Dabartinė dirbtinio intelekto (DI) būklė radiologijoje

Abstract

Artificial intelligence (AI) holds a long history in computer sciences. Since the introduction of the term in the 1960s people have speculated on computers thinking and performing like humans. Due to the recent advancements in computational performance, as well as AI algorithms, we are coming closer and closer to this goal. Most people have already faced AI, even if unknowingly, in the form of social media feeds online, on the road though navigation apps or through autonomous driving, or when unlocking their phone through facial detection. Most recent attention came to AI though advances in chatbots, seemingly even capable of writing full articles with the right instructions. The usage of new AI systems proposes a range of benefits also regarding healthcare and especially a data and technologically driven field like radiology. From the first forms of radiographic imaging in the form of XRAY to the development of newer modalities, the domain of radiology was always linked to advancements in computer technologies. While the first attempts of implementing AI algorithms in the field of radiology were driven by diagnostic inaccuracy and therefore more effort for the radiologist to deal with these mistakes newer systems and advancements attempt to face and overcome these issues. When applying new technologies in medicine these must be regulated profoundly to ensure the safety of the patients. In these regards the FDA regulates the United States market and therefore FDA approved AI systems are a good indicator of applicable safe AI devices. This thesis aims at investigating the state of AI in radiology by asking the question where we are in the domain of radiology of these new computer systems. This is done by looking where AI can be applied in this field and what challenges are faced doing so as well as looking at FDA approves devices as well as FDA approved algorithms, its subspecialties and imaging modalities to get a view on the state of AI in 2023.

Ähnliche Arbeiten