Challenges and Consequences of Artificial Intelligence in Radiology

Černý, M. & Kvak, D. (2023). Výzvy a důsledky využití umělé inteligence v radiologii. In:Miloš Táborský et al. Digitální medicína II. ISBN: 978-80-88506-18-8.


The development of artificial intelligence (AI) in medical imaging presents many challenges and opportunities. AI can help physicians in diagnosis, detection of pathologies and improvement of image quality. If we understand its role as a supportive tool to augment and improve the physician's evidence base for informed decision making, rather than as a substitute for the physician, these technologies can make a significant contribution to improving the quality of healthcare that will benefit patients, healthcare professionals and operators. In the following, we attempt to outline some of the challenges associated with their use. Quite fundamental is the recognition that the accuracy of the model and its ability to provide predictions is dependent on the quality of the data on which it is trained. If the training data is inaccurate, incomplete or biased, it can have a negative impact on the accuracy of the model, especially if there is a systematic error. In situations where a population group is underrepresented in the training data, biased and potentially damaging predictions may result. It is important that models are transparent, so that clinicians can understand what their decisions are based on and can trust their recommendations. However, for AI to be truly effective and beneficial to physicians, it is essential that it is seamlessly integrated into existing clinical processes and does not increase physician workload. In the conclusion of the text, we briefly discuss the general principles of AI regulation in radiology and the specific form of regulation in the EU.


Related articles

awarded by