Dear AuntMinnie Member,
Could ChatGPT help reduce inappropriate imaging for emergency department patients? Not at this time, according to a new study.
Our featured story in this issue reports on how ChatGPT tends to recommend unnecessary imaging exams and underperform residents. You can read all about it here.
Speaking of ChatGPT, the GPT-4 Vision model also isn’t reliable enough in its current form to reliably interpret radiologic images, according to a group from the U.S. National Institutes of Health.
The news hasn’t been all negative for large language models (LLM), however. Another study concluded that an LLM can generate radiology report impressions that are professional and linguistically appropriate for a full spectrum of imaging exams.
In other developments, researchers have found that an AI algorithm can help predict future breast cancer risk and identify high-risk women. Also, AI applications are increasingly being used for a variety of CT applications, a recent presentation reported.
For example, AI can locate incidental, extrapulmonary structures on chest CT scans, helping clinicians foretell all-cause mortality risk. And machine-learning models can predict interstitial lung abnormalities on CT.
Be sure to stop by our Artificial Intelligence content area for all of the latest news in radiology AI.
Erik L. Ridley
Editor in Chief
AuntMinnie.com
