POSSIBILITIES OF APPLYING MACHINE LEARNING TECHNOLOGIES IN THE SPHERE OF PRIMARY PREVENTION OF CARDIOVASCULAR DISEASES

Abstract

Highlights The review analyzes the studies devoted to the possibility of using machine learning methods to predict the occurrence of atrial fibrillation, cardiovascular risk factors, carotid atherosclerosis, and total cardiovascular risk. The combinations of machine learning methods with mobile, cloud and telemedicine technologies have significant prospects. In the near future, such technologies are expected to be used for atrial fibrillation screening and risk stratification using cardiac imaging data. Based on machine learning methods, mobile preventive technologies are being developed, particularly for nutritional behavior management. Abstract The article reviews the main directions of machine learning (ML) application in the primary prevention of cardiovascular diseases (CVD) and highlights examples of scientific and practical problems solved with its help. Currently, the possibility of using ML to predict cardiovascular risk, occurrence of atrial fibrillation (AF), cardiovascular risk factors, carotid atherosclerosis, etc. has been studied. The data of questionnaires, medical examination, laboratory indices, electrocardiography, cardio visualization, medications, genomics and proteomics are used in ML models. The most common classifiers are Random Forest, Support Vector, Neural Networks. As compared to traditional risk calculators many ML algorithms show improvement in prediction accuracy, but no evident leader has been defined yet. Deep ML technologies are at the very early stages of development. Mobile, cloud and telemedicine technologies open new possibilities for collection, storage and the use of medical data and can improve CVD prevention. In the near future, such technologies are expected to be used for atrial fibrillation screening as well as cardiovascular risk stratification using cardiac imaging data. Moreover, the addition of them to traditional risk factors provides the most stable risk estimates. There are examples of mobile ML technologies use to manage risk factors, particularly eating behavior. Attention is paid to such problems, as need to avoid overestimating the role of artificial intelligence in healthcare, algorithms’ bias, cybersecurity, ethical issues of medical data collection and use. Practical applicability of ML models and their impact on endpoints are currently understudied. A significant obstacle to implementation of ML technologies in healthcare is the lack of experience and regulation.