Abstract
On a global scale, a radical transformation of the healthcare sector is taking place right before our eyes. The last few years have become a turning point in terms of the number of new directions, the emergence of innovative diagnostic and treatment methods, and the introduction of digital platforms. Digital medicine uses information and communication technologies to address the numerous problems associated with ensuring the quality and accessibility of medical care already available today. The rapid development of neural networks and artificial intelligence (AI) provides doctors with ample opportunities to predict the course of diseases and calculate the risks to the health of patients. Manufacturers of medical devices offer consumer a wide range of software and products with AI embedded. Despite the tremendous advances in the application of AI in medicine, the medical community is highly concerned about some of the intractable problems associated with the too rapid and ubiquitous use of these digital platforms. A highly trained neural network is an extremely complex computer program consisting of a large number of internal hidden layers with customizable parameters. The more complex the neural network and the number of computational operations it performs, the more difficult it is to understand the processes in its inner layers. The functioning of AI systems in a black box format makes explaining the results of its work a very non-trivial task. Therefore, in the future, research will certainly be required assessing the reliability of these systems and interpreting their decision-making processes which, will affect the neural networks of the latest generations.
Subject
General Materials Science