Vol 4, No 1S (2023)

BACKGROUND: One of the current issues of fundamental and clinical medicine is the early diagnosis of pathological changes that may lead to the development of cardiovascular disease.

AIM: The study aimed to investigate age-related changes in the cardiovascular system in healthy individuals by a comprehensive assessment of the correlations between hemodynamic and microcirculatory parameters at different age.

METHODS: The study was conducted on 136 adult volunteers with no clinical or laboratory manifestations of cardiovascular disease. The material was divided into five age groups: adolescence, I and II periods of adulthood, old age, and senile age. Males and females accounted for 50.73% and 49.27%, respectively. Central and peripheral hemodynamic parameters were measured using high-resolution compression oscillometry, ultrasound duplex scanning of aortic arch branches, echocardiography, and reactive hyperemia test on the brachial artery. Microcirculation indices in the skin of the right index finger were studied using a laser Doppler flowmeter before and after the treadmill test. The digital data were processed by the methods of variation statistics using the JASP 0.16.4.0 statistical software package.

RESULTS: An age-related decrease in skin microcirculation indices was observed. A significant statistical dependence of changes in microcirculation indices after physical activity in the groups of old and senile age was found. Left ventricular ejection fraction, cardiac output, cardiac index, stroke volume, and stroke index progressively decreased with increasing age; diastolic blood pressure increased; pliability of the brachial artery vascular wall decreased; total and specific peripheral resistance increased. Blood flow indices in the common carotid arteries (diameter and linear and volumetric blood flow velocity) decreased, whereas resistance index and intima-media thickness increased. Blood flow indices in vertebral arteries (linear and volumetric blood flow velocity) and indices of total blood flow to the brain progressively decreased with age. In addition, the dependence of some indices on gender was revealed.

CONCLUSIONS: With increasing age in healthy people, changes in central and peripheral hemodynamic parameters occur, which do not lead to cardiovascular diseases. Age-related changes in macro- and microcirculation are best detected during functional stress tests and should be considered in clinical practice as a manifestation of the normal aging process.

Inflammatory bowel diseases (IBDs) represent a group of nosologies including Crohn’s disease and ulcerative colitis. These diseases lead both to the deterioration of the patient’s physical condition and to a pronounced decrease in the quality of life. Currently, healthcare costs of managing patients with IBDs are steadily increasing and are commonly associated with the need for examinations and therapy with expensive drugs and unplanned expenses of healthcare resources for unscheduled consultations, emergency hospitalizations, and surgical interventions. Deterioration of patients’ condition may be associated with self-cancellation of therapy after normalization of the state of health. The use of telemedical technologies (TMTs) is promising for maintaining remission of IBDs, improving adherence to therapy, and reducing the frequency of unscheduled visits to the clinic and the cost of treating IBD complications (including surgery). The efficiency and benefits of TMTs in patients with IBDs were assessed in comparison with standard face-to-face follow-up. A literature search relevant to the review was conducted in the PubMed database. Papers were selected that assessed the efficiency of telemedical interventions among patients with IBDs as compared to the standard face-to-face follow-up group. The evaluation criteria included the degree of the disease activity, risks of IBD recurrence, quality of life of patients, adherence to therapy, healthcare costs, and frequency of visits to healthcare facilities. Telemedical interventions were conducted using the Internet, mobile applications, various telemedical platforms, SMS messages, and phone calls. The telemedical follow-up groups received access to educational materials and online consultations with a specialist upon request and completed questionnaires to assess disease activity, adherence to therapy, and quality of life. The degree of IBD activity and risk of recurrence and the level of adherence to therapy did not differ significantly between the telemedical intervention and standard face-to-face follow-up groups. The study by L. Pang et al. showed that the patients’ quality of life in the group of telemedical intervention was significantly higher (p=0.03) [8]. Another advantage of using TMTs is the provision of access to educational materials about IBDs and the possibility of remote contact with the treating physician. The patient’s ability to obtain necessary information reduces the need for unplanned visits to medical facilities and the cost of organizing medical care for patients with IBDs. The use of TMTs allows to monitor IBD activity as effectively as face-to-face follow-up, improve the patients’ quality of life with IBDs, and reduce costs for organizing care for patients with these diseases by decreasing the frequency of unscheduled visits to the doctor and conducting unnecessary additional examinations.

BACKGROUND: According to the World Health Organization, global population aging is the main trend in the development of the society on the planet. Russia ranks 116^th worldwide in terms of life expectancy. The development and implementation of measures aimed at preserving the able-bodied population of the country and slowing down the current trend of aging of the population with a focus on increasing life expectancy is manifested in national projects such as Health and Demography. Health-saving technologies for prolonging effective professional longevity of medical workers of older age groups will preserve a labor resource of the country and exclude economic losses of the state.

AIM: The study aimed to determine the significance of individual risk factors affecting the professional effective longevity of physicians and form a set of measures to prolong the professional competence of physicians of older age groups.

METHODS: In this study, the unit of observation is a physician providing medical care in the outpatient segment of Moscow and the Moscow region. The total number of subjects was 378. Sociological, analytical, and statistical methods and the method of organizational modeling were used. The main method was a face-to-face survey that included questionnaire data, questions on training, seniority, category, workload, and information assurance, and two blocks of questions to assess professional competencies and cognitive functions.

RESULTS:

1. An age-related decrease in the level of the physician’s professional competence.
2. An age-related decrease in the effectiveness of the prescribed treatment.
3. A tendency toward an age-related increase in the proportion of cognitive disorders among physicians.
4. Data analysis showing the need to implement a set of measures aimed at supporting the professional effective longevity of physicians of older age groups.

CONCLUSIONS: The organizational technology represented by a complex of measures providing prolongation of professional effective longevity of physicians of older age groups was proposed. Essentially, the organizational technology is the prevention of chronic non-infectious diseases in conjunction with measures to optimize work activities and stimulate the professional competence of the physician. Therapeutic Learning Technology is a parallel annual dispensary examination of physicians of older age groups followed by a therapeutic and rehabilitation course based on the morbidity profile and combined with a cycle of continuing medical education. The therapeutic and rehabilitative course should include older physicians undergoing cognitive training and hemodynamic stabilization. Information support (a block of federally approved training materials, i.e. normative legal acts regulating medical activity, clinical guidelines, video manuals on schools for different categories of patients, and other materials) is placed at the geriatric contact center, providing each physician from the target group with prompt access both for the period of treatment and rehabilitation course and the entire period of the physician’s medical activity.

Active deployment of artificial intelligence (AI) systems in medicine creates many challenges. Recently, the concept of responsible artificial intelligence (RAI) was widely discussed, which is aimed at solving the inevitable ethical, legal, and social problems. The scientific literature was analyzed and the possibility of applying the RAI concept to overcome the existing AI problems in medicine was considered. Studies of possible AI applications in medicine showed that current algorithms are unable to meet the basic enduring needs of society, particularly, fairness, transparency, and reliability. The RAI concept based on three principles — accountability for AI activities and responsibility and transparency of findings (ART) — was proposed to address ethical issues. Further evolution, without the development and application of the ART concept, turns dangerous and impossible the use of AI in such areas as medicine and public administration. The requirements for accountability and transparency of conclusions are based on the identified epistemological (erroneous, non-transparent, and incomplete conclusions) and regulatory (data confidentiality and discrimination of certain groups) problems of using AI in digital medicine [2]. Epistemological errors committed by AI are not limited to omissions related to the volume and representativeness of the original databases analyzed. In addition, these include the well-known “black box” problem, i.e. the inability to “look” into the process of forming AI outputs when processing input data. Along with epistemological errors, normative problems inevitably arise, including patient confidentiality and discrimination of some social groups due to the refusal of some patients to provide medical data for training algorithms and as part of the analyzed databases, which will lead to inaccurate AI conclusions in cases of certain gender, race, and age. Importantly, the methodology of the AI data analysis depends on the program code set by the programmer, whose epistemological and logical errors are projected onto the AI. Hence the problem of determining responsibility in the case of erroneous conclusions, i.e. its distribution between the program itself, the developer, and the executor. Numerous professional associations design ethical standards for developers and a statutory framework to regulate responsibility between the links described. However, the state must play the greatest role in the development and approval of such legislation. The use of AI in medicine, despite its advantages, is accompanied by many ethical, legal, and social challenges. The development of RAI has the potential both to solve these challenges and to further the active and secure deployment of AI systems in digital medicine and healthcare.

Due to improvements in the medical techniques used, virtual reality technologies are increasingly used to treat various diseases. Simultaneously, various ethical issues emerged. Knowing exactly what problems the treating physician and the patient may be facing is important to increase the effectiveness of therapy. This will help minimize or at least prevent negative consequences. A qualitative secondary study was conducted on the ethical issues arising from the application of virtual reality technologies in treatment. Papers suitable for analysis were searched using the keywords “virtual reality”, “ethical issues”, and “medicine”. The works of the last 10 years were selected, most of the sources being at most 5 years old. After analyzing papers and studies which describe the experience of applying virtual reality technologies in medicine, the range of ethical problems that are currently relevant was determined most accurately. Firstly, the vulnerability of patients is evident, which arises from the realism of the avatars and the deep immersion in the virtual world. Without the physician’s supervision, the patient may experience a high degree of psychological pressure, which has a negative effect. Secondly, strong emotional involvement is possible, which may develop attachment to the artificial intelligence and cause derealization, depersonalization, and other mental disorders. Thirdly, active interaction with the virtual world may be accompanied by long-lasting effects, such as gaming disorder. Finally, patients may have difficulties in distinguishing between reality and simulation, which affects autonomous choice and compromises informed consent. Virtual reality technologies may potentially change the lives of trauma patients and improve their quality of life. However, these changes are accompanied by ethical issues that must be addressed with respect to the vulnerability and nature of trauma.

BACKGROUND: Facial neuropathy at the peripheral level (unilateral muscular weakness of the entire half of the face) is a common neurological disorder. Assessment of the facial nerve dysfunction grade is necessary to track the dynamics of treatment and monitor the effectiveness of rehabilitation. For this purpose, worldwide clinical practice uses grading systems, the most popular of which are the House-Brackmann, Yanagihara, and Nottingham scales. Such methods are non-universal and based on visual diagnosis, which relies solely on the subjective experience of the physician. Consequently, objective measurements and automation are needed to track the dynamics of recovery. With the use of image processing and computer vision techniques, this task became feasible.

AIM: To develop a method of automated assessment of the facial nerve dysfunction grade by biometric facial analysis to monitor the patient’s recovery dynamics.

METHODS: As part of the collaboration with the Herzen Moscow Research Institute of Oncology, a database of target group patients with grades IV (4 people), V (4 people), and VI (11 people) of facial nerve disfunction according to the Haus-Brackmann scale was compiled. A control group consisted of 20 students from the Bauman Moscow State Technical University. During registration, subjects were asked to perform a series of following mimic tests: raising the eyebrows, closing the eyes, smiling, smiling with effort, inflating the cheeks, pouting the lips, and articulating with effort. Control points of the eyebrow, eye, and mouth areas were used to assess the degree of facial asymmetry. The two-dimensional MultiPIE model, implemented in the dlib library and containing 68 control points, was used as a facial model. A program code in Python was written, which calculates asymmetry coefficients based on changes in the coordinates of control points when the patient performs mimic tests.

RESULTS: A study was conducted to determine statistically significant differences between asymmetry coefficients in the control group and patients. Based on the Mann-Whitney criterion, asymmetry parameters during some mimic tests showed statistically significant differences (p <0.05). Thus, asymmetries in the forehead when raising the eyebrows (0.00 < 0.05), in the mouth when smiling (0.026 < 0.05), in the mouth when smiling with effort (0.00 < 0.05), in the mouth when pouting the lips (0.039 < 0.05), and in the mouth when articulating with effort (0.004 < 0.05) were revealed.

CONCLUSIONS: The results prove the performance of the proposed method and show the need for additional research, particularly the search for differences between groups of patients with different severity and the development of a classification model for machine learning.

BACKGROUND: Patients who, due to great distance or for other reasons, are unable to visit the medical center, must be provided with high quality and accessible medical care. Quality assurance and improvement of two-way communication between the patient and the medical staff must be ensured. The work is conducted at https://nefrocentr.ru.

AIM: The study aims at improving the efficiency and quality of medical care for patients of nephrological profile, i.e. patients with chronic kidney disease (CKD) of grade IIIb–V (pre-dialysis stages), patients undergoing renal replacement therapy (hemodialysis and peritoneal dialysis), and renal transplant recipients to reduce emergency hospitalizations and mortality and improve overall survival, as well as developing and implementing telerehabilitation and improving quality of life.

METHODS: The Telenephrocenter platform was created with the support of the Moscow Health Department, the Moscow Center for Innovative Technologies, and the Moscow Social Development Complex (Grant project No. 25-12-1/22 for the creation of a digital ecosystem for the Moscow Nephrology Cluster to promote alternative models of home nephrology care for patients with grades IIIb-IV-V CKD and kidney donor recipients).

• Application and improvement of existing online platforms for remote monitoring: video consultations, use of questionnaires and health assessment scales, therapy correction, dialysis program, and group telerehabilitation.
• Creation of educational video tutorials for patients to improve literacy with the further formation of a library of knowledge to teach the rules of life, nutrition, self-control, physical activity, hygiene, therapy, and peritoneal dialysis procedures.
• Telepatronage and follow-up of patients on peritoneal dialysis.

RESULTS: The system is a multidimensional information base that allows collecting data on the patient’s health status outside of the medical facility. Over 150 patients after autologous kidney transplantation, 80 patients on peritoneal dialysis, 25 patients on hemodialysis, and 23 patients on pre-dialysis CKD stages are under remote monitoring. The system enables the doctor to adjust the prescription of drugs, provide an urgent response to patients’ problems, and select the correct algorithm for behavior in a critical situation. Monitoring is based on questionnaires to which the patient answers according to a schedule set by the doctor. The questionnaire contains a set of symptoms that the patient chooses based on the intensity of the color. The aggregate of the responses forms a “heat map” of health status, which, when viewed remotely, allows the doctor to determine whether the patient has a problem and what kind of problem he or she has. Thus, the doctor makes informed decisions about further treatment. The resource allows patients to review a wide range of information material. Online consultations for patients, i.e. tele-educational activities for groups of nephrological patients depending on the type of pathology with the development of video tutorials from doctors on major problems and side-effects, are conducted. The effectiveness of the ongoing therapy (drug therapy and renal replacement therapy) is assessed remotely to adjust, cancel, and select the drug therapy. Weekly classes on physical rehabilitation and dietary support and nutrition are offered in an online format.

CONCLUSIONS: The creation of the digital ecosystem of the Moscow nephrology cluster to provide telesupport to all nephrology patients plays an important role in everyday life. The ability to get quality timely care allows for quality control of patients. Questionnaires, including assessment of the psychological state of patients, provide early detection of complications after correction of immunosuppressive therapy and increase patients’ adherence to therapy. Conducting online patient schools on dietary and physical activity modification helps to reduce anxiety and improve mood, increase physical activity, form good eating habits, and adjust dietary intake. These conclusions are based on weekly monitoring of patient diaries.

BACKGROUND: According to various data, orofacial diseases occur in the population in 20%–85% of cases. The most common cause of maxillofacial pain unrelated to the dental system is temporomandibular joint (TMJ) dysfunction, which is defined as a disruption of the anatomical relationship during natural movements. Magnetic resonance imaging (MRI) is the method of choice for diagnosing joint pathology.

AIM: MRI with static, pseudodynamic, and dynamic protocols were optimized for targeted diagnosis of TMJ dysfunction.

METHODS: Classical examination of the TMJ is performed in a closed and open mouth position. However, the images obtained do not reflect the location of the intra-articular disc at all stages of mandibular motion.

RESULTS: After a static examination in two oral positions, pseudodynamic sagittal T1 weighted imaging (WI) scans were obtained. The patient was asked to divide the movements from full closure to full mouth opening into five steps. The examination was conducted with the MRI operator and the patient acting together. The scans assessed the position and shape of the meniscus at five stages of mandibular movements. The performed program is leading in the assessment of the disc shape changes at each stage of the mouth opening. Wide coverage of the studied area allows to visualize both articular relations of internal structures and surrounding soft tissues, including contraction of the lateral wing muscle and superficial and deep parts of the masseter muscle. The final stage was a dynamic T2 WI scan performed by the patient individually. The obtained series of images represents sequential movements of the condyle, maximally approximating the natural mouth opening. The pulse sequence helps to assess the amplitude of motion of the mandibular condyle and hypermobility, which can be difficult when performing static images with standard mouth expanders.

CONCLUSIONS: An optimized dynamic and pseudodynamic TMJ study protocol is an essential part of the precise diagnosis of TMJ dysfunction and allows for differential diagnosis between muscle spasm and intra-articular disc adhesion and reliable visualization of intra-articular relationships during mouth opening.

BACKGROUND: Endometriosis is one of the most common causes of infertility, affecting approximately 6%–10% of women of reproductive age. For many decades, laparoscopic surgery has been considered the “gold standard” for diagnosing various forms of endometriosis. However, diagnostic laparoscopy, despite its widespread use, is an invasive and expensive procedure with certain risks. Current scientific literature increasingly shows that the methods of radiological diagnosis (ultrasound and magnetic resonance imaging [MRI]) are the main and most promising for verification of endometriosis. The advantages of MRI over ultrasound are multiplanar images, high tissue contrast, and smaller size of detectable heterotopias.

AIM: To assessing the sensitivity and specificity of MRI for diagnosing different forms of endometriosis.

METHODS: A retrospective analysis of the medical histories and instrumental studies of 129 women of reproductive age (mean age of the participants was 30.5±4.6 years) with a clinical diagnosis of infertility and suspected genital endometriosis was conducted. At the first stage of the study, the patients underwent a comprehensive MRI with one-stage magnetic resonance hysterosalpingography to assess the pelvis and tubal patency. Further, laparoscopic surgery (LS) was performed to confirm/refute the diagnosis of genital endometriosis, provide surgical treatment, or further search for possible causes of infertility. The results obtained with MRI and LS were compared, and the sensitivity and specificity of MRI for diagnosing external and internal genital endometriosis were assessed.

RESULTS: According to the data obtained, the specificity of MRI for diagnosing peritoneal endometriosis was 74% (95% confidence interval [CI], 57%–85%), and the sensitivity was 94% (95% CI, 87%–99%). Diagnostic accuracy of the technique was 86% (95% CI, 79%–91%). The most frequent localization of endometrioid heterotopias was retrocervical (41%). The specificity of this technique for diagnosing endometrioid ovarian cysts was 92% (95% CI, 82%–97%), and the sensitivity amounted for 98% (95% CI, 90%–100%). Diagnostic accuracy of the technique was 94% (95% CI, 88%–96%). The specificity and sensitivity of MRI for the diagnosis of adenomyosis were 96% (95% CI, 92%–99%) and 99% (95% CI, 87%–100%), respectively. Diagnostic accuracy of the technique was 97% (95% CI, 93%–99%). Our results were consistent with the literature data on the accuracy of MRI for diagnosing endometriosis.

CONCLUSIONS: Based on the results, MRI is a promising non-invasive method for diagnosing various forms of endometriosis. The performance of a comprehensive MRI allows obtaining sufficiently accurate information about the condition of the pelvic organs, detecting the manifestations of genital endometriosis, and reducing the time for examining women with infertility.

BACKGROUND: Artificial Intelligence (AI) is actively implemented in medicine. Since medical students are future physicians, an assessment of their awareness and attitudes toward AI is important.

AIM: To compare the degree of awareness and attitudes toward AI among Russian-speaking students of the Orenburg State Medical University (OrSMU) from the Russian Federation and English-speaking students from the Republic of India.

METHODS: From March 12 to 25, 2023, a voluntary anonymous survey (28 questions) was offered to OrSMU students using the Google Forms platform. An English-language version was prepared for foreign students. All responses were analyzed statistically (calculation of mean values using Likert scale, Student’s t-test, and Pearson’s chi-square test).

RESULTS: A total of 331 students participated in the survey, including 214 Russian-speaking and 117 English-speaking participants (127 males, 202 females, and two did not indicate gender). All participants were divided into 2 subgroups: junior (1–3 years, 200 participants) and senior year (4–6 years, 131 participants) students. The vast majority of respondents (92.3%) knew what AI is, with a higher percentage (p<0.001) among Russian-speaking students (95.8%) versus English-speaking students (84.6%). Only 34.1% of Russian-speaking and 46.2% of English-speaking students (p=0.032) were aware of the possibility of using AI in medicine. A total of 28.5% and 23.4% of Russian-speaking and 44.4% and 38.5% of English-speaking respondents, respectively, were aware of the use of AI in diagnostic radiology and pathological anatomy (p=0.004). Students expressed the greatest agreement with the statement that AI will play a significant role in the development and support of medicine in the future (mean Likert scale value of 4.23). Students were the least likely to agree with the statement that AI’s diagnostic abilities are superior to the human physician’s clinical experience (mean of 2.84). In the case of a disagreement between the AI and the physician, 76.7% of respondents would trust the latter to make the final decision. Most respondents considered diagnostic radiology, electrocardiogram analysis, and pathological anatomy to be promising areas for the use of AI (91.3%, 71.3%, and 70.4%, respectively). For the rest of the statements about attitudes toward AI, the average values ranged from 3.63 to 4.33. Among the disadvantages of using AI, the threat of data leakage was reported. The advantages were quick data analysis and assistance in diagnosis.

CONCLUSIONS: English-speaking students were more aware of the use of AI in medicine, whereas students from Russia showed a more positive attitude toward AI. However, in the case of a disagreement between the physician and the AI, both groups of respondents would trust the physician to make the decision.

Artificial intelligence (AI) systems are highly efficient. However, their implementation in medical practice is accompanied by a range of ethical issues. The “black box” problem is basic to the AI philosophy, although having its own specificity in relation to medicine. A selection of relevant papers for the last three years by citations and their analysis through PubMed and Google Scholar search engines was conducted to study the problems of the AI implementation in medicine. One of the central problems is that the algorithms to justify decisions are still unclear to doctors and patients. The lack of clear and reasonable principles of AI operation is called the “black box” problem. How can doctors rely on AI findings without enough data to explain a particular decision? Who will be responsible for the final decision in case of an adverse outcome (death or serious injury)? In routine practice, medical decisions are based on an integrative approach (understanding of pathophysiology and biochemistry and interpretation of past findings), clinical trials and cohort studies. AI may be used to build a plan for disease diagnosis and treatment, while not providing a convincing justification for specific decisions. This creates a “black box”, since the information that the AI considers important for making a conclusion is not always clear, nor is it clear how or why the AI reaches that conclusion. Thus, Juan M. Durán writes, “Even if we claim to understand the principles underlying AI annotation and training, it is still difficult and often even impossible to understand the inner workings of such systems. The doctor can interpret or verify the results of these algorithms, but cannot explain how the algorithm arrived at its recommendations or diagnosis”. Currently, AI models are trained to recognize microscopic adenomas and polyps in the colon. However, doctors still have insufficient understanding of how AI differentiates between different types of polyps despite the high accuracy, and the signs that are key to making an AI diagnosis remain unclear to experienced endoscopists. Another example is the biomarkers of colorectal cancer recognized by AI. The doctor does not know how algorithms determine the quantitative and qualitative criteria of detectable biomarkers to formulate a final diagnosis in each individual case, i.e., a “black box” of process pathology emerges. For the trust of doctors and patients to be earned, the processes underlying the work of AI must be deciphered and explained, describing how it is done sequentially, step by step, and a specific result is to be formulated. Although the “black box” algorithms cannot be called transparent, the possibility of applying these technologies in practical medicine is worth considering. Despite the above problems, the accuracy and efficiency of solutions does not allow to neglect the use of AI. On the contrary, this use is necessary. Emerging problems should serve as a basis for training and educating doctors to work with AI, expanding the scope of application and developing new diagnostic techniques.

BACKGROUND: Speech recognition is becoming increasingly common in the national healthcare system. One of the first specialties to implement this technology on a large scale was radiology. However, the efficiency of voice input and its effect on the length of time required to complete medical records remain unresolved.

AIM: To assess the efficiency of speech recognition in generating radiological protocols of different modalities and types.

METHODS: The retrospective study was conducted at the Moscow Reference Center of the Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies of the Moscow Department of Health. A total of 12,912 radiological reports on fluorography, mammography, chest computed tomography (CT), contrast-enhanced magnetic resonance imaging (MRI) of the brain, and contrast-enhanced CT of the abdomen and pelvis were included in the study by simple random sampling. The size of all samples exceeded 766 reports, calculated with regard to the size of the general population of over 100,000 reports. The Voice2Med software was used to fill in the radiological protocols. Intergroup comparison was performed using the Mann–Whitney U-test with a statistical significance level of 0.05.

RESULTS: The average duration of generating fluorographic protocols in the keyboard and voice input groups was 189.9 s (0:03:09) and 236.2 s (0:03:56), respectively (p <0.0001). For mammographic reports, the duration was 387.1 s (0:06:27) and 444.8 s (0:07:24), respectively (p <0.0001). For radiographic reports, it amounted to 247.8 s (0:04:07) and 189.0 s (0: 03:09), respectively (p <0.0001), and for chest CT, it was 379.7 s (0:06:19) and 382.7 s (0:06:22), respectively (p=0.12). For MRI of the brain, the protocols were generated for 709.9 s (0:11:49) and 559.9 s (0: 09:19), respectively (p <0.0001), and for contrast-enhanced chest, abdominal, and pelvic CT scans, it took 2714.6 s (0:45:15) and 1778.4 s (0:29:38), respectively. Voice input slowed down the preparation time of mammographic and fluorographic protocols. This is due to the use of a structured electronic medical document in medical facilities to describe the results of the examinations. Speech recognition showed the greatest efficiency in generating MRI and CT protocols. Such reports contain a large number of pathological changes, both target and incidental findings, which requires a detailed description by the radiologist in the examination protocol.

CONCLUSIONS: Speech recognition in generating radiological protocols showed different efficiency depending on the modality and type of the radiological protocol filled in using the voice input system. This approach is optimal for describing CT and MRI scans.

BACKGROUND: Artificial intelligence (AI) is a system based on machine learning of neural networks. AI structure resembles nerve tissue, having the so called “neurons”, i.e. mathematical codes. A neural network has three levels, including the input layer (information enters the system), the hidden layer (multidimensional data is analyzed), and the output layer (the system generates conclusions). Current neural networks use a “perceptron”, i.e. a neuron consisting of a large number of interconnected input and hidden layers, which makes the system capable of self-learning and analysis of non-linear data and generalization and processing of incomplete information, including the method of projection onto latent structures.

AIM: To develope a program based on multivariate data analysis to determine the chewing efficiency at the stages of prosthetics.

METHODS: In 2016, the Department of Orthopedic Dentistry and Orthodontics developed and tested a program for determining chewing efficiency based on the analysis of digital occlusiograms obtained by scanning dental prints on a wax plate. The results were processed by mathematical methods of multivariate data analysis using the projection on latent structures (the partial least-squares [PLS-2] model), which allowed assessing the relationship between the value of chewing efficiency and the area and brightness characteristics of the areas corresponding to occlusal contacts. The program compared the measurement results with the reference occlusiograms in the database and gave a conclusion. The experiments yielded statistically significant results for the efficiency of the program compared to traditional chewing tests. Due to the relevance of implementing AI in orthopedic treatment, the decision was made to improve the training methods of the program to update the existing array of reference data. Starting in 2019, 24 occlusiograms were added to the previously received data with dental defects from 9 to 12 teeth. Using an expanded database, the program, when analyzing the occlusiogram of a new patient, allowed to consider changes in chewing efficiency obtained earlier with the Trezubov chewing test for various defects of the dentition and compare the reference and minimally achievable values of chewing efficiency. The program algorithm was verified by the researchers using the classical Trezubov chewing test. Chewing efficiency was measured as a percentage.

RESULTS: The obtained combinations of the digital algorithm parameters for assessing chewing efficiency resulted in increased accuracy in the range of 4%–6% compared to the traditional Rubinov and Ryakhovsky chewing tests.

CONCLUSIONS: A digital algorithm for assessing chewing efficiency allows for quick and accurate assessment without the use of time-consuming analog tests.

BACKGROUND: The technique that allows detecting various pathologies and clearly identifying the cause of death and other associated diseases without disturbing the integrity of the skin refers to postmortem radiological diagnostics and is called “virtopsy”. The number of autopsies decreases worldwide, which creates the need to use alternative non-invasive methods, including virtopsy. The main method of postmortem radiology is multislice computed tomography (CT), with its advantages including high sensitivity to bone pathology. Postmortem radiological diagnostics allow supplementing the classical autopsy. This multidisciplinary approach helps visualize a wide variety of pathologies and assist the pathologist in the diagnostic search.

AIM: To assess the possibility of applying radiological diagnostic methods in the pathological anatomical practice and features of non-invasive autopsy (virtopsy) and its differences from the classical autopsy.

METHODS: The archival material including macro preparations of two lower limbs with tumor-like masses of unclear genesis (the upper third of the femur and the calcaneal tuber) was used. The biomaterial was previously encased in a specially treated gelatin medium, which ensured its preservation and left parts of the CT scanner intact. Multislice CT was used for radiological assessment of the preparations. From the primary CT sequences obtained, scans with the best visualization in the bone and soft tissue windows were selected, and three-dimensional (3D) reconstruction was applied. Scans and 3D reconstructions of images of tumor-like masses of the heel and thigh bones were obtained. Interpretation of the borders of the mass, characteristics of morphology and possible source of growth of the neoplasm allowed assuming a malignant nature of the tumor, suggesting its bone origin. Further histologic examination in both cases confirmed the neoplasm (osteogenic sarcoma).

RESULTS: CT scans and 3D reconstructions of images of tumour-like masses in the heel and femur were obtained. The interpretation of the boundaries of the mass, the morphological characteristics and the possible source of growth of the neoplasm suggested a malignant nature of the tumour, suggesting a bony origin. Further histological evaluation in both cases confirmed the nature of the neoplasm (osteogenic sarcoma).

CONCLUSIONS: The use of postmortem radiological diagnostic elements provides a sufficiently accurate verification of the nature of some pathologies. However, the absence of tissue damage and disturbances of intact topographic characteristics play not the least role in the postmortem diagnosis. In contrast to autopsy, this technique allows to return to the original appearance of the structures examined and, if necessary, start the diagnostic search anew. Moreover, specialists from other regions can be remotely involved by sending the data of CT scanning.

Environmental problems have a tremendous impact on the entire world population, particularly on human health, which plays a leading role in individual well-being. Environmental pollution, according to some estimates, kills approximately 9 million people every year. The introduction of artificial intelligence (AI) systems in many areas has enormous potential in reducing human impact on the environment; however, such systems have negative effects. The potential of AI systems to improve healthcare is inextricably linked to the ethical challenges posed by the complexity of these systems and their impact on the lives and health of communities, patients, and staff. In addition to aspects that relate directly to the algorithms, data, and clinical application of AI systems, long-term risks exist that are not obvious at first glance. One of these risks is the negative impact of AI systems on the environment, which may harm human health indirectly. AI systems are more than software, having physical components that are necessary for their functioning, such as processors, memory, and sensors. The manufacture and the energy consumption of the components has a profound effect on the environment. One study showed that when a single AI algorithm is trained, carbon emissions may reach values corresponding to the total carbon emissions from five cars’ lifetime.

This study analyzes existing literature linking the development of AI systems, especially in healthcare, to their effects on the environment. The study is intended to complement the emerging AI Ethics Code for healthcare, specifically the principles of sustainability that will be included in this code.

The study concludes that the environmental impact of AI systems should be considered when formulating ethical standards for AI in healthcare. These standards must be considered during the development, testing, and application phases of AI systems. All the people involved in the creation and use of AI systems (developers, physicians, and regulators) must monitor the environmental impact and minimize the environmental consequences of such systems at all stages of their existence. This principle calls for minimizing negative impacts, improving the energy efficiency, and disposing physical components in strict compliance with current legislation. Moreover, the rapid development of AI systems and the ethical dilemmas require that solutions be proposed jointly and ethical standards be developed in a manner that is consistent and sensitive to emerging technologies.

BACKGROUND: One of the common problems in labeling medical images is inter-observer variability. The same image can be labeled differently by doctors. The main reasons are the human factor, differences in experience and qualifications, different “radiology schools”, poor image quality, and unclear instructions. The influence of some factors can be reduced by proper organization of the annotation; however, the opinion of doctors frequently differs.

AIM: The study aimed to test whether a neural network with an additional module can learn the style and labeling features of different radiologists and whether such modeling can improve the final metrics of object detection on radiological images.

METHODS: For training artificial intelligence systems in radiology, cross-labeling, i.e., annotation of the same image by several doctors, is frequently used. The easiest way is to use labeling from each doctor as an independent example when training the model. Some methods use different rules or algorithms to combine annotation before training. Finally, Guan et al. use separate classification heads to model the labeling style of different doctors. Unfortunately, this method is not suitable for more complex tasks, such as detecting objects on an image. For this analysis, a machine learning model designed to detect objects of different classes on mammographic scans was used. This model is a neural network based on Deformable DETR architecture. A dataset consisting of 7,756 mammographic breast scans and 12,543 unique annotations from 19 doctors was used to train the neural network. For validation and testing, a dataset consisting of 700 and 300 Bi-Rads-labeled scans, respectively, was taken. In all data sets, the proportion of images with pathology was in the 15%–20% range. A unique index was assigned to each of the 19 doctors, and a special module at each iteration of the neural network training found a vector corresponding to this index. The vector was expanded to the size of the feature map of each level of the feature pyramid, and then attached by separate channels to the maps. Thus, the encoder and the decoder of the detector had access to the information about which doctor labeled the scan. The vectors were updated using the back-propagation method. Three methods were chosen for comparison:

1. Basic model: Combining labels by different doctors using the “voting” method.
2. New stylistic module: For predictions on the test dataset, a single doctor’s index was taken, which showed the best metrics on the validation dataset.
3. New stylistic module: The indexes of the five doctors with the best metrics on the validation dataset were used for predictions on the test dataset. Weighted Boxes Fusion was chosen to combine the predictions.

The area under the receiver operating characteristic curve (ROC-AUC) was used as the primary metric on the test dataset (Bi-Rads 3, 4, and 5 categories were referred to pathology). The sum of maximum probabilities of detected malignant objects (malignant masses and calcinates) by cranio-caudal and medio-lateral oblique projections was assumed as the probability of malignancy for each method.

RESULTS: The following ROC-AUC metrics were obtained for the three methods: 0.82, 0.87, and 0.89.

CONCLUSIONS: The information about the labeling doctor allows the neural network to learn and model the labeling style of different doctors more effectively. In addition, this method may obtain an estimate of the uncertainty of the network’s prediction. The use of embedding from different doctors, leading to different predictions, may mean that this data is difficult for an artificial intelligence system to process.

INTRODUCTION: Infectious endocarditis (IE) in injecting drug patients with human immunodeficiency virus (HIV) is manifested predominantly by damage to the tricuspid valve (TV). The pathogen of the disease is Staphylococcus aureus. The IE peculiarities of the TV in this category of patients are multiple septic emboli in the small circulation, which may result from flotation vegetations, and predominance of pulmonary symptomatology over cardiac phenomena.

CASE PRESENTATION: A 30-year-old female HIV-infected injecting drug user was dynamically monitored for the laboratory-confirmed IE development by transthoracic echocardiography during a year. The patient was twice admitted to the intensive care unit in a serious condition with purulent-septic complications in the lungs in the form of bilateral polysegmental destructive (septic) pneumonia confirmed by X-ray and multispiral computed tomography. After intensive therapy, the patient refused further ARV treatment and consultations with a cardiac surgeon. A clinical follow-up demonstrated IE of the TV and right-sided lower lobe pneumonia in an HIV-infected drug-dependent patient after a cesarean section. Immunodeficiency without appropriate ARV therapy had a negative impact on the course of the disease. According to transthoracic echocardiography, an increase in TV vegetations from 7 mm to 16 mm and regurgitation were observed, whereas left ventricular ejection fraction decreased from 60% to 47% during the follow-up. Complications of cesarean section, immunodeficiency, and absence of ARV therapy could be the trigger factors of the IE development. Injection drug use with possible violations of asepsis and circulation of microorganisms of the causative agents in the blood contributed to the rapid development of the disease and was a determining factor of the infectious process in the TV.

CONCLUSIONS: Echocardiography remains the primary radiological diagnostic method for examining HIV-infected drug-dependent patients with fever and/or after medical manipulations. However, this group of patients more frequently clinically manifests the disease with bright pulmonary pathology. Therefore, radiological methods should necessarily be a priority during the diagnostic search. The sources of septic embolism in the small circulation in IE may be loose and floating TV vegetations.

BACKGROUND: Non-alcoholic fatty liver disease is a common chronic liver condition characterized by pathological accumulation of fatty droplets, which is not associated with alcohol consumption. The main reason in the development of complications and adverse outcomes of the disease is fibrotic changes in the liver, leading to cirrhosis.

AIM: To determ the informative value of two-dimensional shear wave elastography in hepatosis by a comparative analysis of morphological and ultrasound elastometric findings.

METHODS: The study included 40 patients aged 35–60 years. The control group consisted of 25 healthy people with a body mass index of 24–26. Among 40 patients, F1, F2 and F3 stages were detected in 15 (37.5%), 20 (50%), and 5 (12.5%) cases, respectively. The diagnosis was based on clinical, laboratory, and ultrasound examinations conducted with a Mindray Consona N9 device using a convex transducer at 2.0–3.5 MHz. In the absence of contraindications, all patients underwent puncture biopsy under ultrasound navigation in compliance with aseptic and antiseptic requirements.

RESULTS: At the first (F1) stage of hepatic fibrosis (n=15), the elastometry index averaged 5.4 (4.8–6.2) kPa, and the shear wave velocity was 1.1±0.2 m/s. Histomorphological examination in this group revealed mild fatty liver dystrophy, with liver tissue having varying degrees of dystrophic changes. In the second (F2) stage of fibrosis (n=20), the mean elastometry values were 6.9 (6.3–7.9) kPa, while the shear wave velocity increased to 1.5±0.2 m/s. In patients of this group, histological examination showed that moderate to large fatty dystrophy developed in hepatocytes. In the F3 stage (n=5), the elastometry index averaged 8.3 (8.1–13.5) kPa, and the shear wave velocity reached 1.7±0.2 m/s. In this case, hepatocytes with fatty dystrophy in the focus surrounded by a cluster of lipophages were determined. Coincidence of fibrosis stage determined by elastometry data with morphological examination of liver biopsy specimen was observed in 36 (90%) patients with fatty hepatosis.

CONCLUSIONS: Comparison of the results of two-dimensional shear wave elastography and morphological changes of liver parenchyma in hepatosis showed high (90%) correlation with the data of the above examinations. If puncture biopsy is contraindicated, shear wave elastography may serve as a valuable diagnostic method to clarify the stage of fibrosis in hepatosis.

Among many diseases of the joints, the most common is osteoarthritis and, in particular, knee osteoarthritis. Current medicine requires new and improved imaging techniques to assess patients with knee osteoarthritis, which enable early diagnosis, improved understanding of the pathophysiology of osteoarthritis, and the efficacy of the ongoing treatment. Scientific data on the use of magnetic resonance imaging (MRI) as an objective method to assess the impact of shock wave therapy (SWT) on the restoration of damaged cartilage and bone marrow in patients with knee osteoarthritis was analyzed. A search for publications in the PEDro and PubMed databases for the period from 2017 to 2022 was performed. The search was done using the following keywords: “osteoarthritis”/“knee osteoarthritis”, “shock wave therapy”, “magnetic resonance imaging”. Five foreign studies were included in the review as a result of the search (no Russian publications were available). Two studies analyzed the effects of SWT on altered cartilage in patients with knee osteoarthritis. Patients who received oral non-steroidal anti-inflammatory drugs were selected as a comparison group. The authors assessed SWT effects on cartilage using quantitative estimates of T2 relaxation time (T2 mapping) and the severity of bone marrow edema on MRI images. At 24 weeks after treatment, differences in relaxation time on T2-weighted MRI images of the patella, acetabulum, and medial and lateral articular surfaces of the femur and tibia showed no statistical significance compared to those before treatment (t= –1.859; P= 0.076). In conclusion, the authors suggested that the resulting effect of SWT in patients with knee osteoarthritis was due to the use of low energy and stated that research will continue using higher SWT parameters. Another study reported that patients with knee osteoarthritis showed a statistically significant reduction in the area of the subchondral bone marrow lesion 6 months after SWT. In addition, the authors found a correlation between the reduction in subchondral bone marrow lesion size and the improvement in pain scores resulting from the SWT treatment. In another study, MRI scans in patients with knee osteoarthritis documented a positive SWT effect by reducing bone marrow edema. After one year of follow-up, MRI images showed complete regression of bone marrow edema among patients who received SWT. The number of studies that investigated SWT effects on cartilage and bone marrow restoration in patients with knee osteoarthritis using MRI is limited, indicating the need for continued research in this area.

BACKGROUND: Aortic aneurysms, known as “silent killers”, are frequently asymptomatic, leading to vessel rupture and death. Annual rates for ruptures, stratifications, and sudden deaths are up to 3.6%, 3.7%, and 10.8%, respectively. Timely diagnosis and early treatment save a patient’s life. The use of artificial intelligence (AI) technologies helps to detect aortic aneurysms, which significantly improves the quality of diagnosis and saves patients’ lives.

AIM: To assess the efficiency of using AI technologies to detect thoracic aortic aneurysms on chest computed tomography (CT) and determining the possibility of using AI technologies as an assistant to the radiologist during the primary description of radiological images.

METHODS: The study retrospectively assessed the results of AI technologies aimed at detecting thoracic aortic aneurysms on chest CT scans. No contrast enhancement was performed primarily. The sample consisted of 84,405 observations of patients over the age of 18 years; of these, 86 scans with thoracic aortic aneurysms were selected according to AI data. The selected examinations were retrospectively assessed by radiologists and vascular surgeons for the probable presence of a thoracic aortic aneurysm. In 44 cases, an aortic aneurysm was initially detected by the radiologist. In 31 cases, an aneurysm was not initially described by the radiologist, 6 were excluded from the sample (due to the absence of the radiologist’s report in the Unified Radiological Information Service), and 5 scans had false-positive results according to AI findings.

RESULTS: The use of AI technologies allows detection and labeling of pathological changes in the aorta on the images. AI technologies increase the detectability of thoracic aortic aneurysms in the description of chest CT scans by 38.8%. The incidence of ascending aortic aneurysm was 0.3%, which corresponded to the literature data of 0.16%–1.6%. According to the results, 22 surgical interventions for aortic stenting were performed.

CONCLUSIONS: The use of AI in the primary chest CT description may help increase the detectability of clinically significant pathological conditions, such as thoracic aortic aneurysm. Further development of routing for this category of patients in the cito mode for surgical treatment is relevant. Expansion of retrospective screening by chest CT scans using AI systems will improve the quality of diagnosis of concomitant pathologies and prevent adverse outcomes for patients.

BACKGROUND: Traumatic peripheral nerve damage is a significant clinical and social problem, which is characterized by a high level of disability in young patients.

AIM: To assess the diagnostic efficiency of ultrasound in the diagnosis of peripheral nerve damage in combat trauma.

METHODS: A total of 163 patients (362 peripheral nerves) were examined. The duration of traumatic nerve damage was 2–273 days. All patients were men aged 20–48 years. Ultrasound was performed with 7–17 MHz linear transducers on an ACUSSON S2000 device using standard technique in B-mode, in longitudinal and transverse scanning planes, with the use of Doppler techniques. Statistical analysis was conducted to assess diagnostic efficiency. Calculation of operational (sensitivity and specificity) and integral (accuracy) characteristics was performed by the qualitative assessment of the reference method (surgical intervention) and the method under study (ultrasound).

RESULTS: Peripheral nerve damage resulted from combat trauma. In 120 (73.6%) cases, injuries of the limbs were accompanied by injuries of the osteoarticular apparatus and vessels. A total of 274 (75.7%) nerves had signs of traumatic damage. Multiple nerve injuries were observed in 95 (58.3%) people. Nerves of the upper extremities were damaged more frequently (185 [67.5%]) compared to nerves of the lower extremities (89 [32.5%]). Contusional structural changes were observed in 181 (66%) nerves. Impairment of anatomical integrity was diagnosed in 46 (16.8%) nerves, while complete impairment was found in 29 (10.6%) cases with the presence of diastasis between the nerve ends. Early after injury, a wound canal and hematomas were visualized near the nerve. In 4 cases, a foreign body of metallic density was visualized in the nerve sheath. After 3 weeks from the moment of injury, terminal neuromas were observed. The neuroma sizes for the proximal and distal ends were 0.5×0.3 to 1.6×0.6 cm and 0.4×0.2 to 1.3×0.6 cm (avascular), respectively. Adjacent sections of the nerve over 3–5 cm were thickened and characterized by echo structure, however, with thickening of all fasciculi and blood flow observed along the periphery of the nerve. Marginal nerve damage was observed in 17 (6.2%) people. In case of marginal neuroma, a significant nerve thickening of 1.4–3.2 times over a short distance (from 0.4 cm to 1.5 cm) with loss of bundle differentiation of the part of the nerve, pronounced decrease in echogenicity, and absence of Doppler signal was detected. Nerve compression in 47 cases (17.1%) was accompanied by 1.2–2.3 times thickening of nerve trunks, indistinct contours, decreased echogenicity, and significant changes in the nerve structure. Compression was due to cicatricial changes, hematomas, foreign bodies, bone fragments, and — in 2 cases — spokes from external fixation devices. A total of 106 patients underwent surgery.

CONCLUSIONS: The ultrasound sensitivity and specificity were 96% and 67%, respectively. The peculiarities of the examination included extensive soft tissue defects and external fixation devices, which significantly complicated the examination. The diagnostic accuracy was 91%. A pronounced cicatricial process (70%) was the main cause of false-positive (6.6%) and false-negative (2.8%) results.

BACKGROUND: The main reason for the development and implementation of artificial intelligence (AI) technologies in neuro-oncology is the high prevalence of brain tumors reaching up to 200 cases per 100,000 population. The incidence of a primary focus in the brain is 5%–10%; however, 60%–70% of those who die from malignant neoplasms have metastases in the brain. Magnetic resonance imaging (MRI) is the most common method for primary non-invasive diagnosis of brain tumors and monitoring disease progression. One of the challenges is the classification of tumor types and determination of clinical parameters (size and volume) for the conduct, diagnosis, and treatment procedures, including surgery.

AIM: To develope a software module for the differential diagnosis of brain neoplasms on MRI images.

METHODS: The software module is based on the developed Siberian Brain Tumor Dataset (SBT), which contains information on over 1000 neurosurgical patients with fully verified (histologically and immunohistochemically) postoperative diagnoses. The data for research and development was presented by the Federal Neurosurgical Center (Novosibirsk). The module uses two- and three-dimensional computer vision models with pre-processed MRI sequence data included in the following packages: pre-contrast T1-weighted image (WI), post-contrast T1-WI, T2-WI, and T2-WI with fluid-attenuated inversion-recovery technique. The models allow to detect and recognize with high accuracy 4 types of neoplasms, such as meningioma, neurinoma, glioblastoma, and astrocytoma, and segment and distinguish components and sizes: ET (tumor core absorbing Gd-containing contrast), TC (tumor core) = ET + Necr (necrosis) + NenTu, and WT (whole tumor) = TC + Ed (peritumoral edema).

RESULTS: The developed software module shows high segmentation results on SBT by Dice metric for ET 0.846, TC 0.867, WT 0.9174, Sens 0.881, and Spec 1.000 areas. The testing and validation were done at the international BraTS Challenge 2021 competition. The test dataset yielded DiceET 0.86588, DiceTC 0.86932, and DiceWT 0.921 values, placing the developed software module in the top ten. According to the classification, the results demonstrate high accuracy rates of up to 92% in patient analysis (up to 89% in slice analysis), a very high potential, and a perspective for future research in this area.

CONCLUSIONS: The developed software module may be used for training specialists and in clinical diagnostics.

Currently, artificial intelligence (AI) plays an important role in various fields as a primary worker or assistant, especially in healthcare. AI has many functions that can be performed even better than humans do due to the high speed of compiling a large amount of data from various sources (Internet and electronic health records), which increases the productivity of doctors. Less discussed, though equally important, is the discrimination of AI in healthcare. Doctors, who have a lot of responsibility, cannot rely on AI due to multiple vulnerabilities. First, AI collects huge amounts of data. However, AI does not guarantee error-free results, despite its non-autonomy and operator control, since human factors can play a role, becoming a source of inaccuracy. Therefore, a high risk of using poor quality data in further decision-making processes exists. Errors, biases (ethnic, gender, age, and social), and data gaps worsen AI outcomes, leading, for example, to discrimination against minorities and inaccurate prescriptions. In addition, AI may act unethically or even violate laws (e.g., Title VI, Section 1557 of the Affordable Care Act, which prohibits discrimination by race, color, national origin, sex, age, or disability in certain healthcare programs and activities).

Secondly, AI may be called secretive due to the indeterminacy of the algorithms. This means that no one can explain how or why the AI reached this or that decision, as stated by Judea Pearl in The Book of Why: The New Science of Cause and Effect. Therefore, doctors cannot check the facts and make sure that the analysis is done correctly and the conclusion is accurate. Difficulties in solving real-world problems using AI that cannot be solved by formal and mathematical rules of logic as humans do (e.g., natural language and face recognition) must be recognized. AI makes the work of medical professionals easier, however, posing many unresolved problems that mislead doctors and contribute to wrong decisions.

BACKGROUND: The medical market offers interesting solutions that use artificial intelligence (AI) technologies; however, such solutions often remain at the startup level or are used locally. The question is how to achieve the maximum when introducing AI systems into medical practice.

AIM: To answer the question of why the large number of existing developments in AI and medical decision support systems are not used as widely as medical information systems, telemedical consultations, and other health IT solutions. Possible ways to improve the efficiency of implementing AI technologies and medical decision support systems in the work of physicians were presented.

METHODS: Theoretical and general scientific (analysis of literature and Internet sources on the problem of research, synthesis, generalization, comparison, and systematization) and empirical (observation, interview, and testing) methods were used.

RESUTLS: The main barriers to the effective implementation of AI systems in medical practice and possible options to solve the following problems were highlighted.

• Problem 1: incorrect data collection. Solution: care must be taken with the accumulation of materials used for the analysis and training by AI systems in medicine.
• Problem 2: incompetent developers. Solutions: involvement of effective third-party specialists or training one’s own.
• Problem 3: medical workers’ and/or patients’ aversion to AI technologies. Solutions: education in the successful application of AI technologies and medical decision support systems in healthcare and involvement of practicing physicians as experts during the AI conceptualization.

CONCLUSIONS: If the above criteria for the development and use of AI technologies and medical decision support systems are met, the effect of their introduction into medical practice will tend to be maximized.