Emergence of Artificial Intelligence in Cardiology - The Future View PDF

We live in a world where AI is pervasive. Besides auto-completing sentences as we type, it also populates Google searches before we finish our thoughts, enables cars to drive themselves, enables us to speak to our phones, and supports language translation. It has been used in medicine to identify pathologic specimens, identify mammogram lesions, and identify retinal pathology with a level of skill exceeding that of trained ophthalmologists [1, 2].

Despite being vilified as a tool that will result in mass unemployment and economic disruption, it has also been lauded as a potential savior that will liberate humanity from tedious tasks and enable people to engage, interact, and exist on a higher level, while at the same time vilified as a tool that will lead to massive unemployment and economic disruption [3].

ML of various types, but most often deep neural networks, is called AI [4]. Although there are many types of neural networks, broadly speaking, deep neural networks permit sophisticated recognition of subtle patterns in a nonlinear manner using models that contain many layers of data abstraction and synthesis, which leads to an uncanny ability to “read” mammograms and electrocardiograms (ECGs) or recognize faces. The term “artificial intelligence” refers to ML of various types, such as deep neural networks [5, 6]. By using models containing many layers of abstraction and synthesis of data, deep neural networks can recognize subtle patterns in a nonlinear manner. In this way, mammograms, ECGs, and faces can be read uncannily. In spite of the fact that deep neural networks provide deep intelligence, they are currently limited in their ability to provide spontaneous adaptability or general intelligence, as they provide only a narrow type of intelligence with very focused skills [7-10]. This narrative review used search terms related to AI and medicine and cardiology subspecialties to search PubMed and MEDLINE databases without date restrictions. Relevance was considered when selecting articles for inclusion.

Mounting evidence reveals that ML will power the new tools that drive cardiovascular medicine in the near future. The article highlights the rapidly emerging role of ML in cardiovascular medicine [8]. The use of AI has included interpreting echocardiograms, automatically identifying heart rhythms from ECGs, uniquely identifying individuals from ECGs as biometric signals, and detecting left ventricular dysfunction from the surface ECG as a sign of heart disease. Cardiovascular specialists are unlikely to be replaced by AI, however [9-11]. It may instead serve as a tool for skilled practitioners to expand their clinical capabilities, make more accurate and timely diagnoses, and improve care delivery [12].

AI is unlikely to replace cardiovascular specialists, however. AI has the potential to expand the clinical abilities of skilled practitioners, to increase the accuracy and speed of diagnoses, and to improve care delivery as a whole [13].

It is essential that we understand AI’s strengths, limitations, opportunities, and risks, just as we do with any statistical method or tool. We explore the nature of ML, how it is developed, the types of problems it poses, and its limitations, as well as some of its current and promising applications in cardiovascular medicine. Meanwhile, we will emphasize potential risks, such as a potential bias when AI is applied to populations outside those represented in a training set, data security threats, and data ownership concerns. We must embrace the powerful emerging tools enabled by AI to ensure that they are properly applied for the benefit of humanity as we care for patients and practice cardiovascular medicine. In this review, we aim to provide a foundation for understanding benefits.