Artificial intelligence is no longer a topic of the futureo. It is increasingly penetrating medicine - from diagnostics and analysis of medical data to personalizing treatment and organizing healthcare services.

What does this mean for doctors, patients, and medicine itself? Will AI be a tool that expands the capabilities of medical professionals, or will it fundamentally change the way we understand the medical profession?

Career Daily News reached out to leading professors, doctors, and experts from various fields of medicine with one key question:

“How do you think medicine will look in the next 2-3 years in the era of increasingly developing artificial intelligence?”

The responses show different perspectives on the future of healthcare - from the enormous potential of technology to the risks associated with its use.

One thing, however, seems increasingly clear: AI will not simply add another tool to medicine. It will change the way doctors work, make decisions, and interact with patients.

The question is how we will find the right balance between technological capabilities and human expertise.

But let's give the floor to our local luminaries in medicine, what do they think?

Acad. Prof. Ivan Milanov, MD, PhD, DSc — Director, Medical University Hospital, Sofia, Bulgaria

"Artificial intelligence will assist doctors by summarizing the available information, but it will not soon replace clinical intuition, which is very important in the diagnostic process."

Assoc. Prof. Nikolay Gabrovski, Head of the Neurosurgery Clinic and Deputy Director for Medical Activities at UMHATSM “N. I. Pirogov”, Chairman of the “Council for Brain Health” Foundation

The introduction of large language models into our daily lives can be defined as revolutionary and nothing less. If the internet and digitization made vast amounts of information accessible to all of us, so-called artificial intelligence (AI) has provided an impressive opportunity to quickly extract information from these large datasets, synthesize and analyze it. It is undeniable that this will change every aspect of our lives and will inevitably affect medicine as well.

General expectations are that this will happen as early as tomorrow, but in my opinion, in healthcare, AI as a tool will enter more evolutionarily than revolutionarily. The reason for this is that medicine cannot tolerate even the smallest analytical errors, and refining reliable tools will take time. Currently, the focus is primarily on training models with high-quality and reliable information, and this is starting to bear fruit.

Although not immediately, daily life in medicine may be changed by AI through the introduction of decision support systems. There are high expectations in the field of imaging diagnostics. AI is already performing a significant part of the calculations in complex spinal (as well as various other types of) surgery. I hope that one constant burden accompanying the medical profession - extensive documentation and paperwork - will be significantly altered by AI, which will synthesize what has been done during the treatment process, integrate voice notes, and generate a considerable part of the documents we need. AI will also be very useful in planning patient movement and navigating them through the treatment process, especially when it comes to complex, multi-step diagnostic and therapeutic procedures. I believe these will be the first steps.

At a more fundamental level, AI is expected to change the search for new drug molecules and, most importantly, to change the way their capabilities are tested. It is likely that entirely new classes of medications will be discovered or significantly higher efficacy will be achieved by modifying already existing drugs.

In the medium term, I see enormous progress in connecting our brain and nervous system with external devices and machines (brain-machine interface). This will provide incredible opportunities for patients with impaired motor, visual, and other functions on one hand, and for direct, easy communication between us and the surrounding technologies, machines, and devices on the other.

Recently, there has been much talk that thanks to the exponential development of AI, we will soon be able to treat every disease and even stop the aging of cells. This may indeed be possible one day, but I am convinced that it will take us not years, but decades. Our body is an incredibly complex biological machine, with processes that have evolved over millions of years, and the idea of controlling it with a few commands from a computer keyboard is naive. But there is no doubt - we are embarking on a very interesting, long, and largely unpredictable path in our development as individuals and as a society.

Implementing AI without a strategy is like buying a plane without a pilot

Prof. Dr. Kostadin Angelov, MD (Chairman of the Health Care Committee of the National Assembly)

“In my opinion, in the next 2-3 years, AI will not replace the doctor. It will start to replace the doctor who refuses to work with AI. This is the essential difference. In medicine, we are entering a period where the speed of technological development outpaces the ability of systems to adapt to it. AI can already analyze vast amounts of medical information, imaging studies, laboratory results, and scientific publications in a time that a human cannot achieve. The next step is for this information to start actually changing clinical decisions. But I would not entrust medicine to an algorithm. Because medicine is not just mathematics. It is judgment, experience, intuition, communication with the patient, and above all, bearing responsibility. Therefore, the future, in my opinion, will not be “doctor versus AI,” but doctor + AI versus doctor without AI. However, there is something we talk about much less - security. When medicine becomes increasingly dependent on AI, data, and connected systems, the hospital will no longer be just a healthcare institution. It will be part of the critical infrastructure of the state. A cyberattack on a hospital could halt surgeries, diagnostic equipment, electronic records, medications, and communications. At some point, this is no longer an IT incident. It is a risk to human life and potentially a matter of national security. Therefore, the future of AI in medicine must also be a future of AI security. We will need to know not only whether an algorithm is accurate but also who has access to it, what data it was trained on, whether it can be manipulated, what happens when it gives an incorrect result, and who is responsible for the consequences. And here, in my opinion, will be the greatest challenge. Because the most dangerous scenario is not that AI makes mistakes. The most dangerous scenario is that a person stops checking because they trust the machine. In the coming years, we will also see another significant change - medicine will gradually shift from treating an already existing disease to predicting and preventing risk. AI will allow us to see patterns that a human can hardly detect on their own. This could change healthcare much more than automation. But in the end, I believe in one thing: AI will not make the doctor redundant. It will make the good doctor even stronger. And the best doctors of the future will not be those who know everything. They will be those who know what to ask the machine, when to trust it, and even more importantly, when not to trust it.”

Prof. Dr. Ivan Kostov, MD, PhD – Executive Director of University Obstetrics and Gynecology Hospital “Maichin Dom” – Sofia, specialist in Obstetrics and Gynecology

“AI will raise the level of knowledge among both doctors and patients, and everyone will benefit.”

Professor Dr. Ruja Nikolova, MD (Faculty of Public Health “Prof. Dr. Tsekоmir Vodenicharov, MD, Medical University of Sofia)

Regarding: Medicine in 2026 - 2029 in the time of artificial intelligence from the perspective of the complex interaction in the socio-technical system of the components “human/doctor – artificial intelligence – social environment”

“The current opinion examines the likely development of medicine in the period 2026–2029 through the prism of the accelerated introduction of artificial intelligence (AI) systems, including generative and multimodal models, clinical decision support algorithms, medical imaging analysis systems, predictive models, and intelligent automation of administrative processes.

The main thesis is that in the next three years, medicine will not transition to a model where AI replaces the doctor. It is more likely that a new hybrid model of medical practice will form, based on the complex interaction between three main components: the human and the doctor – artificial intelligence – the social and institutional environment. The quality of future medicine will depend not only on the accuracy of algorithms. It will be the result of how technologies are integrated into clinical thinking, professional responsibility, the relationship between doctor and patient, the organization of the healthcare system, access to technologies, legislation, and public trust.

Artificial intelligence is likely to significantly change the diagnostic process, the processing of medical information, the personalization of treatment, medical documentation, scientific research, and the management of health systems. At the same time, its use creates risks associated with algorithmic error, data bias, automated trust, weakening of professional skills, violation of privacy, unequal access, and the transfer of excessive power to technological structures.

Therefore, the central question in the next three years will not be whether medicine will use AI, but under what conditions the human will remain the active bearer of medical responsibility and how technologies will be organized to enhance rather than undermine human relationships in healthcare. The period 2026–2029 is likely to be a transitional phase in which AI will move from limited experimental applications to broader integration into real medical practice. The most significant change will not simply be the emergence of new algorithms, but the transformation of the relationships between the patient, the medical professional, the technological system, and the social environment.

Therefore, the future of medicine should not be viewed through the simplified dilemma of “doctor or artificial intelligence.” It is more accurate to analyze it as a socio-technical system, in which the clinical outcome is formed by the interaction between three interdependent components: human/doctor ↔ artificial intelligence ↔ social and institutional environment. Scientific data show that AI has significant potential to assist in diagnostics, medical data analysis, treatment personalization, medical documentation processing, and the organization of healthcare services.

The current opinion defends the thesis that during the period 2026–2029, medicine will develop most successfully not by replacing the human with an algorithm, but by building a human-centered system of joint decision-making in which AI expands the cognitive and organizational capabilities of the medical professional, while the social environment provides the ethical, legal, and institutional conditions for its safe use.

During the period 2026–2029, medicine will enter a phase of accelerated integration of AI. However, this change should not be understood as a technological revolution that occurs independently of humans and society. AI will change diagnostics, information processing, labor organization, and the relationships between patient and doctor. At the same time, it will create new risks - algorithmic errors, biases, excessive dependence on automated decisions.

Therefore, the main task for medicine in the next three years is to build a human-centered socio-technical system. In this system, the human must remain the goal of medical care, the doctor - the bearer of critical professional judgment and responsibility, and artificial intelligence - a tool for expanding human capabilities. The social environment has the task of creating the necessary conditions through education, regulation, data protection, equitable access, and public oversight.

Therefore, the future of medicine should not be formulated as a question of whether AI will replace the doctor. The more essential question is whether healthcare systems will succeed in organizing the interaction between humans and technology in such a way that AI improves medical care without undermining human autonomy and professional responsibility. It is precisely this balance that will determine the state of medicine by 2029. The most successful model will not be the one with the most algorithms, but the one that manages to combine technological efficiency, scientific reliability, and human responsibility.”

Prof. Dr. Svetoslav Yovev (Cardiologist, specialist in invasive cardiology and electrophysiology, Head of the Department of Cardiac Pacing and Electrophysiology at UMHAT “Prof. Dr. Alexander Chirkov”)

“In my opinion, in the next 2–3 years, artificial intelligence will enter medicine much more seriously, but it will not replace the doctor. It will be a tool that will help us work faster and more accurately.

In electrophysiology, this is already evident. AI can analyze vast amounts of information, assist in mapping and localizing arrhythmias, and there is data that it can shorten the duration of some procedures. But when it comes to the actual intervention, the hand and experience of the operator are still irreplaceable.

The same applies to cardiac pacing. Pacemakers, defibrillators, and CRT systems provide a vast amount of information. AI will increasingly assist in analyzing it, in the early detection of problems, and in the individual monitoring and programming of patients.

For me, the main question is not whether AI will replace the doctor. It will not replace it. But the doctor who knows how to use it correctly will have an advantage over the one who does not use it.”

Prof. Dr. Alexander Oscar, PhD - (Head of the Department of Ophthalmology, MU - Sofia)

“In my opinion, in the next 2–3 years, medicine will change more than we are used to seeing in a decade. The main driver of this change will be artificial intelligence.

AI will not replace the doctor, but the doctor who knows how to use artificial intelligence will have a huge advantage over the one who does not use it.

We will see an increasingly wider application of algorithms that analyze imaging studies, laboratory results, and vast amounts of medical information in seconds. They will assist in early diagnosis, assess the risk of developing diseases, and allow us to choose increasingly personalized treatment for the specific patient.

In ophthalmology, this is already happening. The analysis of retinal images and OCT studies can assist in the detection of diabetic retinopathy, glaucoma, macular degeneration, and other diseases in their early stages. The next step is not just to detect the disease but to be able to predict its development more accurately.

Artificial intelligence will also change the daily life of the doctor by reducing the time for administrative work, assisting in the analysis of medical documentation, and providing us with much faster access to current medical knowledge. But there is something that technologies will not be able to replace - the human connection between the doctor and the patient. Medicine is not just about diagnosing based on data. It is trust, empathy, responsibility, and the ability to see the person behind the disease.

Therefore, I believe that the future will not be “artificial intelligence instead of the doctor,” but the doctor and artificial intelligence together. And paradoxically, the more technological medicine becomes, the more important it will be for it to remain human.”

Prof. Dr. Georgi Chernev (Founder and Owner of “ONCODERMA” and Head of the Clinic of Dermatology, Venereology, and Dermatological Surgery at the Medical Institute of the Ministry of Interior)

Perhaps it is appropriate to introduce a brief retrospective on the topic: cognitive activity and subsequent decisions: the machine (AI) or the human?)

Alan Turing is an English mathematician who introduced the concept of “machine intelligence” back in 1950, proposing the Turing test to assess whether a computer can think like a human.

John McCarthy is an American computer scientist who introduced the term “artificial intelligence” and organized the historic seminar at Dartmouth in 1956, which is widely considered the beginning of AI as a distinct academic discipline.

In 1943, Warren McCulloch and Walter Pitts created the first mathematical model of artificial neurons, laying the foundations for modern neural networks. Eighty years later, this model becomes more relevant than ever and begins to displace humans from various types of activities.

Medicine is one of the fields that can be seriously influenced by AI. And here arises the question: can AI change medicine for the benefit of humanity?

Medicine globally does not look good at the moment, even in the absence of AI. The frighteningly rising incidence of all forms of cancer, for example, which does not correspond with the loud daily discoveries of new, revolutionary therapies, is solid evidence of the lack of adequate solutions in the healthcare sector. I mean “adequate human solutions,” whose results are also “unpleasant.”

Will humanity listen to the recommendations of AI? Will it be given a chance to re/manage the healthcare map and health policies, starting from ground zero? Or will its application be limited to avoid confronting the unpleasant truth about health lobbying and its regulators?

These are personally the dilemmas I face.

AI and Regulatory Policies in Medicine

At the national level, there is an oversupply of personnel - doctors and institutions/hospitals that cannot be secured as jobs, pay, and good qualifications. This raises the question: can they be replaced by AI?

The answer for me is a big yes!

The reason is the motto of the new era: mass instead of class! And quick profit at any cost! The lack of control over the effectiveness and quality of medical activity, the number of healthcare facilities, and their actual accreditation creates serious problems.

Will AI help?

The answer is unclear, but rather positive. I think it should be given a chance, even though AI will limit to some extent the intervention of the human factor - and perhaps for the better.

The replacement or use of AI in medicine is likely possible initially in certain areas such as pathology, imaging diagnostics, administrative activities, quality control, and digitization.

But can AI find optimal organizational solutions to problems that have remained unresolved until now? Can it introduce innovations, guide towards new horizons, and plan health insurance systems in certain regions?

At the moment, the answer is at least unclear for me. But probably - yes.

Can AI bear responsibility in medicine?

This is an extremely important question. AI cannot bear responsibility and engage in risky, life-saving operations in medicine. Even da Vinci robots are operated by humans. At least for now.

In specialties such as surgery, reconstructive, dermatological and plastic surgery, neurosurgery, and vascular surgery, the human factor remains paramount.

The question is whether AI will be given the green light to verify the accuracy of its decisions and to conduct a comparative analysis with the human factor.

AI and Regulatory Health Policies

The lack of real objectivity in making certain human decisions in leading health policies could certainly be replaced not only by AI but also by most eighth graders in the country. But AI has no legislative power and cannot change the system on its own.

Due to this fact, I see a potentially serious role for it not only in decision-making but also in shaping future policies.

But will it be allowed to deal with corrupt models in medicine?

Independence and incorruptibility - these are probably the two strong qualities of AI.

AI could reduce the profits of certain circles, but it also has a chance to guarantee progress. And progress is an expensive gift, and someone has to pay the price.

The fear of its introduction in certain areas of medicine, in my opinion, is primarily based on the danger of terminating unhealthy health policies that benefit certain lobbies. Sometimes the decisions are human and unhealthy, personal, selfish, and well-funded.

There are numerous perspectives that speak in favor of artificial intelligence and its application in medicine, but at the moment this application is rather limited.

For younger specialists, it is certainly helpful - educational and informational. I myself refer to it regarding various topics.

But decisions for one or another type of therapy can only be made by clinicians - influenced and verified by AI. Having access to an objective information array that searches, analyzes, and provides data saves time and can help the human factor make more adequate medical decisions.

AI is a golden opportunity for those seeking solutions and needing quick analyses of large amounts of data and searching for connections between information arrays and various spheres of our existence, including medicine.

But responsibility in medicine is personal and is always borne personally.

Humanoid robots will likely disrupt this balance!”

Prof. Dr. Venelin Gerganov, MD (Professor of Neurosurgery, Senior Physician at the International Neuroscience Institute, Hanover, Germany, and Neurosurgeon at UMHATSM “N. I. Pirogov”, Sofia.)

“The application of artificial intelligence (AI) in medicine will lead to profound changes in almost all its aspects. The scope and consequences of these transformations can currently only be partially predicted.

When AI is used as an extension of the capabilities of the doctor or researcher - as a tool for solving problems previously considered unsolvable - it will allow for much deeper exploration of fundamental questions related to the emergence and development of diseases, as well as the functioning of the human body. Particularly significant progress is anticipated in the fields of neuroscience, oncology, and infectious diseases.

AI will make possible the planned and targeted creation of new medications, significantly shortening the time to synthesize the final product. The path to their clinical application will also be shortened thanks to new capabilities for simulating and analyzing large datasets. This will likely reduce the need for prolonged studies of the effects of drugs on animal models and limited groups of people. Thus, the period from fundamental discovery to its application in a specific patient - “from bench to bedside” - will be substantially shortened.

Particularly exciting will be discoveries in the field of brain-computer interfaces, which will allow for the restoration of lost brain functions through technological solutions. Along with the enormous therapeutic potential, however, a number of ethical dilemmas will arise regarding the possibility of using these technologies not only to restore impaired functions but also to augment and expand the natural capabilities of the human brain.

Clinical practice will also undergo significant change. Many of the processes involved in making medical decisions will be supported, and in certain cases - entirely taken over by algorithms. This applies to both daily medical work and tasks on a more global scale: monitoring, analyzing, predicting, and controlling epidemiological processes, as well as solving problems in the field of hygiene and social medicine. A large part of routine activity will be transferred to AI.

Mastering basic clinical skills will become easier, and the training process will be more effective. In this way, the average level of clinical practice could be raised and unified, providing higher quality and accessible healthcare.

However, top medicine will continue to require highly competent doctors who master and apply AI-based technologies while retaining their critical, controlling, and guiding role. The most human aspect of the medical profession will remain unchanged - the conversation with the patient, the need for empathy and understanding, the ability to share pain and instill confidence.

According to some authors, the introduction of AI cannot be compared to the Industrial Revolution. The steam engine, locomotive, and assembly line were new “tools of labor” in the hands of humans, revealing new horizons. Workers retrained, and their labor became more efficient and productive.

AI is also a new “tool of labor,” but one that at some point could replace not just individual activities but the very person who uses it. When AI ceases to be a tool in the hands of the doctor and displaces the doctor themselves, medicine will change fundamentally. A topic for reflection…”

Prof. Dr. Kosta Kostov (Professor and Pulmonology Consultant, Scientific Director of INSPIRO Medical Center, Chairman of the INSPIRO Foundation, and Editor-in-Chief of INSPIRO Magazine)

“I would like to start with a phrase that I often use, probably mine, until someone else appears with authorship claims:

The human mind is like the stomach - it digests knowledge if it is healthy or returns it if it is sick.

The same will happen with the doctor’s mind when using AI. Either it will be used healthily, processing and analyzing its data and advice, or it will become ill from its mistakes.

I do not know what awaits us in the more distant future when a real threat looms over humanity to lose control over AI. Ray Kurzweil in his book “The Singularity is nearer: When We Merge with AI” warns of the ongoing exponential, not linear development of technologies and AI, which will soon reach human levels of intelligence and begin to merge with it. Thus, around 2045, we will enter the stage of the so-called “singularity” - the merging of human and artificial intelligence. There will be a gigantic increase in collective human intelligence to the point where humans will gain access to control over their own biology. This sounds frightening if humanity does not safeguard itself and allows unclean forces to seize control over long-term health and longevity - both physical and digital. It sounds both inspiring and optimistic, but also frightening, especially if control over AI is lost and it begins to live a completely independent life, turning us into its subordinates.

But let’s not rush, we still have a few years of independent life from AI to use for the development of our own minds and to prevent merging, but only friendly assistance. In the next few years, the doctor will still not be replaced by artificial intelligence, but will rely on it as the closest assistant in daily work with patients and in scientific laboratories. It will continue to assist more intensively in the operating room, in therapeutic units, in laboratories, and scientific research. I also consult it when analyzing data and for quicker orientation in new developments and articles from medical journals, as I do not have that powerful capacity for quick and selective analysis.

AI gives the doctor a tool with a richer memory, something like an additional, added “brain,” which enhances our cognitive abilities, speed, and accuracy in decision-making through the interpretation of diverse clinical data for more precise diagnosis and more justified treatment based on evidence-based medicine. AI helps us with a gigantic memory, based on an immense database and powerful analytical capacity, with which the doctor’s intuition checks and corrects itself. In other words, the doctor has access to an informational capacity that enriches their knowledge and intuition to make fewer mistakes. AI saves the doctor immeasurable time for reading and wandering through literature, presenting summaries of vast amounts of scientific literature, daily information on medical news, selected news, and summarized articles that can be reviewed in the little free time during the workday. These are added capabilities beyond human cognitive capacity.

AI adds something fundamentally different: it integrates a vast number of information sources and offers clinical hypotheses, differential diagnostic possibilities. The information from the imaging study of a specific patient can be compared, contrasted with data from millions of patients, and guide the imaging diagnostician towards the most correct decision. A good and prepared doctor can rationally use the medical data of the patient, processed and analyzed by AI, which they interpret through their own experience and intuition to make an informed decision about the accurate diagnosis and appropriate treatment. Diagnosis and treatment will become a decision of a joint “consilium” involving the human brain and AI, which will facilitate the work of the doctor, who needs an interdisciplinary team, and decisions will be made in a narrower circle, with the “collective” participation of the “colleague” AI. For me, such “consiliums” happen daily. AI is my closest collaborator, but I do not allow a change of roles - the final decisions are made by me. This is the right and morally justified model of working with AI - a backstage “assistant” who stands beside me and at the bedside because the sick person needs understanding, compassion, and trust in the healer, not the cold flesh of the machine.

I have many chronic patients, long-term sufferers with several folders of medical documentation, with dozens of files of medical information - discharge summaries, laboratory tests, functional respiratory tests, imaging studies of the lungs, several hospitalizations, accompanying diseases, and therapeutic prescriptions with corresponding side effects, complex medication schemes that require careful reading. In my busy schedule, I often need more time for the more complex cases when there is diverse chronic pathology and a lot of available data to analyze. This time is needed for the patient, not for reviewing a vast amount of documentation. With AI, my time for processing and analyzing data will be saved, and I will have more time for each individual patient and will review more patients. AI can analyze the entire history of the patient in seconds if it has access to a unified database for all patients. This will allow the doctor, with the help of AI, to quickly prepare exit documentation, time that doctors today spend on administration rather than treatment. This is a completely different medicine, closer to its holistic version, which views the patient as an individuality in three dimensions - mind, soul, and body - rather than the modern conventional medicine that treats the patient primarily as a sick body. This is not a personalized approach, far from the concept of Marcus Aurelius and René Descartes. The processing of medical documentation will happen very quickly, and this change will be met with enthusiasm by doctors because it will increase job satisfaction and pleasure. Today, the doctor spends more time in front of the computer than at the patient’s bedside, and the time saved in front of the computer will go to the benefit of the patient. This will significantly increase diagnostic and therapeutic success, and the benefits will be entirely for the patients. The doctor will talk more with the patient and less with the computer. This, in my opinion, will be one of the most positive effects because the patient wants to share their problems with a compassionate soul, to share the psychological burden of suffering, and this is an important and indispensable part of the therapeutic outcome. While the doctor talks with the patient, AI can process the information from the history and the details of the conversation and provide a quick direction towards the likely diagnosis to guide the doctor in the decision. It is important for the doctor to rely on their own mind and not to overtrust AI, as they may become infected by its “hallucinations” or develop a dependency on it, which will be fatal for their professional intellect. Boundless trust in the competence of AI and its infallibility is a threat to the professional reflexes and skills of the doctor, who loses the ability to recognize its mistakes. AI should assist clinical thinking, not replace it.

The greatest revolution with the help of AI is already happening in personalized medicine, where AI does incredible things beyond human capabilities - analyzing the human genome and adapting treatment models, metabolic impacts, lifestyle, external influences to minimize the risks for certain diseases embedded in the structure of the genome. This is a fundamental change. Two patients with the same diagnosis are not viewed as identical but as different individuals - not the disease, but the suffering person. Such a personalized approach is gaining momentum most noticeably in oncology, but has been underway for at least 20 years in pulmonology, in socially significant diseases such as COPD and bronchial asthma.

Fundamental changes are also occurring in medical scientific research, which seeks a personalized approach, and the results will be directed towards increasingly refined groups of patients who share similar endotypes (disease characteristics).

The role and structure of medical specialties will also change. Some specialties will be more strongly affected than others, especially those that are part of the so-called para-clinics, supporting the work of clinical specialties. Radiology, pathology, and laboratory medicine will be particularly affected. I assume that AI will significantly impact general medicine specialists, as AI will provide more opportunities for quick “home” diagnosis of patients and will redirect them more often to narrow specialists and give them opportunities for remote access to medications. However, I am sure that the need for well-prepared doctors will increase, as the competition from AI will increase the need for good, capable, thorough, patient-dedicated doctors who can integrate all data and make informed decisions. AI will save technological time for the doctor, but will increase the value of true clinical thinking for those who strive for it and will more clearly highlight the valuable doctors, and they will remain the most competitive.

Medical education will also change dramatically. Students will receive not only basic medical knowledge but also training models on how to use AI to reach diagnostic and therapeutic decisions, but also how to recognize the mistakes and “hallucinations” of AI. Future doctors will be better than today’s, as they will benefit from additional, enhanced intelligence.

The use of AI will require even more scientific, specialized knowledge to avoid mistakes for which the doctor will bear responsibility, not AI. From here arise some legal questions - how far will the responsibility of doctors extend when using AI, how far will the responsibility of institutions, manufacturing companies, society, and the state extend.

The most important function of the doctor cannot be taken away by AI - the compassionate relationship with the patient, the meeting of two souls - the suffering and the co-suffering. The doctor will continue to carry the deepest meaning of the profession - mercy, human presence, compassion - at the core of therapeutic success. Doctors dedicated to the profession, understanding its fundamental role - to give hope and support in the difficult days of pain and suffering - will remain sought after. No one can replace the moral responsibility for the fate of the sick. AI should not be allowed to escape the control of the doctor to dictate the fate of the sick.

I expect that in the next 5 years, specialized national and international consensus (recommendations) for behavior in specific diseases will begin to account for the role of AI and introduce its use in diagnostic and therapeutic standards. This will make public healthcare more effective, and treatment more efficient, as they will become less dependent on the limited mental capacity of the individual doctor and will enable things that today require the collective capacity of an entire team.

I also have personal fears related to the likelihood that in highly developed technological societies, machines similar to today’s ATMs will soon appear, which will be used by people as doctors for outpatient problems - common infections, physical and mental discomfort, to receive advice from them, to prescribe medications, and why not even directly dispense medications according to a quick analysis of complaints. Then it will become frightening.”

Prof. Dr. Ivan Gruev, MD (Deputy Director for Medical Activities at NMTB “Tsar Boris III”, Professor of Cardiology and Elected Chairman of the Bulgarian Cardiologists Association for the term 2026–2028)

“Medicine will receive powerful support from AI for more precise diagnostics and for analyzing large databases, but the responsibility for decisions and human communication will remain entirely the priority of doctors.”

Prof. Dr. Boryana Deliyska, Nephrologist, Long-Standing Lecturer at the Medical University – Sofia

“AI may provide the correct diagnosis and treatment for a patient, but will it ever be able to replace the doctor’s smile or words of encouragement to the patient, which are also part of the treatment?”

Prof. Dr. Dimitar Maslarov, MD, PhD, DSc (Neurologist and Head of the Department of Neurology at First University Hospital – Sofia “St. John the Baptist”; Lecturer in Neurology at Medical University – Sofia)

“The future will not be man versus machine, but man plus machine.”

Artificial intelligence is already changing the way we work, communicate, and access information. Medicine is no exception. In some fields, AI is already helping physicians make diagnoses, analyze medical imaging, and assess patient risk.

Its potential is particularly significant in neurology – a specialty that deals with vast amounts of information from imaging studies, electrophysiological tests, laboratory results, and clinical data.

We spoke with Prof. Dr. Dimitar Maslarov about the role of artificial intelligence in neurology, its potential, and the risks it presents for medicine.

How close are we actually to the medicine of the future?

In fact, we are already there. Artificial intelligence is gradually evolving from a technological novelty into a real tool in medical practice.

The reason is simple – medicine generates enormous amounts of data. A single patient may have imaging studies, laboratory results, electroencephalography, medical records, genetic information, and numerous clinical indicators.

Humans can analyze this information, but our capabilities are limited. Computer systems can process huge datasets in a very short time and identify patterns that would be difficult for us to detect otherwise.

And why is neurology particularly well suited to the use of AI?

Because neurology is an extremely information-intensive specialty.

We work with brain structure, function, movement, behavior, electrical activity, and many other indicators. We have computed tomography, magnetic resonance imaging, EEG, electromyography, and other methods.

Artificial intelligence can help neurologists bring all this information together and analyze it in ways that would be difficult for a human to accomplish alone.

But here we need to make one important clarification – AI is not a doctor. It is a tool in the hands of a doctor.

Where do you already see the most significant applications?

One of the clearest examples is stroke.

In stroke care, time is of enormous importance. The faster the problem is identified and a treatment decision is made, the greater the chance of saving brain tissue and reducing the consequences.

AI can assist in the analysis of medical imaging and the rapid detection of certain pathological changes.

Another very important area is the diagnosis of neurodegenerative diseases – such as Alzheimer’s disease and Parkinson’s disease. Artificial intelligence can analyze images and other data and look for characteristic patterns.

Can AI detect a disease before a doctor does?

Potentially – yes, in certain situations.

This is one of the most interesting possibilities. An algorithm can be trained on a huge number of cases and recognize very subtle changes.

But this does not mean that the machine is necessarily “smarter” than the doctor. It simply has a different set of capabilities.

The doctor has clinical experience and understands the context. The algorithm can analyze enormous amounts of information with exceptional speed.

The combination of the two can be much more powerful than either one on its own.

So this is not about replacing doctors?

No. At least I do not see the future that way.

Rather, the role of the doctor will change.

It will probably become less necessary for physicians to spend their time processing huge volumes of information and increasingly important for them to interpret that information, communicate with patients, and make decisions.

I would put it another way: a doctor who uses AI is more likely to replace a doctor who does not use it than AI is to replace the doctor.

That is quite a strong statement.

But I think it is realistic.

Imagine two specialists. One has access to a tool that can analyze thousands of images in seconds and flag certain suspicious findings. The other performs everything manually.

Of course, the first is not automatically a better doctor. But if they use the technology critically and wisely, they can work faster and more efficiently.

Technological literacy will gradually become part of a physician’s professional competence.

And can AI predict how a disease will progress?

This is one of the most promising areas.

If we have enough high-quality data, we can look for links between patient characteristics and the subsequent development of a disease.

This is how we are gradually moving toward personalized medicine – treatment and monitoring tailored not only to the diagnosis, but also to the individual characteristics of each patient.

Of course, an algorithm’s prediction should not be treated as a final verdict. It is information that the physician must evaluate alongside everything else.

Epilepsy is also an area where there is a great deal of data.

Yes. An EEG, for example, can contain an enormous amount of information. During prolonged monitoring, a physician may have to analyze hours or even days of recordings.

AI can assist in identifying characteristic electrophysiological patterns and epileptic activity.

This can be particularly valuable during prolonged monitoring, when the volume of information is enormous.

And what will happen to rehabilitation?

There is also significant potential there.

For example, recovery after a stroke can take months. AI and robotic technologies can monitor movement, assess progress, and adapt exercises.

The development of brain-computer interfaces is particularly interesting. The idea is to use brain activity to control external devices.

For people with severe motor impairments, this could be of enormous importance.

You are also the author of a publication focusing on different generations and artificial intelligence. Why is this issue important for medicine?

Because different generations have different attitudes toward technology.

For younger people, AI can much more easily become a natural part of everyday life. For other generations, it may seem unfamiliar and even intimidating.

But there is something important to understand – being able to use technology does not automatically mean that you understand medicine.

A young person may work extremely well with AI but lack the clinical knowledge needed to properly interpret its results.

On the other hand, an experienced physician may have enormous clinical expertise but need additional training to use new technologies effectively.

That is why the future requires a meeting between experience and technology.

Where are the dangers?

There are several.

The first is putting too much trust in the algorithm.

AI can make mistakes. If a physician automatically accepts every result it produces, this can be dangerous.

The second concerns medical data. When we use huge datasets, we must ensure their security and the protection of personal information.

And the third issue is accountability. If an algorithm makes an incorrect recommendation – who is responsible?

These questions will not be solved by technology alone. We need medical professionals, IT experts, lawyers, ethicists, and institutions.

How does medical education need to change?

Doctors do not need to become programmers. But they do need to understand how these systems work, what their capabilities and limitations are.

They must be able to assess the quality of the information and understand that an algorithm may produce a probabilistic result rather than an absolute truth.

The most important thing is not to lose critical thinking.

How do you envision neurology ten years from now?

It will probably be very different.

AI will be integrated into medical imaging, prediction, patient monitoring, rehabilitation, and probably many other areas that are only now beginning to develop.

But despite all the technological changes, I believe that the human being will remain at the center of medicine.

A patient does not need only a diagnosis. They need someone to listen to them, explain what is happening, give them hope, and take responsibility for their treatment.

That is difficult to replace with an algorithm.

If you could send one message to people who are afraid that AI will replace doctors, what would you tell them?

We should neither be afraid of artificial intelligence nor regard it as infallible.

It is a tool – an extremely powerful tool.

The question is how we will use it.

I believe the future will not be “man versus machine.”

It should be “man plus machine.”

And if we use this technology wisely, the biggest winner will not be artificial intelligence.

It will be the patient."

Prof. Dr. Nikola Vladov, MD, PhD
(Head of the Department of Hepatobiliary and Pancreatic Surgery and Transplantation, Military Medical Academy (MMA); specialist in hepatobiliary and pancreatic surgery and transplantation)

“In recent years, we have witnessed an exceptionally rapid development in surgery, both in general surgery and, in particular, in hepatobiliary and pancreatic surgery. This progress has been driven by modern diagnostic imaging techniques, advances in minimally invasive surgery—particularly laparoscopic and robotic surgery—as well as the introduction of modern electrosurgical technologies.

Artificial intelligence represents another major turning point in the development of medicine, opening up new horizons for physicians. AI is expected to fundamentally transform oncological surgery by enhancing anatomical visualization and treatment planning through tools such as 3D reconstructions, radiomics, and virtual hepatectomy.

These techniques enable surgeons to plan resections based on tumor volume, proximity to major vessels, and the volume of the future liver remnant, while also allowing the simulation of resections of different extents, whether anatomical or atypical.

The introduction of radiomics enables the integration of imaging data from computed tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography (PET). The analysis of these combined imaging data allows for more precise preoperative planning and helps minimize surgical risk.

In robotic or laparoscopic surgery, virtual planning improves the surgeon’s ability to navigate within confined anatomical spaces and perform precise resections with less blood loss and greater operative efficiency.

The integration of radiomics into preoperative planning represents a promising step toward precision surgery. By improving preoperative risk stratification and supporting data-driven decision-making, radiomics- and AI-based systems may help optimize patient selection for surgery, tailor operative strategies, and ultimately improve oncological outcomes.

The use of artificial intelligence during surgery remains limited. The term ‘surgomics’ refers to the collection and analysis of intraoperative data; however, distinguishing between the different phases of an operation remains challenging, which currently limits the potential for real-time intraoperative AI assistance.

All of the above demonstrates that artificial intelligence cannot yet replace surgeons. However, it can provide substantial support for precise preoperative planning and the postoperative management of complications. AI also has an important role in reducing risk in patients with oncological diseases and may contribute to improving their prognosis.”

Prof. Dr. Kiril Karamfilov, MD, PhD – Head of the Cardiology Clinic at University Hospital “Alexandrovska,” specialist in Cardiology and Interventional Cardiology, President of the Bulgarian Society of Cardiology

“AI is changing the way we detect diseases, not the person who treats them. The next two to three years will not be the period when artificial intelligence replaces the physician. They will be the period when medicine begins to detect diseases long before the patient experiences the first symptoms. In my view, this will be the greatest transformation – the transition from reactive to truly proactive medicine.

In cardiology, this AI transformation has already begun. Algorithms can analyze a standard electrocardiogram and identify signs of heart failure, atrial fibrillation, cardiomyopathies, or valvular disease that may remain invisible to the human eye. Artificial intelligence assists in the analysis of echocardiographic images, the assessment of coronary atherosclerosis through computed tomography, and the integration of vast amounts of clinical data to predict an individual patient’s risk. This is no longer science fiction – it is the direction in which medicine is developing today.

The real transformation, however, will not be the algorithm itself. It will be access to quality healthcare. When a standard ECG or echocardiogram can be analyzed with the help of validated AI models, highly specialized expertise will no longer be a privilege reserved only for major university medical centers. A patient in a small town will increasingly have the opportunity to receive timely and accurate diagnosis comparable to that available in specialized clinics. This is a major opportunity to reduce inequalities in healthcare.

At the same time, artificial intelligence will transform the physician’s daily work. It will process medical documentation, analyze scientific literature, synthesize clinical information, and support decision-making. The most valuable resource AI will give us will not be faster data processing, but more time for the patient. Because good medicine begins with a careful conversation, the right interpretation, and trust between physician and patient.

We must not forget, however, that every technology has its limitations. Algorithms can be inaccurate, can be affected by the quality of the data on which they are trained, and can sometimes appear convincing even when they are wrong. This is precisely why the physician’s role will not diminish – it will become even more responsible. Artificial intelligence can provide probabilities, but clinical judgment, responsibility, and the final decision remain human.

I am convinced that the future belongs neither to the physician alone nor to artificial intelligence alone. It belongs to their partnership. The best physician of the next decade will not be the one who knows the most, but the one who knows how to critically use the capabilities of AI without losing clinical reasoning, human compassion, and a sense of responsibility. Because algorithms can recognize patterns, but only a physician can understand the person behind the diagnosis.”

Prof. Dr. Plamen Bozhinov, MD, PhD – Neurologist, specialist in Clinical Neurophysiology and Brain Stimulation, Medical University – Pleven, National Research Institute, UMHAT Medica – Ruse, Galileo Medical Center – Pleven, Neurovita Medical Center – Sofia

“In medicine, the uncritical use of AI can lead to a decline in clinical skills or even prevent them from being properly developed, particularly among doctors in training. Prolonged reliance on automated systems may weaken physicians’ independent judgment and ultimately impair their ability to make rapid decisions in critical situations.”

Prof. Dr. Milena Staneva, MD, PhD (Head of the Angiology Clinic, Acibadem City Clinic UMBAL Tokuda)

"The most important message is: AI assists, but the doctor makes the decision.

Over the next 2–3 years, AI will begin to transform medicine not so much by replacing doctors, but by changing the way physicians analyze information, assess risk, and make clinical decisions.

In vascular medicine, I see particularly great potential in early detection and risk prediction. Modern medicine has access to enormous amounts of information – imaging studies, laboratory parameters, functional tests, electronic health records, and data from monitoring devices. The challenge is no longer simply to obtain this information, but to be able to integrate and interpret it quickly and accurately enough.

AI can support precisely this process – identifying patterns that are difficult to recognize when analyzing individual parameters, identifying patients at increased risk, and directing the physician’s attention toward the need for additional testing or earlier intervention.

This could also change the philosophy of healthcare – from treating an already manifested disease to earlier detection of risk and prevention of its consequences. This is particularly important in chronic vascular diseases, because a significant proportion of the risk accumulates years before severe clinical manifestations appear.

AI will also support the personalization of medical care. Instead of looking at a patient solely through the lens of a diagnosis, we will increasingly be able to integrate their individual risk profile, comorbidities, test results, and response to previous treatment. This will support the choice not simply of the “right treatment for the disease,” but of the most appropriate approach for the individual patient.

But there is another, in my view fundamental, change – physicians will need to acquire a new type of competence. Doctors do not need to become programmers, but they do need to understand how AI systems work, what their limitations are, what data they have been trained on, and when their outputs may be misleading.

This is particularly important because AI can provide a convincing answer even when that answer is incorrect. Therefore, the most valuable competence for physicians will not simply be the ability to use AI, but the ability to critically evaluate AI-generated results and place them in the context of the individual patient.

In this sense, AI will not diminish the importance of medical expertise. On the contrary, it will probably raise the standards expected of it. The more tasks that can be automated, the more important clinical reasoning, interpretation skills, communication with patients, and responsibility for the final decision will become.

Therefore, in the coming years, I would expect not “medicine without doctors,” but medicine with more powerful intelligent tools in the hands of well-trained physicians.

For this to happen safely, however, we will need validated systems, high-quality data, clear rules regarding responsibility, protection of personal information, and training for healthcare professionals. The WHO also emphasizes the need for AI to complement rather than replace human judgment, as well as the importance of human oversight and multidisciplinary assessment.

Ultimately, the greatest benefit of AI will not be simply to make medicine more technological. It will be to help us detect diseases earlier, assess risk more accurately, and make more informed decisions – while preserving the human approach and the physician’s responsibility.

AI will not replace the doctor, but a doctor who knows how to use AI critically and responsibly will gradually have significantly greater opportunities for diagnosis, prevention, and personalized treatment."

Prof. Ognyan Brankov, MD, PhD (Department of Pediatric Surgery, St. Anastasia Multiprofile Hospital, Burgas, Bulgaria)

“A doctor should use artificial intelligence without losing the human being within.”

I recently turned 80, and I continue to work as a pediatric surgeon. This gives me a particular perspective on medicine today. During my professional career, I have witnessed enormous changes in surgery—in diagnostics, medical imaging, surgical techniques, anesthesiology and intensive care, medical equipment, minimally invasive procedures, and robotics. I remember threading coarse silk sutures through the eye of a needle. I remember the first ultrasound machine and the first CT scanner. I have seen things that seemed like science fiction at the beginning of my career gradually become part of everyday medical practice.

I have always been fascinated by science fiction literature, and I have read about things that are now beginning to become reality.

Today, we are already in the midst of another major transformation—the use of artificial intelligence. And this is no longer a subject of science fiction. Artificial intelligence, let us call it “it,” is gradually becoming a real tool for physicians, and its importance will continue to grow. It is becoming involved in establishing diagnoses, analyzing medical data, medical imaging, preoperative planning, and patient follow-up. It can process information within seconds, compare vast amounts of data, identify patterns, and assist physicians in making decisions. It will also transform medical education and scientific research.

But the more I think about the future of medicine, the more clearly I realize that the most important question is not what artificial intelligence can do, but what it cannot—or should not—be allowed to replace.

These thoughts have particular relevance to pediatric surgery. A child is not simply a smaller version of an adult. A child is a world of their own, with their own rules and physiological patterns. Their physiology and psychology are fundamentally different. And pediatric surgery is about much more than diagnosis and an operation.

There is a frightened child in front of you, and you have to do something that does not appear in any medical algorithm. You have to earn the child’s trust. Standing beside the child are the parents—equally anxious and frightened, also looking to the physician for reassurance and security.

Artificial intelligence can be extraordinarily intelligent. But intelligence and human sensitivity are not the same thing. It can analyze the physical condition of a sick child, but it cannot sense the fear in the child’s eyes. It can suggest diagnostic possibilities, but it cannot feel a mother’s anxiety. It cannot assume responsibility for a human decision.

That is why I do not see artificial intelligence as a competitor to the physician. I see it as an exceptionally reliable colleague. It may sound paradoxical, but the more powerful technology becomes, the more important the human side of medicine will be. If machines can take over part of the informational and analytical burden, physicians will have even more time for what machines cannot do—to talk, to listen, to understand, and to inspire trust.

In the years to come, artificial intelligence will fundamentally transform the medical profession, and I am convinced that this transformation is necessary. My only concern is that, in our pursuit of being modern, we may forget why we became doctors in the first place. Because ultimately, behind every diagnosis, there is a human being.

And in pediatric surgery-there is a child.

And when a child looks into our eyes, they are not looking for artificial intelligence. They are looking for a person they can trust."

Prof. Dr. Petko Karagyozov, MD, PhD – Head of the Gastroenterology Clinic and Head of the Digestive Endoscopy Unit at Acibadem City Clinic University Hospital Tokuda, Sofia; Professor of Gastroenterology at Medical University – Pleven

"ARTIFICIAL INTELLIGENCE IN ENDOSCOPY – A TOOL FOR HIGHER QUALITY

Artificial intelligence does not replace the endoscopist; it expands their capabilities. The technology has been present in the endoscopy room for years, and some of the first approved medical AI systems were developed precisely in gastroenterology. In screening colonoscopy, computer-aided detection systems improve the detection of precancerous lesions, while some platforms also assist in polyp characterization. Published data show that the greatest benefit is seen among endoscopists early in their training and those with lower adenoma detection rates. However, increased detection does not automatically translate into greater patient benefit or better long-term outcomes. Important questions remain: Does excessive reliance on AI lead to a loss of skills? Does detecting such tiny lesions actually change the patient’s prognosis? What is the real benefit to the patient? On the other hand, our hopes that polyps could be removed without sending them for histological examination have not yet been fulfilled, while the impact on colorectal cancer morbidity remains unclear.

The increasingly widespread use of systems that objectively assess the quality of bowel preparation for colonoscopy, the completeness of the examination, how the procedure is performed, and the extent to which it meets quality criteria is expected. For the endoscopist, this means objective feedback, standardization, and the ability to monitor the quality indicators of each gastroenterologist, providing a tangible practical benefit.

AI is also becoming increasingly prominent in upper endoscopy. Platforms exist that assist in detecting early neoplasia in Barrett’s esophagus, monitor the completeness and quality of the examination, or support the detection of early gastric cancer. The goal is not to replace clinical judgment but to reduce the likelihood of missed lesions.

The greatest potential for change lies in endoscopic ultrasound. Systems are being developed to improve the detection of early pancreatic tumors, monitor examination completeness, and support the training of novice endosonographers. Platforms also exist that guide the selection of the biopsy site, thereby improving the diagnostic accuracy of the method. AI can combine multiple macroscopic imaging features, for example in the early diagnosis of chronic pancreatitis. Such integration of images and imaging modalities is beyond the capacity of the human mind.

In interventional endoscopic ultrasound, AI is expected to support difficult decision-making, such as selecting a drainage strategy or the optimal route. In endoscopic retrograde cholangiopancreatography (ERCP), AI is expected to predict difficult cannulation and the risk of complications.

AI can also free up time outside procedures. Workflow management platforms assist in generating reports from completed procedures, structuring findings, and maintaining consistent and organized documentation. This could give physicians more time for clinical decision-making and patient care.

AI is already helping standardize quality and reduce differences between operators. It increases the detection of precancerous lesions, but its long-term benefit for patient prognosis has yet to be established.

A significant transformation in endoscopic ultrasound is expected through accelerated training and improved detection of early pancreatic lesions, but AI is not expected to replace the physician’s training, experience, and responsibility."

Prof. Dr. Boris E. Sakakushev, MD, PhD, MHM - Professor of General Surgery, Senior Lecturer and Consultant in General Surgery, Medical University of Plovdiv, Faculty of Medicine

“I believe that integrating artificial intelligence (AI) into clinical practice represents a fundamental shift in medical doctrine. It will not replace the surgeon, but will fundamentally reshape surgical education, intraoperative decision-making, and clinical responsibility.”

Prof. Dr. Plamen Kinov, MD, PhD – specialist in orthopedics and traumatology, Head of the Department of Orthopedics and Traumatology at VITA Hospital, and Head of the Department of Orthopedics and Traumatology at Medical University – Sofia

“Medicine will become increasingly technology-driven, but good outcomes will continue to depend on a physician’s knowledge, experience, and clinical judgment. AI will not make good specialists obsolete – on the contrary, it will expand their capabilities.”

Prof. Dr. Dobromir Dimitrov, MD, PhD

Rector of Medical University – Pleven

“If we look at the development of medicine, major technological revolutions have always expanded the boundaries of what doctors can see, understand, and do. Artificial intelligence is ushering in a new stage in this development. What distinguishes it is not simply its ability to process vast amounts of information, but its ability to identify patterns, connect data from different sources, and turn this information into a basis for more precise clinical decisions. Therefore, over the next 2–3 years, in my view, we will not simply witness the adoption of AI in medicine. We will witness a gradual reshaping of the way medicine ‘sees’ the patient.

And perhaps the most natural place for this transformation to begin is the physician’s perspective. AI will expand the physician’s diagnostic perspective and gradually become that additional ‘second pair of eyes’ that will make it possible to detect even the most subtle abnormalities through the simultaneous analysis of imaging, pathological and endoscopic findings, and laboratory data. In this way, technology will not replace clinical judgment, but will enhance it by enabling more precise diagnosis.

But when the machine can already see more, the natural next question is: can it also help the physician act more precisely? This brings us to the next significant area – robotic surgery. Sixth-generation systems will integrate AI that, in real time, will recognize anatomical structures, blood vessels, nerves, and critical areas. And in my view, the next step will not be the autonomous robot that operates on its own. More interesting is the robot that ‘understands’ the operative field – one that can warn the surgeon of a potential risk, provide additional information, and support decision-making at a critical moment.

The development of these technologies opens up another extremely important possibility – for distance to cease to be a medical boundary. This is where telesurgery and tele-mentoring can be extremely important. The combination of robotics, AI, and high-speed communication networks will allow specialists to reach patients even when they are thousands of kilometers away. This could change not only access to highly specialized surgery, but also the way knowledge is transferred and young surgeons are trained.

But if we can overcome physical distance, the next challenge is to turn the enormous volume of medical data into a basis for an individualized decision for the specific patient. This is where Big Data and personalized medicine come in. AI will be able to combine genomic data, imaging studies, laboratory parameters, electronic medical records, and treatment outcomes collected on a scale that has previously been difficult to achieve in medical practice. And this could gradually change a fundamental question in medicine. Instead of asking only, ‘What usually works for this diagnosis?’, we will increasingly be able to ask, ‘What is most likely to work best for this particular patient?’

And from personalizing treatment, we naturally arrive at the next step – treatment itself. AI has the potential to analyze enormous numbers of molecules that could serve as the basis for new medicines, predict their effects, their interactions, and their potential toxicity, and thus significantly accelerate the discovery and development of new drugs, including antibacterial and anticancer agents. This could shorten one of the most complex stages – the path from scientific discovery to potential therapy.

But perhaps the deepest transformation will not be in diagnosing faster or treating more precisely. It will be in beginning to predict what lies ahead. This is the direction of predictive medicine. AI could gradually shift healthcare from a model in which we react after a problem appears to a model in which we try to identify it before its clinical manifestation. The risk of sepsis, decompensation, readmission, postoperative complications, or disease progression could be identified earlier, when there is still time for prevention or timely intervention.

And this is where another, often underestimated aspect of this revolution actually emerges. AI will not only change the way we treat patients. It will also change the way the healthcare system itself functions. AI will gradually enter the organization of healthcare – from documentation and triage to hospital resource management, patient monitoring, and prevention – these are areas in which AI can take over a significant part of the informational and administrative burden. In this way, technology will not simply save time, but will allow physicians to focus on where human presence is irreplaceable – in clinical judgment, diagnosis, and treatment, and above all, in a meaningful physician–patient relationship. Because the more intelligent technologies become, the more clearly the value of the human element in medicine will stand out – clinical judgment, responsibility, empathy, and trust.

And if I have to summarize what I expect over the next 2–3 years, I would say the following: artificial intelligence will not replace the physician. It will expand the physician’s senses, increase their ability to analyze information, and give them the opportunity to make decisions based on a much broader data foundation.

And perhaps the most impressive aspect is that this transformation will gradually remove the last boundary that for a long time we have taken for granted – distance. Imagine a physician who can analyze more, predict earlier, and be there for the patient even when physically located thousands of kilometers away. This is precisely where, in my view, the three major technologies of our time converge – artificial intelligence, robotics, and telecommunications. And their integration could transform not simply medical technology, but the very model of healthcare.”

Prof. Dr. Todor Al. Popov (Consultant Professor at St. Ivan Rilski University Hospital in Sofia, former President of the Union of Bulgarian Medical Specialists, former President of the World Allergy Association “INTERASMA”, Vice-President of the European Academy of Allergy and Clinical Immunology)

“Before my eyes are the frightening images, widely circulated on social media, of the torrent caused by the melting glacier in Nepal. I am astonished to subsequently learn that Google, using a program based on Artificial Intelligence (AI), has determined that AI was involved in their creation!

With this example, I want to illustrate my concerns about the uncertainty we face on the threshold of the AI Era we are entering. We are confronted with a cognitive power that surpasses human capabilities many times over. Among the leading creators and custodians of AI, some predict everything from global prosperity (Elon Musk) to the risk of the extinction of humanity (Geoffrey Hinton, known as the Godfather of AI).

In my opinion, in the short term – over the next 2–3 years – the integration of AI into our daily lives is likely to bring more benefits than harm in the field of medicine. It will penetrate the healthcare infrastructure – processing and creating documentation, analysis, planning and optimization of inpatient and outpatient activities at various levels. At the level of research and technological development, AI will contribute to diagnostic clarification, including detailed imaging and laboratory technologies. The AI-driven design of new medicines, which is already underway, will accelerate progress in the treatment of chronic and currently incurable diseases. AI, in its role as a “training” and “monitoring” factor, could raise the level of healthcare professionals’ training, but, on the other hand, reassess the need for specialists in different medical fields.

It will be very interesting to see how the use of AI will affect patients and the general public – to what extent laypeople’s overreliance on it could lead to undesirable consequences or, alternatively, to a decline in trust in doctors.

A race to adapt to the new conditions lies ahead: success will come to those organizations and doctors/researchers who manage to navigate the selection of the optimal AI agent and master the art of communicating with it.

I remain hopeful that we will be able to distinguish between reality and AI reality, and that benevolent rather than malicious AI will prevail – both for humanity as a whole and for medicine in particular.”

Prof. Dr. Elena Piperkova, MD, PhD, DMSc
(Head of the Nuclear Medicine Clinic at the University SBAL of Oncology “Prof. Ivan Chernozemsky”, Sofia. Lecturer to medical students, supervisor of nuclear medicine residents and PhD students in the scientific field of Oncology. Author of more than 250 scientific publications, with contributions to textbooks and monographs and presentations at European, international and national scientific forums. Holds three medical specialties and a specialty in Medical Management, as well as two doctoral degrees: Doctor of Medicine and Doctor of Medical Sciences. Recipient of various professional distinctions)

"Natural intelligence, with its knowledge and databases, creates AI, and AI will support high-tech medicine. But the doctor and the patient remain HUMAN!"

Prof. Dr. Albena Zlatareva
(Department of Pharmacology, Toxicology and Pharmacotherapy, Faculty of Pharmacy, Medical University – Varna)

"Artificial intelligence is opening a new era in medicine. The ability to process vast amounts of information, identify patterns, and support clinical decision-making deserves serious attention and professional support. I believe that this technology can significantly improve medical practice. Precisely for this reason, I believe it is necessary to clearly define its place: it expands our capabilities, but it does not relieve us of our responsibility to the person seeking medical care.

When I reflect on this boundary, I return to the monologue of the psychologist Sean in Good Will Hunting. He reminds the brilliant Will that he may know Michelangelo and his work extraordinarily well without ever having stood beneath the ceiling of the Sistine Chapel and experienced what it smells like there. Sean directs the conversation toward the difference between accumulated knowledge and lived experience, toward social experience. This distinction carries particular weight in medicine.

For me, the analogy lies in the way we come to know a person. We may have detailed information about a patient's illness without yet understanding how it is changing their life. Medical records describe symptoms, test results, and diagnoses. In the encounter with the patient, we learn what these mean for their family, work, social integration, and vision of the future with the disease. This understanding requires a human approach.

The quantitative accumulation of data and computational capabilities is already bringing about a qualitative transformation in medical work. AI can assist in image analysis, the summarization of medical records, the use of scientific literature, and research processes. When used appropriately, it can relieve physicians of part of the information burden and create more time for patients. The benefit depends on the specific system, task, and manner of implementation.

The evidence for some applications is compelling. An analysis of the randomized MASAI trial published in 2026, involving more than 105,000 women, reported higher sensitivity in AI-supported mammographic screening: 80.5% compared with 73.8% with standard double reading, with the same specificity of 98.5%. The result demonstrates the potential of a specific model of collaboration between technology and radiologists. It does not mean that every AI system is reliable or that physician involvement is unnecessary. Applying such findings requires clinical judgment. A system evaluated for a particular task and population must also be assessed in the conditions in which it will actually operate. Algorithmic accuracy is only one part of the evaluation. Missed cases, unnecessary examinations, consequences for treatment, and the team's ability to recognize errors are also important. The novelty of a technology, by itself, does not demonstrate added value for the patient.

A physician's social experience is accumulated through communication with people in different circumstances of life. A patient who does not follow prescribed treatment may be afraid of adverse effects, may not understand the instructions, or may not have the financial resources needed for treatment. To help, we must create an atmosphere of trust in which the patient feels able to share these concerns. Information about such difficulties can also be processed by AI, but reaching that information often begins with a careful human conversation.

Empathy also requires precision in how we discuss it. A 2025 study published in npj Digital Medicine found that participating patients with cancer rated chatbot responses as more empathetic than physicians' responses to written questions. This presents an opportunity to improve communication and a reason to critically examine our own way of communicating. However, the study assessed the perception of written responses, not a system's ability to replace a sustained therapeutic relationship.

A 2026 study published in Communications Psychology is also instructive: across four studies, participants preferred a human source of empathy, even though they rated the AI-generated responses highly. These psychological experiments are not evidence of clinical effectiveness, but they highlight the importance of personal choice. For patients, it matters who listens to them, who knows their history, and who will remain responsible for their care after the conversation ends.

Therefore, today's AI cannot be regarded as a complete substitute for the physician's social experience, empathy, and critical thinking. It can propose hypotheses and arguments, flag inconsistencies, and support reasoning. The physician, however, must verify whether the underlying data are reliable, whether the recommendation is applicable to the particular person, and whether the expected benefit justifies the risk. This judgment includes the patient's values and preferences, which cannot be determined on their behalf through a statistical prediction.

Human judgment is also vulnerable to errors and biases. Professional maturity requires us to accept well-validated assistance, including when it challenges our initial opinion. At the same time, maintaining our own competence and pursuing continuous learning are essential. An observational study published in 2025 involving colonoscopy found lower adenoma detection rates in examinations performed without AI after its introduction into practice. The study design does not establish a causal relationship, but it raises a serious question about the risk of dependence on automated assistance.

The most important question for me remains the ethical one. The pursuit of faster implementation, economic efficiency, or technological leadership does not justify compromising patient safety, privacy, informed choice, or dignity. The International Code of Medical Ethics requires independent professional judgment, responsibility for medical decisions, and protection of the patient's interests from commercial influences. These obligations also apply when an algorithm is involved in the decision-making process.

The World Health Organization warns about inaccurate or biased responses from generative models, excessive reliance on automation, and risks to health data. Therefore, implementation must include training, verification of outputs, and ongoing safety monitoring. Responsibility should be clearly distributed among developers, healthcare institutions, and medical professionals.

As a physician and educator, I believe that the training of future colleagues must simultaneously develop skills for working with AI and independent clinical reasoning. They must be able to evaluate sources, recognize whether data have been validated or not, and justify their decisions. It is equally necessary for them to be able to speak with a person who is afraid, explain the limitations of medicine, and acknowledge when help from another specialist is needed. Technological literacy expands professional competence, but it does not lower the standards expected of it.

My hope is that AI will help us devote more attention to these human dimensions of patient care. Time saved has meaning when it is transformed into better listening, clearer explanations, and more well-founded decisions. If we use technological progress solely to accelerate the flow of patients, we will miss an essential part of its potential value.

Sean's lesson remains relevant: vast knowledge does not exhaust the experience of living. AI will continue to expand our access to knowledge and our ability to use it. Our task is to turn this power into care that preserves human presence. In front of the patient, there must be a physician capable of considering the recommendation, understanding the person, and taking responsibility for the decision."

Prof. Dr. Naydenka Ivanova Zlatarеva, MD, PhD, Professor of Cardiology, Cardiologist, St. Sofia Hospital – Sofia

“AI has significant potential to help. However, let us not forget that all the information contained in this platform has been accumulated and is continuously enriched thanks to the efforts and functions of the human brain.

Without human intervention, adequate treatment cannot be provided. And this is of crucial importance for the health and life of every patient.

As a lecturer, I believe that the information students can obtain from AI is not sufficient to develop into good doctors. To a large extent, the same applies to medical residents.

Quite a few misunderstandings can arise when patients decide to use this platform. Sometimes it takes considerable effort on the part of the doctor to explain to the patient what exactly the AI is saying.”

Prof. Dr. Petar Petrov, DSc – Professor of Obstetrics and Gynecology and Honorary Professor at the Department of Midwifery, Medical University – Plovdiv; Chairman of the Bulgarian Scientific Society of Obstetrics and Gynecology.

Artificial Intelligence – A New Stage in the Development of Medicine

Artificial intelligence (AI) is no longer a technology of the future but a real part of modern medical practice. In recent years, advances in machine learning, deep neural networks, and generative artificial intelligence have created opportunities that are changing the way physicians collect, analyze, and interpret medical information.

One of the most significant applications of AI is diagnosis. Algorithms can process enormous amounts of data, including medical imaging, laboratory results, electronic medical records, genetic information, and findings from previous examinations. For certain tasks, these systems can identify patterns that are difficult for the human eye to recognize. This is already finding applications in radiology, pathology, cardiology, oncology, ophthalmology, and many other medical specialties.

The use of AI for the early detection of diseases is particularly promising. In medicine, time is often a critical factor. The earlier a disease is detected, the greater the opportunities for effective treatment. Artificial intelligence can support physicians in identifying patients at risk and analyzing complex diagnostic indicators.

AI also has the potential to transform preventive medicine. By analyzing large volumes of information, individual risk models can be developed, enabling a more personalized approach to prevention and patient monitoring.

However, it is important to emphasize that AI should not be regarded as a substitute for the physician. It is a tool that expands the capabilities of medical professionals. The final decision should remain in the hands of a qualified physician who combines technological analysis with clinical experience, medical ethics, and the individual characteristics of the patient.

II. Personalized Medicine, Treatment, and Healthcare Organization

The impact of artificial intelligence is not limited to diagnosis. One of the most important areas in which AI can transform medicine is personalized treatment. Every patient is different, with unique genetics, lifestyle, comorbidities, and responses to therapy. Analyzing multiple individual characteristics makes it possible to develop more precise models for treatment selection and monitoring.

In oncology, for example, AI can support the analysis of histological specimens, medical imaging, and the molecular characteristics of tumors. This creates opportunities for more accurate risk assessment and the selection of therapeutic strategies. Similar approaches are gradually being introduced into other medical specialties as well.

The application of AI in surgery is developing particularly rapidly. Robotic systems, computer vision, and intelligent analysis of surgical data can support surgeons in planning and performing interventions. In the future, increasingly close interaction between physicians, robotic systems, and data-analysis algorithms is expected.

There is also significant potential in obstetrics and gynecology. Artificial intelligence can support the interpretation of ultrasound images, the assessment of high-risk pregnancies, the analysis of cytological and histological samples, the early detection of oncological diseases, and patient monitoring. In reproductive medicine, AI can be used to analyze embryos, assess reproductive potential, and optimize certain stages of assisted reproduction.

Beyond direct medical activities, AI can also transform the organization of healthcare. Automating administrative processes, processing medical documentation, resource planning, and communication with patients can reduce the administrative burden on physicians. This would allow more time for the most important part of medicine – the relationship between the physician and the patient.

III. Challenges, Ethics, and the Future of Medicine

Alongside its enormous opportunities, artificial intelligence also raises serious questions. One of them concerns the reliability of algorithms. AI is trained on data, and if that data is incomplete, inaccurate, or insufficiently representative, there is a risk that the algorithm may reach an incorrect conclusion. Therefore, AI-based medical systems must be validated, monitored, and continuously evaluated.

Another fundamental issue is the protection of personal medical data. Health information is among the most sensitive categories of personal data. Its use for training and operating artificial intelligence systems must comply with legal requirements, the principles of medical confidentiality, and patients' rights.

The issue of responsibility is equally important. If an artificial intelligence system provides an incorrect diagnosis or treatment recommendation, it must be clearly established who is responsible for the final decision. This is precisely why human oversight remains a key principle in the implementation of AI in medicine.

Medical professionals also face a new professional challenge – the need for digital and technological literacy. The physician of the future will not simply be a specialist who understands medical science. They will also need to understand how intelligent systems work, what their capabilities and limitations are, and when the results produced by an algorithm should be questioned.

Artificial intelligence will not eliminate the need for physicians. Rather, it will transform the role of the physician itself. From a specialist who primarily collects and interprets information, the physician will gradually become a professional who manages vast amounts of information and uses intelligent technologies to make more informed decisions.

The future of medicine will probably not be a confrontation between humans and machines. It will be a partnership between physicians and artificial intelligence. A machine can analyze millions of pieces of data, but it cannot replace human empathy, moral judgment, or the trust that a patient develops with their physician.

Therefore, in my view, artificial intelligence can make a good physician even better, but it cannot turn a weak physician into a good one.

The real question is not whether artificial intelligence will change medicine. It is already changing it. The question is how we will use this technology to make medicine more accurate, faster, more accessible, and more personalized, without losing the most important thing in it – the human being.

Prof. Dr. Nikolay Nikolov, MD, PhD – Rheumatologist and Head of the Department of Rheumatology at Heart and Brain Hospital – Pleven.

Artificial Intelligence in Medicine – Predictions and Hopes

"Over the past 15 years, we have witnessed the rapid development of new technologies, which have made a powerful impact across all areas of our lives. Perhaps nowhere has this been felt more profoundly than in medicine – in diagnostics, genetic engineering and new biological therapies, computer technologies, and new approaches to medical education and training. The emergence of artificial intelligence was a natural next step.

Its accessibility and ease of use are leading to its increasingly rapid adoption in medicine. It seems that our patients have learned to use it even faster, most often for convenience and, in some cases, to avoid a visit to the doctor. Very often, they consult AI about their diagnosis and treatment before entering the doctor's office and then begin to “test” the specialist.

In my opinion, artificial intelligence is a powerful tool with virtually unlimited potential that will be extremely useful in diagnostic and therapeutic practice. We simply need to learn how to use it effectively so that it can build upon and enhance our routine knowledge and skills.

This makes it necessary to introduce educational modules into the curricula of medical universities as quickly as possible, as well as courses and online training for medical specialists and general practitioners.

We cannot turn back time. Artificial intelligence is our future, and we must learn how to work with it in the best interests of our patients."

Prof. Dr. Lyubomir Aleksiev – Head of the Department of Aviation and Maritime Medicine at the Military Medical Academy (MMA); Chairman of the Bulgarian Association of Aviation, Maritime and Space Medicine

"Let us not forget that AI is a product of natural human intelligence. Just as many innovations and discoveries throughout human history have contributed to development and progress, while others have gradually or rapidly exhausted their significance, and still others have been rejected as unnecessary or even dangerous, AI too will find its place in the life cycle of humanity. Whether it will bring benefit or harm is still too early to predict, but this human achievement will continue to develop until it either proves useful and moves to another level of necessity and refinement, or is rejected and the human mind once again demonstrates its innate capacity for self-preservation and purification.

We are currently in a period of rapid and, at times, unpredictable development of AI. This should not frighten us; on the contrary, it should encourage society to seek ways to regulate and adapt to the processes it has itself created.

As a physician working in aviation medicine and, to some extent, space medicine, I believe that aviation, as it has done many times before, provides lessons and models for the implementation of innovations in the medical, technical, and management fields. Progress regarding the role of the human factor in effective and safe medicine has been considerably slower and more conservative, based on the fundamental principles of aviation – flight safety and efficiency.

Aviation is the safest mode of transport and one of the fastest-developing sectors in technological terms. This is also demonstrated by the efforts to rapidly implement AI in the aviation industry, which once again is advancing ahead of a number of areas of economic and social development.

One of the leading reasons is the globalization of aviation activities and their regulation by international and global organizations such as ICAO, EASA, FAA, and others. These organizations establish laws, rules, and regulations for all participants operating in the world's airspace, which are mandatory to comply with. Roadmaps, discussion documents, and projected timelines for the implementation of AI in aviation have already been developed, clearly indicating that AI will be introduced, and that this is expected to happen within the coming years.

In medicine, however, there are still no detailed, universally binding international regulations, and in some countries not even national ones, that would guide and organize medical activities according to unified standards for effective and safe healthcare. Progress in this direction would contribute to a clearer and more predictable future for the role of AI in medical science and practice.

A number of institutions working on the implementation of AI in medicine already have a vision of how it will enter healthcare, and in some cases there are already working models for specific specialties and activities. At this stage, AI is used primarily for collecting and processing data, developing programs and approaches, rapidly retrieving information from enormous databases, and even providing suggestions for solving clinical cases.

In all cases, the principle of human oversight and a critical attitude toward what artificial intelligence provides remains essential. We are still far from creating a human–AI team (teaming) in which there will be genuine partnership and a distribution of rights and responsibilities in the traditional understanding of teamwork.

An interesting question is that of the human being who controls AI. Who should this person be, and what qualities should they possess? I hope we do not allow a selection between people who can and cannot communicate with AI. This relationship should begin at an early age so that every person can be fully prepared for the future human + AI society."

Prof. Dr. Lyubomir Kovachev, MD, PhD – Surgeon and Long-Standing University Professor at the Medical University of Pleven

"Artificial Intelligence in Medicine: What Will Change by 2029 and What Will Remain the Same

Over the next two to three years, artificial intelligence will not “revolutionize medicine.” It will do something less spectacular but more useful: it will take over part of the administrative burden and assist with the interpretation of medical examinations. These are the areas where there is already substantial evidence. Everything else — autonomous diagnosis, treatment without a physician, and new drugs created by machines — remains beyond this horizon.

First: Documentation. This is Already Happening

If you work in a hospital, the most likely change you will experience by 2029 will not be in diagnosis, but in documentation.

The reasons are measurable. Physicians can spend up to two hours documenting for every hour spent interacting with a patient. For nurses, documentation tasks can consume up to 60% of their working time. Other studies show that physicians spend more than one hour on electronic health record tasks for every hour of direct patient contact, while more than half of their working day can be devoted to documentation — compared with only about one quarter spent face-to-face with patients.

The evidence here is no longer based solely on observation. In a pragmatic randomized real-world study, the use of ambient AI reduced burnout and interpersonal disengagement among clinicians, while documentation time decreased without compromising diagnostic accuracy, billing compliance, or the quality of medical records.

There is also an important caveat: the results are not consistent everywhere. Other studies have found no significant improvements in productivity or patient experience. In other words, the benefit depends not on the model itself, but on how it is implemented within a specific workflow.

Second: Medical Imaging. The Best-Established Use Case

The scale of the data can no longer be ignored.

A prospective German real-world study conducted across 12 centers and involving 119 radiologists included 463,094 women, of whom 260,739 were screened with AI support. The breast cancer detection rate in the AI-supported group was 6.7 per 1,000 compared with 5.7 per 1,000 in the control group — 17.6% higher. The recall rate for additional examinations was 37.4 per 1,000 compared with 38.3 per 1,000, meaning it was slightly lower. The positive predictive value was 17.9% compared with 14.9%, while the positive predictive value of biopsy was 64.5% compared with 59.2%.

The randomized MASAI trial adds the second pillar: among 105,934 women, cancer detection rates were similar, but the reading workload was reduced by 44.3%. A subsequent publication reported a 29% increase in cancer detection, with only minor increases in recall and false-positive rates — 8% and 1%, respectively.

A third approach—partial autonomy—has been prospectively tested on 31,301 women. Radiologists’ workload falls by 63.6%, the detection rate increases by 15.2% (from 6.3 per 1,000 to 7.3 per 1,000), while the recall rate rises by 14.8%.

But the three studies demonstrate three different trade-offs between detection, recall, and workload. The choice among them is clinical and organizational, not technical.

Third: Clinical Decision-Making. Here, Machines Are Not Ready Yet

This is the most important distinction in the entire field—the distinction between performing a test and making a clinical decision.

In a study involving 2,400 real-world patient cases and four common abdominal pathologies, current leading language models did not arrive at the correct diagnosis across all pathologies, followed neither diagnostic nor treatment guidelines, and could not interpret laboratory results. The difference in mean diagnostic performance between physicians and the models ranged from 16 to 25 percentage points. Diagnostic accuracy among clinicians varied—the German hospital residents performed slightly worse (87.50%) than a more senior U.S. hospitalist (92.50%). Most models managed to match human performance only on the simplest diagnosis: appendicitis.

The positive side of the same picture is that models perform better as assistants than as autonomous diagnosticians. In a study of 302 difficult real-world cases assessed by 20 clinicians, physicians assisted by the models achieved 51.7% top-10 accuracy, compared with 36.1% without assistance and 44.4% when using search engines.

A systematic review of 550 studies confirms this asymmetry: there are benefits, but there is also a risk of over-reliance on the systems.

Fourth: Acute Care. The Greatest Promise at the Highest Cost

This is where the gap between the scientific literature and reality is greatest.

In sepsis, despite nearly 500 published articles, there are only two randomized clinical trials evaluating benefits in terms of patient outcomes.

The positive prospective data are specific: following the implementation of a deep-learning model, hospital mortality fell from an expected 10.3% to an actual 9.5%, while compliance with the sepsis bundle increased from an expected 48.3% to an actual 53.4%.

There is also a more striking figure. In an 18-month cluster-randomized study across 16 institutions, a tiered-alert approach resulted in a 23% reduction in unnecessary transfers to the intensive care unit and an 18% reduction in sepsis mortality—one life saved for every nine patients treated. At the same time, the study reported a 14% increase in nurses’ cognitive workload.

That is the key point. Approximately 70% of alerts in intensive care units are identified as false alarms. The benefit and the cost are inseparable: without redesigning the workflow, the system saves lives while simultaneously exhausting the people who use it.

Fifth: Surgery. A Quality Tool, Not an Autonomous Operator

For the surgical community, the topic is directly relevant—and it shows how easily statistical significance can be confused with clinical significance.

A meta-analysis of five randomized trials involving 4,354 patients found a higher overall adenoma detection rate with computer-assisted detection—36.6% versus 25.2%. The effect was most pronounced among beginners (60.0% versus 41.9%).

But the other side has also been documented. A multicenter randomized community-based trial found no statistically significant difference. In another study, the adenoma detection rate was 47.1% versus 37.3%, again without statistical significance. A microsimulation study projected a minimal reduction in the 10-year incidence of colorectal cancer—from 82 to 71 cases per 10,000 people—with two fewer deaths per 10,000 (0.02%), alongside a 6.4% increase in follow-up procedures.

The conclusion is nuanced: AI improves the detection of small, non-advanced lesions, without a clear improvement in advanced lesions. Traditional techniques remain central—narrow-band imaging provides twice the likelihood of detecting an adenoma.

Sixth: New Drugs? Not on This Horizon

Around 90% of potential drug candidates fail in clinical trials. Bringing a new drug to market takes an average of 10–15 years and costs between $1.5 and $2.0 billion.

AI is already helping predict protein structures, perform virtual screening, and recruit patients for clinical trials. But the greatest impact depends on something different—changing the rate of clinical success, in other words, improving the selection of which compounds should be developed further. That cannot be achieved within two years.

Seventh: The Workforce. Reorganization, Not Job Cuts

The argument that “AI will reduce the need for specialists” is not supported by the available measurements. A study of diagnostic radiologists’ workload found that most current applications require additional time for post-processing rather than taking tasks off their hands.

The change, however, is real: the specialist’s role will increasingly depend less on image interpretation itself and more on how effectively they manage, validate, and clinically contextualize automated systems.

Historical experience is reassuring: automation typically does not eliminate jobs but reshapes roles—people shift toward tasks that require a human element.

Eighth: Regulation. The Most Boring and the Most Important

This is where it is decided what will actually reach the patient.

The European Union’s AI Act entered into force on August 1, 2024, and is being implemented in stages over 6 to 36 months—with obligations for digital health tools taking effect after 24 months and for high-risk systems after 36 months. At the same time, a proposal has been submitted to postpone these obligations after August 2026 until harmonized standards become available.

Meanwhile, the capacity of notified bodies is limited. Their number has fallen from 58 to fewer than 40, while average certification timelines under the Medical Devices Regulation have already reached 19.5–21.8 months. The consequences are measurable: 47.6% of large manufacturers report reduced innovation projects in Europe, while 54% of surveyed industry professionals say they are planning to withdraw devices from the European market.

The practical conclusion: over the next two to three years, clinical practice will be changed primarily by already approved devices and by unregulated administrative applications, rather than by a wave of new high-risk systems.

Ninth: Bulgaria’s Starting Point

Any forecast for Bulgaria should begin with measurement, and the relevant data are documented.

Only 39% of Bulgarian hospitals have electronic medical records shared across all clinical departments, compared with an EU27+3 average of 57%. The national e-health profile is classified as below average.

The reasons have been identified: regulatory barriers, lack of infrastructure, financial constraints, and cultural factors. A study covering 14 hospitals found largely negative or lukewarm attitudes—probably as a result of a large number of failed projects and unmet expectations.

And here is the most important Bulgarian finding, which is positive in substance. An earlier study of computer literacy found that only 21.8% of physicians and 25.8% of nurses considered the country ready for e-health, but 90.4% of physicians and 82% of nurses acknowledged the need for electronic health records, while 46.1% of physicians and 22.2% of nurses said they would use such a system if adequate software were available.

The resistance is not to the technology. The resistance is to the conditions under which it has been introduced so far.

What to Do If You Work in a Healthcare Institution

Six things that do not depend on decisions in Brussels:

1. Create a register of the AI tools in use. What each tool does, what it is used for, who the provider is, when it was validated, and who is responsible for it.
2. Validate locally before implementation. Test performance on the local population, not just on the vendor’s promotional data.
3. Document human oversight. Record who reviewed the system’s output and accepted or rejected it.
4. Monitor performance after implementation. Track false alarms, data drift, and real-world outcomes.
5. Train staff on the systems’ limitations, not just their features.
6. Build the economic case in advance—including infrastructure, integration, training, and maintenance. Most analyses omit these costs and therefore cannot serve as a basis for planning.

Conclusion

By 2029, artificial intelligence will most visibly transform the administrative and diagnostic periphery of medicine rather than the core of clinical decision-making. The benefits emerge in assisted, rather than autonomous, modes. The workforce will not shrink—it will be reorganized. And the most important decisions will not be technological, but regulatory and economic.

If no action is taken, the quality gap between institutions will widen—and the reason will not be a lack of access to the technology."

Dr. Yavor Vladimirov, MD (Medical Director of the In Vitro Clinic “Sofia”)

“In the next 2–3 years, artificial intelligence will become part of the daily work of the doctor, just like many other technologies, without which today it is hard to imagine medicine.

The greatest benefit will likely be in processing the enormous amount of information we work with every day - medical documentation, laboratory and imaging studies, scientific literature. This can save a lot of time for the doctor and, I hope, leave them more time for the patient.

However, I do not think that AI can replace clinical thinking. Medicine is rarely black and white. Two patients with the same diagnosis may require different approaches, and the result that an algorithm offers is not always the right one for the specific person.

Therefore, in my opinion, the role of the doctor will even become more responsible. It will not be enough just to have access to a vast amount of information. You will need to be able to assess which of it is true, which is applicable to the specific patient, and when the technology is wrong.

And there is something else we must not lose. The patient does not come to us just for a diagnosis or test result. They need someone to explain what is happening, to understand their concerns, and to make a decision together with them. This is difficult to reduce to an algorithm.

Therefore, I do not expect AI to make the doctor less necessary. Rather, it will change the way we work. A good doctor will need to use the new technologies but also to know when not to trust them. And the responsibility for the decision must ultimately remain theirs.”