A recent study published in *Radiology* evaluated the performance of the large language model GPT-4o in selecting optimal protocols for abdominal and pelvic CT scans. The study involved over 1,400 patients and found that GPT-4o achieved optimal protocol selection in 96.2% of cases, outperforming unassisted radiologists who succeeded 88.3% of the time. The research highlighted that GPT-4o not only matched but exceeded radiologists across various experience levels, indicating potential for standardization in protocoling and improving workflow efficiency without compromising safety. Researchers noted that while there is variability in imaging requisition interpretations, the findings support the feasibility of LLMs in automating CT protocoling and reducing the time radiologists spend on non-interpretive tasks.
Wed, 07 Jan 2026 11:39:18 GMT | diagnosticimaging.com