New research published in the journal 'Radiology' by South Korean researchers has identified key factors that help radiologists avoid being misled by incorrect advice from large language models (LLMs) like ChatGPT. While these AI models can effectively solve diagnostic problems, they sometimes produce unreliable outputs. The study involved 10 radiologists interpreting chest imaging from 100 patients, and it found that successful collaboration between radiologists and LLMs was linked to both the confidence in the AI model and the radiologists' expertise in chest imaging. The findings suggest that while high confidence in the model often correlated with correct decisions, having expertise acts as a protective measure against misleading rationales. The research emphasizes that effective interaction with LLMs depends on factors beyond model performance, notably reader expertise and confidence. The study utilized curated data from the Korean Society of Thoracic Radiology and compared sessions with and without the use of AI, highlighting the importance of clinical expertise in guiding AI outputs and reducing inaccuracies.
Wed, 12 Aug 2026 06:04:39 GMT | Radiology Business