The specialized large language model generates higher quality summaries than GPT-4o, particularly for radiology reports.
Performance differences are particularly notable in the context of CT and MRI imaging reports, highlighting the model's strengths.
Assessment based on radiology report summarization tasks demonstrates the superior efficacy of the specialized model over the general-purpose alternative.
These findings may encourage further development of specialized models for specific medical applications in radiology.
Abstract
A specialized large language model (LLM) for report summarization had better performance than GPT-4o (OpenAI), a general-purpose LLM, in generating CT and MRI radiology report summaries.