Radiology errors in the interpretation of CT and MR images.

A recent study of 10,090 body CT scans, reviewed by attending physicians in the abdominal division, found no significant difference in error rates between day and night interpretations. The study recorded 44 errors in 2,164 day scans and 226 errors in 7,446 night scans.

Additional Study

Research published in the Journal of the American College of Radiology shows that up to 44% of diagnostic errors are linked to incorrectly ordered laboratory tests and radiology exams, with CT and MRI imaging being particularly affected. A study led by Dr. Ariadne DeSimone at Brigham and Women’s Hospital found that tailored order options and educational interventions significantly reduced these errors by 83%, from 16% to 3%. The research emphasizes improving the imaging order process to reduce diagnostic errors and optimize radiologists’ workflow by minimizing time spent correcting misordered exams.

A separate retrospective study examined the use of ChatGPT-4 in assisting radiologists with error detection in radiology reports. The study involved 200 reports, with 150 common errors deliberately introduced. ChatGPT-4 identified 82.7% of the errors, closely matching the performance of senior radiologists (89.3%) and outperforming assistants and residents (80.0%). Notably, ChatGPT-4 completed the task much faster, taking just 3.5 seconds per report compared to 25.1 seconds for radiologists, and was also more cost-effective. While promising, the study suggests that further prospective validation and specialized training are necessary before AI-assisted re-reading can be widely adopted in radiology.