Modern capabilities of artificial intelligence technologies in cardiovascular imaging

Full Text

INTRODUCTION

Cardiovascular diseases (CVDs) are the leading cause of disability and mortality worldwide [1]. Their diagnosis and treatment are based on medical histories, physical examinations, laboratory tests, and invasive and noninvasive imaging. The emergence of new technologies and artificial intelligence (AI) and machine learning (ML) integration have broadened opportunities for doctors to improve the effectiveness of diagnostic and therapeutic measures [2]. Considerable data extracted from electronic medical records, mobile medical devices, and imaging results have driven the rapid development of AI algorithms in medicine. Cardiology is among the few medical specialties wherein AI technologies have been systematically investigated [3].

AI refers to systems that simulate human intelligence and are capable of learning to make decisions. The interest in this field has been growing since the 1950s, when Turing [4] posed the question whether computers can think. In 1955, McCarthy et al. [5] first introduced the term artificial intelligence. This concept supports the development of methods that enable computers to participate in processes such as learning and reasoning. Many methods of this technology are grounded in ML algorithms, which include processes that ensure the adaptation of parameters to solve specific tasks based on training data [6]. Notably, AI systems are capable of making the best possible decisions.

The exponential development of AI technologies, particularly in ML and deep learning (DL), is rapidly attracting the interest of clinicians in creating novel, integrated, reliable, and efficient diagnostic methods to provide medical care. DL is a significant advancement in addressing challenges in accumulating, processing, and differentiating big data. For many years, the medical community has been unable to solve the existing issues. However, DL has been proven highly effective in identifying complex patterns within high-dimensional data and can be applied across various scientific fields [7]. In addition, DL systems are trainable and capable of functioning based on raw data, such as numbers, text, and their combinations [8].

Cardiologists use various imaging-based diagnostic techniques, which provide more extensive quantitative patient data. Despite potential difficulties, the use of AI-associated methods is one of the most effective data-driven approaches to clinical decision-making. AI implementation requires close collaboration among medical researchers, mathematicians, clinicians, and healthcare management specialists. Currently, the effectiveness of AI technologies is being evaluated across multiple areas of cardiology: from decision support systems to imaging data processing and interpretation. Innovative ideas and practical AI-based methods for the diagnosis and treatment of CVDs are increasingly being proposed, providing new prospects for the development of cardiology [9].

Imaging techniques such as echocardiography (EchoCG), cardiac magnetic resonance imaging (MRI), and computed tomography (CT) are among the key diagnostic tools in cardiology. These provide detailed information on cardiac structure and function, enabling the identification of abnormalities and supporting the diagnosis of CVDs [10].

In recent years, AI technologies have been increasingly adopted in cardiology to improve the accuracy of CVD diagnosis and treatment. AI algorithms used for image analysis facilitate the rapid and accurate identification of disorders, whereas predictive models allow for analyzing large datasets to detect patterns and estimate the probability of specific outcomes. AI-based decision-support systems can assist clinicians in determining treatment strategies, and portable devices may enable early detection of CVDs by continuously monitoring heart function. The ongoing development and integration of AI into cardiology practice is expected to significantly enhance the diagnosis and management of CVDs and thus improve patients’ quality of life.

Despite its clear potential, the application of AI technologies in current clinical practice remains limited, underscoring the need for further research.

DATA SEARCH METHODOLOGY

A published data search was conducted using the search engines PubMed, Google Scholar, and eLibrary. Russian and English publications were considered. The search was performed using the following keywords in both Russian and English: искусственный интеллект и визуализация сердечно-сосудистой системы / artificial intelligence and cardiac imaging, визуализация сердечно-сосудистой системы / cardiac imaging, машинное обучение и сердечно-сосудистые заболевания (machine learning and cardiovascular diseases), глубокое обучение и сердечно-сосудистые заболевания (deep learning and cardiovascular diseases), machine learning and cardiac imaging, and deep learning and cardiac imaging. All articles were published between 1950 and 2024. When necessary, an additional search was conducted for other relevant publications related to the clinical and prognostic value of AI technologies and the challenges associated with their use. Titles and abstracts were independently screened, followed by full-text review of relevant articles. The following algorithm was applied for source selection: duplicates were excluded prior to the review; during the search process, article titles and abstracts were screened for relevance to the topic and availability of full text (excluding abstracts, articles, and publications without full-text access); and full-text manuscripts were then assessed for compliance with the inclusion criteria.

Finally, 70 publications were included in the present review. Fig. 1 illustrates the selection algorithm.

 

Fig. 1. Source selection algorithm.

 

ARTIFICIAL INTELLIGENCE IN CARDIOVASCULAR IMAGING

Cardiology is one of the medical specialties wherein ML and DL methods have been widely adopted and have demonstrated promising results [3]. AI encompasses various computational technologies that enables machines to perform tasks traditionally requiring human intelligence [11]. ML includes algorithms capable of learning from existing data and making data-driven decisions. In cardiovascular imaging, these technologies are used to interpret data obtained from various imaging modalities, including EchoCG, CT, MRI, and nuclear imaging [11].

DL, a subtype of ML, utilizes neural networks and has gained considerable popularity in medical imaging owing to its ability to automatically learn from existing data, analyze errors, and make more accurate predictions [12]. Convolutional neural networks (CNNs), a type of DL architecture, are especially suitable for image analysis. They consist of multiple layers that automatically extract relevant features from medical images, enabling effective classification, segmentation, and detection of various structures [13].

CNNs have been used to assess cardiac functional parameters based on EchoCG data [13]. DL algorithms have improved the accuracy of coronary artery stenosis, calcification identification, and plaque characterization using cardiac CT data [14]. Moreover, CNNs have been applied to automatic segmentation of cardiac structure images, tissue characterization, and perfusion analysis using MRI data [13].

Although DL has received substantial attention, other ML algorithms also play an important role in cardiovascular imaging. Support vector machines are used for image classification, tasks, and risk stratification [15]. The random forest method has shown high effectiveness in selecting relevant features and predicting cardiovascular events based on imaging-derived biomarkers [16]. Clustering algorithms, such as the k-means method, have been used to identify pathological patterns in medical images [17] Unsupervised and self-supervised learning approaches, including generative adversarial networks, have demonstrated considerable potential for image color correction, resolution enhancement, and synthetic data generation in cardiovascular imaging [18]. These may improve image quality, reduce radiation exposure, and help overcome data scarcity in the development of ML models.

AI technologies are actively applied for automatic segmentation and quality control of medical images. These algorithms can automatically delineate regions of interest, thereby facilitating data analysis and interpretation. Additionally, they are used in assessing myocardial contractility and coronary blood flow, enabling more accurate evaluation of the cardiovascular system [13].

DL methods have been employed to analyze vascular calcification using CT data. In the CONFIRM registry, 13,054 coronary computed tomography (CT) angiography scans were analyzed using ML algorithms that incorporated clinical variables and coronary artery calcium scores derived from coronary CT angiography. This approach enabled highly accurate assessment of the possibility of obstructive coronary artery disease (CAD) [19]. Studies have demonstrated that ML algorithms outperform visual assessment by radiologists in detecting obstructive forms of CAD using CT coronary angiography data [3].

Furthermore, ML algorithms based on radiomics have potential for identifying coronary lesions [20]. The effectiveness of detecting various features of atherosclerotic plaques, microcalcifications, and vascular inflammation using CT coronary angiography has been shown to be higher when using radiomics-based ML platforms than with visual interpretation by physicians [21, 22].

As AI technologies, particularly ML, continue to develop, their integration opens up new possibilities. Although these tools show considerable promise for improving diagnostic accuracy and efficiency, their implementation requires thorough validation and careful consideration of regulatory and ethical issues.

ROLE OF ARTIFICIAL INTELLIGENCE IN TRANSTHORACIC ECHOCARDIOGRAPHY

Considering the high prevalence of heart failure in the general population, the demand for EchoCG—a key modality for assessing cardiac function—is increasing [23]. A shortage of highly trained specialists contributes to delays in diagnosis and treatment, reducing patients’ quality of life [24]. AI technologies may play an important role in addressing issues related to variability in image acquisition and interpretation by healthcare professionals [25]. EchoCG allows assessing a chamber size, wall motion, valvular function, and left ventricular ejection fraction (LVEF). The use of ML has demonstrated accuracy in LVEF estimation comparable to that of physicians [26, 27]. Integration of AI technologies has shown promising results, reducing time for image acquisition and analysis of LVEF and left ventricular volume by up to 77% [28]. These advances may improve the efficiency and accuracy of EchoCG, ultimately enhancing quality of care amid the ongoing rise in CVD prevalence.

One of the earliest applications of AI technologies in EchoCG interpretation was the assessment of left ventricular volume and function. These tools are capable of enhancing image quality and facilitating subsequent analysis. In 2015, Knackstedt et al. [29] utilized an ML algorithm to evaluate LVEF and longitudinal strain. The mean analysis time per patient was 8 ± 1 seconds, with an accuracy of 92.1%. In 2016, Narula et al. [30] investigated the diagnostic value of ML for automated differentiation between hypertrophic cardiomyopathy and athlete’s heart using speckle-tracking EchoCG. In a cohort of 77 athletes and 62 patients with hypertrophic cardiomyopathy, the sensitivity and specificity for differential diagnosis were 87% and 82%, respectively.

Zhang et al. [31] trained a CNN model using 14,035 echocardiograms collected over 10 years to perform multiple tasks, including view classification, cardiac chamber segmentation, volume and mass estimation, LVEF calculation, and automated strain analysis by speckle tracking. The AI-derived measurements were comparable to manual assessments and in some cases exceeded them. Additionally, CNN models have been developed to support the diagnosis of hypertrophic cardiomyopathy, cardiac amyloidosis, and pulmonary hypertension, demonstrating high efficiency [31].

Notably, certain AI technologies may facilitate the assessment of valvular heart disease. For example, the use of a support vector machine for diagnosing mitral regurgitation demonstrated a 99.38% sensitivity and 99.63% specificity [32]. The accuracy for normal mitral valve function and mild, moderate, and severe mitral regurgitation was 99.52%, 99.38%, 99.31%, and 99.59%, respectively. Playford et al. [33] studied the diagnostic efficiency of AI for aortic stenosis. The authors analyzed data from 530,871 echocardiograms over a mean follow-up of 4.1 years, including 171,571 men and 158,404 women. It was found that AI was able to identify high-gradient aortic stenosis in 95.3% of cases and in 73.9% of cases with the use of the continuity equation. AI has demonstrated effectiveness in patients with preserved and impaired left ventricular systolic function.

A recent advancement in EchoCG interpretation is the development of a video-based DL algorithm that outperformed clinicians in the assessment of LVEF, cardiomyopathy patterns, and left ventricular segmentation [34]. The variability in results obtained using this algorithm was comparable to that observed among cardiologists, or even lower. This confirms the high accuracy and reliability of the algorithm. Additionally, the use of AI technologies has been associated with faster procedure times, which allows for workflow optimization and shorter patient wait times. These improvements contribute to enhanced quality of care. EchoNet-Dynamic is a DL algorithm that employs three-dimensional CNNs with an R2 + 1D architecture. The model processes echocardiographic video sequences composed of 32 frames. Initially, AI performs temporal analysis by segmenting the left ventricle based on its size, followed by spatial feature extraction. The authors used over 10,000 video files for training. This is among the most successful models that are capable of estimating LVEF and other parameters from echocardiographic videos, achieving performance comparable to that of experienced clinicians. The application of such technologies may improve measurement accuracy while reducing time expenditure, thereby enhancing the diagnostic evaluation of various CVDs [34].

CNN-based models demonstrate high accuracy in left ventricular segmentation, LVEF estimation, and ventricular myocardial deformation analysis [35]. These automated approaches help reduce time consumption and improve the quality of results. ML-based systems have shown effectiveness in detecting and classifying valvular heart disease, with diagnostic accuracy comparable to that achieved by expert clinicians [35]. Additionally, AI models can be trained to identify patterns associated with cardiomyopathies, contributing to early diagnosis and risk stratification [3].

ARTIFICIAL INTELLIGENCE AND CARDIAC MAGNETIC RESONANCE IMAGING

Cardiac MRI is used for diagnosing CVDs. It allows for assessing cardiac morphology, function, perfusion, and quantification of interstitial myocardial tissue volume. This method is recommended for patients with inconclusive echocardiographic findings to evaluate cardiac anatomy and function, including systolic and diastolic dysfunctions [36]. Cardiac MRI is widely used in the diagnosis of cardiomyopathies, congenital and acquired heart defects, pericardial disease, CAD, and cardiac tumors. These tasks require high temporal and spatial resolution images, which can lead to longer scan times. Integrating ML may enhance procedural efficiency and improve the accuracy of data interpretation. The most important tasks addressed using DL include image reconstruction, segmentation, and quality control [37].

Although cardiac MRI is performed with high resolution, interpretation of its results by a physician may be time-consuming and susceptible to error. DL techniques have been applied to automate information extraction from MRI images [38]. An automated two-dimensional CNN was used to process cardiac MRI data as input. It analyzed image features using CNNs, aggregated them, and performed pixel-level segmentation [38]. A study assessed the accuracy of left ventricular chamber volume, mass, and LVEF measurements in 110 patients. It was found that the accuracy of the CNN was comparable to that of an experienced specialist; however, the AI performance was 186 times higher [39]. In another study, ML was used to detect diagnostic characteristics of pulmonary arterial hypertension from MRI data. The algorithm enabled more accurate differential diagnosis compared with manual measurements. Moreover, the AI-based assessment required less time (within 10 seconds) and showed reduced variability [40].

Farrag et al. [41] developed a fully automated algorithm for cardiac segmentation that covers the entire region from the apex to the base and operates across all phases of the cardiac cycle. Moreover, an automatic segmentation algorithm was evaluated by Bernard et al. [42], who reported a Dice similarity coefficient ≥0.95 compared with manual measurements. Fahmy et al. [43, 44] introduced a CNN capable of automatically determining left ventricular mass and scar volume using late gadolinium enhancement in patients with hypertrophic cardiomyopathy.

Bai et al. [38] trained a 16-layer CNN to perform automated analysis of MRI images using data from 4875 patients in the United Kingdom. The method was evaluated based on the Dice coefficient and the determination of the following:

• Left ventricular end-diastolic volume (EDV), end-systolic volume (ESV), and mass
• Right ventricular EDV and ESV.

When analyzing MRI images from 600 patients, this algorithm demonstrated high segmentation performance:

• For the left and right ventricles, Dice coefficients were 0.94 and 0.90, respectively.
• For the left and right atria along the long axis, Dice coefficients were 0.93 and 0.96, respectively.

The use of CNNs in cardiac MRI image analysis demonstrates performance comparable to that of experienced specialists [38]. Notably, CNNs achieve high segmentation accuracy when trained on data acquired using the same MRI scanner; however, their performance tends to decline when images are obtained using different scanners. Chen et al. [45] developed a two-dimensional CNN using a dataset comprising results from 3975 patients. In this work, batch normalization was applied after each hidden convolutional layer to stabilize and accelerate the training process. Results showed that the proposed method achieved higher overall segmentation accuracy with lower variance compared with data from other studies [38, 45].

Post-contrast MRI images for detecting delayed myocardial enhancement (late gadolinium enhancement) provide valuable information on myocardial tissue integrity. For example, the presence and extent of late gadolinium enhancement have been associated with adverse clinical outcomes in patients with hypertrophic cardiomyopathy [46]. However, this approach has limitations and should be cautiously used in patients with severe renal failure or gadolinium-based contrast agent allergy [47]. In a recent randomized controlled trial involving 1348 patients with hypertrophic cardiomyopathy, a novel DL-based MRI technique called virtual native enhancement (VNE) was utilized. This method enabled acquiring images that were equivalent to standard late gadolinium enhancement images but without the use of contrast agents [48]. The DL algorithm processed and enhanced signals from native T1 maps (reflecting tissue T1 relaxation) and cine images (sequential images captured during different phases of the cardiac cycle). Each stream employed a U-net encoder–decoder architecture. The encoder extracted fine to coarse image features and generated multiscale representations, which were then integrated by the decoder to produce final outputs. Feature maps from the U-nets were combined and added to a subsequent CNN to generate the final image. The model was trained using generative adversarial networks. Upon comparison, the authors found that the image quality obtained with VNE was superior to that of MRI with late gadolinium enhancement. These methods demonstrated high performance in assessing the spatial distribution and quantification of myocardial injury. Although the outputs were similar, VNE did not require intravenous contrast administration, making it suitable for repeat imaging to confirm study results. This renders the technology attractive for clinical use and shows its potential to expand the role of MRI in diagnosing other cardiac disorders [48].

The application of AI technologies in cardiac MRI interpretation aims to enhance image acquisition, optimization, and analysis [38]. DL methods can be used for automated cardiac segmentation, facilitating faster and more qualitative assessment of cardiac function [49]. Additionally, AI algorithms are capable of evaluating blood flow to detect ischemia and tissue characteristics to identify fibrosis [50]. Integrating AI into the analysis and interpretation of cardiac MRI may support the monitoring of cardiomyopathies, inflammatory heart diseases, and CAD.

APPLICATION OF ARTIFICIAL INTELLIGENCE IN CARDIAC AND CORONARY ARTERY COMPUTED TOMOGRAPHY

CT is among the most rapidly advancing imaging-based diagnostic modalities. AI technologies support CT interpretation for diagnosing various conditions, particularly those involving the myocardium, heart valves, and coronary arteries [50]. These technologies can enhance image quality and perform automated image analysis, including assessment of coronary artery calcification severity.

Risk assessment for CAD development is crucial to reduce the incidence of future cardiovascular events. Traditional prediction models have limitations, such as intercohort variability and the exclusion of key variables. The implementation of ML in clinical practice provides robust predictive tools capable of accurately forecasting CAD development [52]. Coronary CT angiography is a minimally invasive modality that enables evaluation of coronary artery patency. The application of AI technologies allows for qualitative and quantitative assessment of atherosclerosis and determination of stenosis severity [53]. ML techniques are utilized in coronary CT angiography to produce high-quality images and detect pathological patterns, improving diagnostic accuracy and supporting enhanced risk stratification. To compare risk stratification approaches, a study of 8844 patients was conducted, followed by analysis of the receiver operating characteristic (ROC) curve and calculation of the area under the curve (AUC) [54]. With a mean follow-up duration of 4.6 ± 1.5 years, AUC was higher when ML was used. It was also employed to assess the severity of coronary artery calcification in 13,054 patients with established or suspected CAD [19]. The implementation of ML technology increased CAD detection by approximately 9%. In the subgroup of patients aged <65 years, this rate increased to 17% [19]. Han et al. [55] evaluated ML-based risk stratification in a healthy population of 85,945 participants, applying an AI algorithm to predict moderate (coronary artery calcification >100) and high (coronary artery calcification >400) risk of CAD. Moderate and high CAD risk were identified in 8.4% and 2.4% of participants, respectively. The use of ML algorithms outperformed conventional risk prediction scores in moderate- and high-risk groups, indicating the greater effectiveness of these technologies.

In a 5-year study, Motwani et al. [56] evaluated the performance of ML algorithms in interpreting coronary CT angiography results of 10,030 patients with suspected CAD. All patients underwent coronary CT angiography according to clinical indications. ML performed automatic feature selection by ranking the information, building a model using the LogitBoost logistic regression algorithm with cross-validation throughout the process. The primary outcome was all-cause mortality. Over the 5-year follow-up period, 745 patients died. The use of ML for comprehensive predictor assessment enabled more accurate evaluation of all-cause mortality, as indicated by a higher AUC value.

Furthermore, AI technologies can be applied in the diagnosis of acute coronary syndrome [53]. Most cases of acute coronary syndrome are caused by unstable and nonobstructive atherosclerotic plaques. Currently available noninvasive diagnostic methods that detect coronary artery stenosis or stress-induced myocardial ischemia are unable to identify these pathological changes. It is well established that vascular inflammation contributes to the formation and rupture of atherosclerotic plaques, resulting in acute coronary syndrome. One of the relevant indicators, the perivascular fat attenuation index (FAI), measured in Hounsfield units (HU) of X-ray attenuation, enables age- and sex-adjusted interpretation, increasing its clinical value. FAI is beneficial for evaluating coronary inflammation and improves risk prediction by incorporating risk factors and plaque distribution in the coronary arteries [57].

A study evaluated the prognostic value of FAI in 1872 patients who underwent coronary CT angiography [58]. It included measurement of the FAI in three coronary arteries. FAI values around the left anterior descending artery and proximal segment of the right coronary artery were associated with increased all-cause mortality. A threshold of −70.1 HU was identified. An increase in this value correlated with a marked increase in cardiac mortality. The study showed that FAI measurement may enhance the effectiveness of cardiovascular risk prediction. This approach supports early treatment of patients with unstable atherosclerotic plaques in the absence of overt clinical CAD signs, potentially lowering the risk of myocardial infarction in this population.

AI technologies have demonstrated excellent performance in image analysis and cardiovascular risk assessment. DL algorithms are capable of evaluating coronary artery calcification in an automated manner, enabling rapid and accurate diagnosis of the condition [59]. In addition, AI supports risk stratification for CAD. Automatic assessment of the left ventricular myocardium using AI algorithms during a single rest-phase coronary CT angiography scan may be an effective tool for identifying patients with functionally significant coronary artery stenosis without requiring anatomical evaluation. This approach may reduce the number of invasive FFR measurements, thereby optimizing the diagnostic process and minimizing patient risk [60]. Such algorithms can automatically segment coronary artery images, detect plaques, and evaluate their composition, potentially reducing interpretation time and improving diagnostic efficiency. Furthermore, AI models can characterize myocardial tissue based on imaging data and predict adverse outcomes [20]. The application of AI technologies provides new prospects for personalized risk assessment and treatment planning.

Based on coronary CT angiography data, van Hamersvelt et al. [61] assessed the degree of coronary artery stenosis to identify patients with stress-induced myocardial ischemia. Several AI methods were used in this study: first, the CNN was employed to segment the left ventricular myocardium; subsequently, the patients were classified into groups with or without functionally significant coronary artery stenosis using a support vector machine. The use of this method was associated with improved diagnostic accuracy (AUC = 0.76). Sensitivity and specificity were 84.6% and 48.4%, respectively.

Kelm et al. [62] applied the ML algorithm for the automated identification and classification of coronary artery stenosis. A random forest method was utilized to analyze 229 images. The model accurately identified stenosis and evaluated the cross-sectional luminal area, with a mean processing time of 1.8 seconds per case. Zreik et al. [63] trained a CNN to detect atherosclerotic plaques in the coronary arteries, determine their composition, and evaluate the presence of obstruction. The proposed method achieved an accuracy of 0.77 for plaque detection and characterization and 0.80 for identifying stenosis and assessing its anatomical significance.

LIMITATIONS OF EXISTING STUDIES AND FUTURE DIRECTIONS

The performance of AI models highly depends on the quality and standardization of training data. Variability in imaging protocols and scanner calibration can pose significant challenges to image processing and interpretation [64]. Furthermore, the absence of standardized annotation practices and presence of noise and artifacts may compromise the accuracy of AI algorithms. These limitations are mitigated by the development of multi-institutional databases and standardized imaging protocols [65].

Many AI models, particularly those based on DL, function as black boxes, relying solely on algorithms and training data, which makes it difficult for clinicians to understand the basis of their results. This lack of transparency leads to skepticism and reluctance to implement such technologies in clinical practice. Explainable AI methods, such as feature importance analysis and Bayesian networks, are being developed to address this issue [66].

Moreover, the integration of AI technologies raises critical regulatory and ethical concerns regarding data privacy. Effective mechanisms should be established to ensure the safe operation of these algorithms. Notably, the question of liability for diagnostic errors made by AI systems remains unresolved. This raises numerous legal and ethical issues, for example, who among the following should be held accountable in the event of a mistake: the algorithm developer, healthcare institution implementing the technology, or physician making the final decision based on AI-generated results. The challenge lies in the fact that AI systems make decisions based on large data, rendering them difficult to analyze and explain. Therefore, ongoing discussion and development of clear regulatory frameworks are required to define responsibility and ensure patient safety in the clinical application of AI technologies [67].

CONCLUSION

Thus, AI technologies are of great interest in healthcare, as they enable rapid analysis of large data, demonstrating high effectiveness. In cardiology and cardiovascular imaging, these technologies have shown promising results. Considering that the performance of AI systems depends on the quality of input data, it is crucial to address the issue of broader data sharing according to ethical standards. Although AI can support outcome prediction, full automation remains unfeasible owing to the individualized nature of patient-specific parameters.

AI can provide additional assistance to specialists, contributing to enhanced workflow efficiency and improved medical care. However, AI algorithms should undergo rigorous prospective clinical evaluation before being integrated into routine cardiology practice.

ADDITIONAL INFORMATION

Funding source. This article was not supported by any external sources of funding.

Disclosure of interests. The authors declare that they have no relationships, activities or interests (personal, professional or financial) with third parties (commercial, non-commercial, private) whose interests may be affected by the content of the article, as well as no other relationships, activities or interests over the past three years that must be reported.

Authors’ contribution. A.Kh. Islamgulov, A.S. Bogdanova, D.I. Sufiiarov, A.V. Chernyavskaya: concept of the work, coordination of the final version of the manuscript; E.R. Bayrakaeva, A.A. Maksimova, N.V. Nemychnikov, D.R. Bikieva, A.I. Shakhmayeva, L.A. Burdina, A.V. Bolekhan, E.I. Akimov, Z.Z. Shurakova: collection and analysis of literature data, writing and editing the article. Thereby, all authors provided approval of the version to be published and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

About the authors

Almaz Kh. Islamgulov

Bashkir State Medical University

Author for correspondence.
Email: aslmaz2000@rambler.ru
ORCID iD: 0000-0003-0567-7515
SPIN-code: 8701-3486
Russian Federation, Ufa

Alina S. Bogdanova

Kuban State Medical University

Email: balinochka25@gmail.com
ORCID iD: 0009-0004-9333-5164
Russian Federation, Krasnodar

Damir I. Sufiiarov

Bashkir State Medical University

Email: damur_5@mail.ru
ORCID iD: 0009-0004-3516-6307
SPIN-code: 3311-2947
Russian Federation, Ufa

Alina V. Chernyavskaya

Kuban State Medical University

Email: alinaxxx909@gmail.com
ORCID iD: 0009-0007-8071-1150
Russian Federation, Krasnodar

Elena R. Bairakaeva

Bashkir State Medical University

Email: bairakaeva_0@mail.ru
ORCID iD: 0009-0004-7683-5781
Russian Federation, Ufa

Anastasia A. Maksimova

Bashkir State Medical University

Email: antasiamks@gmail.com
ORCID iD: 0009-0003-4115-2887
Russian Federation, Ufa

Nikita V. Nemychnikov

Bashkir State Medical University

Email: nikita.nemychnikov2001@gmail.com
ORCID iD: 0009-0001-8841-3373
Russian Federation, Ufa

Diana R. Bikieva

Bashkir State Medical University

Email: bikieva.dina@mail.ru
ORCID iD: 0009-0006-5453-5686
SPIN-code: 7078-7424
Russian Federation, Ufa

Alsu I. Shakhmaeva

Bashkir State Medical University

Email: shakhmaeva02@mail.ru
ORCID iD: 0009-0002-8805-9172
Russian Federation, Ufa

Lyubov A. Burdina

Pskov State University

Email: lubovburdina19@gmail.com
ORCID iD: 0009-0004-9199-2515
Russian Federation, Pskov

Aleksandr V. Bolekhan

Pskov State University

Email: sasha-x500@mail.ru
ORCID iD: 0009-0009-3458-2858
Russian Federation, Pskov

Egor I. Akimov

Tula State University

Email: egor.akimov.2001@mail.ru
ORCID iD: 0009-0002-2504-5363
Russian Federation, Tula

Zilya Z. Shurakova

Bashkir State Medical University

Email: divaeva.zilya@mail.ru
ORCID iD: 0009-0007-9625-9787
Russian Federation, Ufa

References

Supplementary files