Large Language Models in Medical Imaging: Opportunities, Challenges and Future Trends
Code: G-1468
Authors: Ali Tarighatnia, Masoud Amanzadeh * ℗, Abdollah Mahdavi, Mahnaz Hamedan
Schedule: Not Scheduled!
Tag: Biomedical Signal Processing
Download: Download Poster
Abstract:
Abstract
Background and aims: Medical imaging has long been intertwined with advancements in computer science, with artificial intelligence (AI) playing a transformative role in optimizing diagnostic accuracy and patient care. Recent developments in large language models (LLMs), such as OpenAI's ChatGPT-4 and Google's Gemini, have demonstrated exceptional natural language processing (NLP) capabilities. This study explores the applications and challenges of LLMs in medical imaging, aiming to provide a comprehensive assessment of their role in radiology. Method: This is a review study that was conducted on the applications and challenges of LLMs in medical imaging. A literature search was performed in electronic databases including PubMed, Elsevier, and Google Scholar to identify relevant articles until February 1, 2025. Articles were searched in databases by combining related terms such as "large language model", “LLM”, and “medical imaging”. We conducted title, abstract, and full-text screening based on inclusion/exclusion criteria. Results: LLMs offer various applications in medical imaging, including automated diagnostic report generation, clinical decision support, integration with radiology information systems, medical education, and research support. Studies suggest that LLM-generated reports can match or exceed the quality of reports by experienced radiologists, reducing reporting time and minimizing errors. Additionally, LLMs assist in differential diagnosis, facilitate access to patient data, and enhance the learning experience for radiologists and medical students. However, their adoption comes with notable challenges. Key concerns include data privacy and security, hallucinations, lack of interpretability, overconfidence in outputs, bias in training data, regulatory uncertainties, and resistance to adoption by healthcare professionals. Conclusion: LLMs hold immense potential to revolutionize radiology by improving diagnostic accuracy, optimizing workflows, and enhancing medical education and research. However, their implementation requires overcoming significant technical, ethical, and legal barriers. Collaboration among radiologists, AI developers, policymakers, and stakeholders is essential to establish robust guidelines, improve model transparency, and ensure responsible integration into healthcare systems. Future research should focus on refining these models, mitigating risks, and developing regulatory frameworks to harness their full potential while maintaining patient safety and trust in AI-driven radiology.
Keywords
Medical Imaging, Large Language Model, Radiology
