Abstract:

scular diseases are a public health problem, not only because of their high prevalence, morbidity and mortality rates, but also because of the high costs associated to the healthcare of these patients. In the period 1990–2019, the prevalence of these diseases increased from 257 million cases to 550 million cases globally(1). According to the World Health Organization (WHO), in 2016 there were 17.9 million deaths secondary to cardiovascular diseases, representing 31% of all deaths globally(2).In Europe, the costs involved in the healthcare of these patients each year is estimated to be about 210 million Euros(3). In the last decades, research has focused on technological development as a useful tool for doctors in making clinical decisions. Thus, the concepts of artificial intelligence and machine learning (ML) appeared, cardiology being the specialty with the greatest applicability of artificial intelligence. Artificial intelligence algorithms are widely used in the medical field. Among the applications of these algorithms are the analysis of acquired images necessary for the diagnosis, segmenting and reconstructing an image, controlling the quality of the image, fitting into a certain phenotype or establishing the patient’s prognosis(4). In cardiology, the main applications of artificial intelligence are in electrocardiography, transthoracic echocardiography, cardiac computed tomography angiography, single-photon emission computed tomography (PET-CT), cardiac magnetic resonance imaging and the diagnosis of heart failure. Nowadays, automatic electrocardiogram interpretation is frequently used. However, despite significant progress in the development of computerized electrocardiogram interpretation, there are still numerous limitations. Artificial intelligence methods aim to improve the accuracy of automated electrocardiogram interpretation and establish the patient’s prognosis. Modern ML models can identify the P-waves, T-waves, and QRS complexes, assess and calculate parameters such as heart rate, cardiac axis, and interval lengths and identify ST-segment changes or common rhythm disorders such as atrial fibrillation(4). As a consequence, it is possible to quickly triage the patients and identify those who require an appointment with a cardiologist or arrhythmologist. The echocardiographic evaluation, especially 3D evaluation and speckle tracking, generates a large volume of data that is potentially useful for establishing the diagnosis(5). Most of the time, these data aren’t used to their full potential. Using ML techniques, it is possible to interpret multiple sets of echocardiographic data, automatically and efficiently at the same time(6). In addition, these methods can integrate the clinical and paraclinical data available in the electronic records of the health systems’ databases with echocardiographic data, thus improving the diagnostic management(7). Artificial intelligence proves to be a cost-effective method that allows for more efficient triage of patients, more prompt and accurate diagnosis, reducing the need for invasive and expensive investigations, improving screening strategies, and, implicitly, therapeutic management. Despite its huge potential in cardiology, artificial intelligence is still a new field with many challenges, that raises also a series of ethical dilemmas.

Keywords: