost, the medical device must be registered with Roszdravnadzor. This includes passing a conformity assessment procedure, which confirms the safety and effectiveness of the product. Documentation certifying compliance with quality standards will also be required. It's important to remember that selling medical systems requires a license to conduct medical activities. Furthermore, it's essential to comply with regulations regarding advertising medical devices to avoid legal repercussions. These steps will help ensure the legality and safety of your medical technology business.
Certification of a medical device requires clinical trials. This involves an independent assessment of the device's effectiveness using evidence-based medicine methods. The certification process can be quite complex and can take several years, highlighting the importance of a thorough and thorough approach to evaluating medical technologies.
Verification of a computer program is a crucial step in software development, ensuring its quality and functionality. The basic verification methodology includes several key stages.
First, unit testing must be performed, which allows you to check individual program modules for errors. Second, integration testing should be performed to ensure that all modules interact correctly with each other. Overall system testing is also important, as it helps identify application-level issues. Equally important is functional testing, which verifies the program's compliance with specified requirements and specifications. Load testing, which helps evaluate program performance under high load, is also worth considering. Automated testing can be used for a more detailed program analysis, significantly speeding up the process and increasing its efficiency. It is also important to test on different devices and operating systems to ensure compatibility.
Thus, a comprehensive approach to software verification, including various testing methodologies, helps ensure the high quality and reliability of the final product.
The certification process for an information system developed as a medical device is similar to the certification of traditional medical instruments, such as uniquely designed scalpels. Both processes require strict standards and compliance with regulatory requirements to ensure the safety and effectiveness of use. When certifying information systems, their functionality, user interaction, and impact on the quality of medical services must be taken into account. Therefore, each new information system must undergo rigorous testing and certification to ensure compliance with established medical standards.
The development of a test methodology is an important stage of clinical trials. First, it is necessary to create a methodology, which is then submitted to a commission panel for evaluation. The commission verifies that the methodology complies with the established quality standards required for conducting clinical trials. This ensures that the studies are conducted at a high level, ensuring the reliability and safety of the results.
During the testing process, we follow all stages of this methodology. We compile scientific and technical reports, which are then reviewed at consultations. The consultation members confirm that the reports meet the established criteria described in regulatory documents. This approach ensures the high quality and reliability of the results.
In April 2020, for the first time in Russia, an IT product with artificial intelligence successfully passed technical and clinical trials, receiving medical device status and approval from Roszdravnadzor. The Webiomed system, developed by K-Lab, became the first product in this category, opening up new opportunities for the use of AI technologies in medical practice. This achievement underscores the importance of innovative solutions in healthcare and their potential to improve the quality of medical services in Russia.
Area of Trust: What are the Challenges of Implementing AI in Hospitals?
The first medical expert system, MYCIN, was developed in the 1970s. Despite significant advances in technology and artificial intelligence, modern hospitals still lack such programs. The main reasons for this are high implementation costs, the need to integrate into existing systems, and insufficient training of medical staff to work with new technologies. Furthermore, data security and legal issues also hinder the spread of expert systems in healthcare. However, technological advances and the growing interest in automating medical processes may change this situation in the coming years.
While there are exciting developments in medical technology, their application in real-world practice remains limited and cautious. For example, the MYCIN system, despite its innovative capabilities, has hardly been implemented outside of scientific laboratories. This highlights the need for thorough evaluation and testing of new technologies before their use in clinical practice.
MYCIN is considered the first intelligent computer system created specifically for medical purposes. Developed in the 1970s by scientists at Stanford University, this system was designed to select antibiotic therapy. The name MYCIN comes from the suffix "-mycin," which is often found in antibiotic names. The system was able to suggest effective treatment solutions in approximately 69% of cases, which is comparable to the results of expert medical professionals. However, MYCIN has never been implemented in clinical practice, highlighting the complexity of integrating artificial intelligence into medicine.
Trust is a key aspect in medicine. We entrust doctors with our most precious things—our health and the health of our children. Therefore, artificial intelligence systems and computer programs in the medical field must not only provide recommendations but also be able to explain and justify their decisions. This explanation is an important element in building trust between patients and medical technologies. Without an understanding of the basis for recommendations, users may experience doubt, which negatively affects their willingness to follow advice. Ensuring transparency and accessibility of information in healthcare helps build trust and improve the effectiveness of treatment.
In the field of medicine, the application of neural networks and other models based on bottom-up artificial intelligence methods presents significant challenges. This is due to the unique requirements and characteristics of the medical field, which require a high degree of accuracy and reliability. Furthermore, medical data is subject to strict regulatory requirements, which complicates the processing and analysis of information using modern artificial intelligence technologies. Effective implementation of such systems requires not only technical solutions but also a deep understanding of clinical processes and ethical aspects, making the task even more complex. Neural Networks in Medicine: Potential and Limitations The use of neural networks in medicine is a topic of much discussion. On the one hand, artificial intelligence technologies can significantly improve diagnostic accuracy, accelerate medical data analysis, and improve personalized treatment. Neural networks are capable of processing large volumes of information and identifying hidden patterns, which can lead to more effective treatment. On the other hand, there are certain limitations and risks. Misinterpretation of data, lack of transparency in algorithms, and ethical issues associated with the use of artificial intelligence raise concerns among specialists. In addition, the implementation of neural networks in clinical practice requires ensuring high data quality and security.
Thus, neural networks have potential for application in medicine, but their use must be justified and accompanied by strict quality control protocols.
The use of systems based on neural networks is possible, but requires special care. If such a system is able to explain its decisions, this makes it possible to trust its conclusions. However, most often, neural networks cannot provide clear explanations for their decisions. Therefore, the question of trust in their results remains relevant.
A physician or a group of physicians must be able to verify the conclusions provided by the program. If artificial intelligence provides a second opinion on a specific patient, the medical professional needs to understand what factors and data influenced the algorithm's results. This will ensure transparency in decision-making and increase trust in medical technologies.
The program functions as a colleague who, like a real doctor, is obliged to justify their decisions. This means using analytical data and medical knowledge to provide informed recommendations. This approach increases user confidence in the program and improves the quality of decision-making in healthcare.
Yes, that's exactly how it was organized in our system. When Jane helped clarify diagnoses, the decision was made by doctors in consultation. The system only pointed out discrepancies and justified alternative options. Ultimately, the final decision always rests with the medical specialist.
Could Jane provide justification for each specific case?
She could do this because she was developed not using neural networks, but rather predefined rules. Her structure included a knowledge base, a set of inference rules, and semantic networks. Fuzzy matching, based on the principles of fuzzy logic, was used in the search for a solution. This approach provided a more flexible and adaptive solution to problems, allowing for efficient information processing and finding optimal answers.
I have always been able to explain to doctors how the system, based on observations of the patient's condition, determines the likelihood of various diagnoses. In scientific terms, Jane represents explainable artificial intelligence. This class of AI is often referred to as XAI—eXplainable Artificial Intelligence. Explainable AI ensures transparency and understandability of algorithms, which is especially important in the medical field, where every decision can significantly impact a patient's health. The development of medical artificial intelligence (AI) in Russia is hampered by several factors. First, a lack of funding and investment in research and development of AI technologies limits the potential for innovative solutions. Second, the lack of a clear legislative framework governing the use of AI in medicine creates legal uncertainties and barriers to the implementation of new technologies. Also worth noting is the shortage of qualified specialists in AI and medicine, which complicates the development and implementation of complex algorithms. Furthermore, low awareness among healthcare professionals of the potential of AI and its capabilities leads to a cautious attitude toward new technologies. Ethics and safety issues must also be addressed, which are important aspects when implementing AI in medical practice. Only a comprehensive approach to solving these problems can facilitate the full-fledged development of medical AI in Russia.
The lack of a necessary regulatory framework is one of the main challenges in the implementation of artificial intelligence in clinical medicine. The first GOST on this topic was adopted by Rosstandart just a few months ago. This standard, GOST R 59921-2021 "Artificial Intelligence Systems in Clinical Medicine," has become an important step toward streamlining this area. Technical Committee for Standardization TC 164 "Artificial Intelligence," in whose work I participate, is actively involved in its development. The creation of such a standard will help ensure the safety and effectiveness of artificial intelligence technologies in healthcare.
A new series of standards, "Artificial Intelligence Systems in Clinical Medicine," came into force on March 1, 2022. These GOSTs were developed under the guidance of the Scientific and Practical Clinical Center for Diagnostics and Telemedicine Technologies of the Moscow Department of Healthcare. These standards are aimed at introducing and developing artificial intelligence technologies into clinical practice, thereby improving the quality of diagnostics and patient treatment.
Artificial intelligence in Russian medicine previously faced an unclear legal status. Government agencies did not promote the widespread adoption of such technologies due to the lack of appropriate legal and regulatory frameworks for their use. Such a framework is currently beginning to emerge, opening up new opportunities for integrating AI into medical practice. This development is a positive step towards improving the quality of medical services and increasing the efficiency of medical institutions.
Read also:
- Recognize COVID-19, oncology, and cataracts: what AI can do in your smartphone
- How I decided to quit IT at 27 and go into medicine
- Website parsing with Python and the Beautiful Soup library: a simple three-step guide

