Artificial Intelligence in Radiology

Since its first use in medical purpose in the 1960s, the concept of artificial intelligence has been especially appealing to health care, particularly radiology. With the development of ever more powerful computers from the 1990s to the present, various forms of artificial intelligence have found their way into different medical specialties – most notably radiology, dermatology, ophthalmology, and pathology. Due to the growing presence of such systems, it is paramount for the specialists handling them to get acquainted with them in order to provide the best service for their patients. It is therefore the aim of this article to explain the most basic principles of artificial intelligence, accentuating the most prominent concepts used in radiology today, such as deep learning and neural networks. It will also mention some of the artificial intelligence systems approved for clinical use in the US, such as IDx-DR, used to discover more than mild diabetic retinopathy in patients over 22 years of age; and Arterys, used for cardiac segmentation and discovering liver and lung nodules. Same as in many other fields, there is a constant need for improvement – in construction, testing, and application of these new technologies. Many ethical questions are asked, considering privacy and liability of artificial intelligence systems in clinical use. One of the greatest concerns for radiologists is the possibility of being replaced by these systems. This scenario seems to be far-fetched, at least for the time being. Radiologists should use that time to get to know the “enemy”. If they accomplish this, they might discover that they had had an ally all along