Dear AI Insider,
Is radiology ready for AI -- and is AI also ready for radiology? Consultant Michael J. Cannavo, aka the PACSman, has tackled these questions in a thought-provoking and wide-ranging new column for AuntMinnie.com.
The latest edition of The PACSman Pontificates series is this newsletter’s featured article, and you can access it here.
Meanwhile, ChatGPT demonstrated promise as a tool for answering patient questions about radiation protection for medical imaging exams. The large language model yielded comparable results to answers posted on radiology institutional websites. Click here to learn more.
However, the model's utility for educating patients about cardiac imaging is more in question. On a positive note, another study concluded that ChatGPT could produce “near perfect” pancreatic cancer radiology reports.
In other AI developments, a deep-learning algorithm demonstrated value in helping to predict mortality in patients receiving transcatheter aortic valve implantation (TAVI).
Breast cancer screening is always a hot area for AI in radiology. Researchers from Denmark found that AI software could improve screening performance and decrease the number of false-positive exams – while also reducing breast radiologist workload.
It’s become increasingly important for radiology education to incorporate exposure to AI technology. Recently, an AI certificate course was shown to improve knowledge of AI for radiology residents.
Also, researchers have described how radiology departments and AI vendors can better promote health equity with responsible use of the technology. Finally, radiology practices must consider at least five factors when validating AI algorithms for their use, according to a review.
