Dados do Trabalho

Título

Beyond Writing and Reporting: Opportunities and Challenges of GPT-based models in Radiology

Descrição sucinta do(s) objetivo(s)

To learn definitions and basic concepts of generative AI, natural language processing (NLP) and large language models (LLM) applied to healhcare.
To understand the potencial impact of generative pre-trained transformer (GPT)-based models in the radiology practice.
To be familiar with some potential applications of GPT-based models in radiology, including image manipulation, report generation, educational support, clinical decision support, and data analysis.
To emphasize the challenges and limitations before implementation of these models in clinical practice, with focus on ethical and regulatory issues.

Descrição da(s) doença(s), método(s) e/ou técnica(s)

A review of literature, including cases and illustrations, based on recent articles and editorials about GPT-based models role in radiology beyond reporting, discussing applications, opportunities and current challanges,

Discussão

AI has significantly changed the world as we know it by using massive data processment to simulate human inteligence and accomplish complex automatized tasks. GPT-based models, a type of AI, are expected to revolutionize radiology with multiple applications, although reporting is the most popularized one. Despite their potential, challanges, regulatory issues and ethical considerations must be considered for appropriate use and validation. Applications include image manipulation and improving, medical writing and reporting, clinical decision support, health insurance and advocacy issues, education and many others. Examples of limitations and challanges for appropriate GPT-models use are hallucinations and other common mistakes, privacy, security,r risks on clinical implementation , ethical issues, legal considerations, bias and regulation. GPT-based models appropriate use will depend on human and AI integration, focused on a responsible AI with structured certification programs. High expectations are related to future perspectives, however how we develop and regulate AI will dictate the healthy progress.

Conclusões

GPT-based models have revolutionazed healthcare, specially radiology. As technology have evolved, AI models have improved their overall performance and enabling enhancement in the diagnostic process, treatment planning and research. However, concens about privacy and security, risks on clinical implementation , ethical issues, legal considerations , bias and regulation are still in debate. For a safe implamantation, AI models demand a organized and structured regulation based on cerfitification programs, following the concept of responsible AI.

Palavras Chave

Artificial Inteligence; Radiology

Arquivos