TORONTO -- AuntMinnie.com spoke with AI expert Tyler Bradshaw, PhD, a medical physicist from the University of Wisconsin-Madison at the Society of Nuclear Medicine and Molecular Imaging annual meeting about the emergence of AI algorithms and large language models (LLMs) in nuclear medicine.
Bradshaw noted that almost every application of AI in radiology is also happening in nuclear medicine, including image enhancement, detecting disease, segmenting pathology for automatic quantification, and improving workflows.
“We’re seeing all that presented here at this meeting. In fact, I’d say we’re seeing even a maturation of these tools this year,” he said.
Click here for our full SNMMI coverage.
Ultimately, no, ChatGPT can’t start an IV line, Bradshaw said, but he posed that question in a session on June 8 geared toward nuclear medicine technologists to introduce them to the ideas and history of LLMs. Nonetheless, he envisions in the near future “chatbots” powered by LLMs that nuclear medicine clinicians will be able to have conversations with to optimize protocols and navigate complex electronic medical records.
For nuclear medicine, perhaps most exciting is that LLMs should evolve in the next few years to be truly “multimodal models” – not in the sense of imaging modalities, but in the way that they will be able to operate not just on text, but also on images and audio.
“This is the direction that large language models are going… I think we’re going to see that type of technology come into the clinic and do things we can’t even imagine,” he said.
