GPT-4 outperformed GPT-3.5 in generating highly accurate pancreatic cancer synoptic reports from original reports, aiding surgeons in assessing tumor resectability for patients with pancreatic ductal adenocarcinoma. The AI-generated reports saved time for surgeons and improved the categorization of resectability compared to original reports. However, some clinically significant errors were noted, indicating the need for supervision when using AI-generated reports for management decisions. While GPT-4 showed high accuracy, it did make errors in some areas. The study emphasizes the potential of AI in standardizing reports and enhancing communication between radiologists and surgeons, but cautions against relying solely on AI for clinical decisions.
Source link
Can ChatGPT Improve Pancreatic Cancer Synoptic Reports?
