A study published in the Journal of the American College of Radiology found that large language models can help patients better understand their radiology reports when summaries are provided, but human oversight is necessary. Patients who received reports with summaries generated by these models had a better understanding compared to standard care. The 21st Century Cures Act mandates patient access to radiology reports, but patients may struggle to understand them and prefer lay language or summaries. While large language models like ChatGPT and Gemini are available for medical questions, their generated answers may contain incorrect information. The study included 99 patients divided into different cohorts receiving standard care, Gemini-generated summaries, Scanslated reports, or a combination of both. Patients with summaries had a higher level of understanding compared to standard care, but there was no significant difference in the need to search report contents online. Most large language model-generated summaries required editing before release to remove suggestions of prognosis, treatment, or causality. The study emphasizes the importance of human oversight before widespread clinical deployment of these models. Future research should focus on measuring impact in larger populations, different clinical settings, evaluating return on investment, and refining large language model performance.
Source link
Source link: https://www.auntminnie.com/imaging-informatics/artificial-intelligence/article/15678833/large-language-models-help-patients-understand-radiology-reports
Large language models aid patients in comprehending radiology reports. #MedicalAI
