Dear AuntMinnie Member,
As always, the Conference on Machine Intelligence in Medical Imaging (CMIMI) held this week in Boston presented attendees with the latest in cutting-edge research for imaging AI. But practical talks discussing implementation strategies were also prominent.
For example, the opening keynote talk by Nina Kottler, MD, of Radiology Partners, focused on best practices for implementing radiology AI. She also shared additional thoughts on this vitally important topic in a video interview at the meeting.
Large language models (LLMs) were featured heavily at the Society for Imaging Informatics in Medicine's AI-focused conference, both in the oral presentations and in the posters. We caught up with CMIMI 2024 Co-Chair Eliot Siegel, MD, of the University of Maryland, to get his take on LLMs and other radiology AI trends.
A study presented at the meeting reported that LLMs can be useful in monitoring and validating AI algorithms after deployment. Also, ChatGPT was deemed to be helpful in classifying clinically significant breast pain and was found to be reasonably accurate for identifying radiological terms in chest x-ray reports. And a rules-based approach was shown to improve ChatGPT’s protocoling performance for chest CT exams.
Medical physicists have a role to play in imaging AI, according to another talk. We spoke with Da Zhang, PhD, of Boston Children's Hospital and Harvard Medical School to learn more about the importance of multidisciplinary collaborations for advancing AI in clinical practice.
In other news from CMIMI 2024, a new method that utilizes probability data extracted from radiology reports for specific conditions could aid in differential diagnosis. A deep-learning framework can also identify and accurately describe parotid gland tumors on CT scans. Additionally, a CT-based AI framework was able to boost opportunistic screening of splenomegaly.
See the list below for our top stories from CMIMI 2024 or stop by our special RADCast section.
- CMIMI: What are best practices for radiology AI deployment?
- Video from CMIMI: Strategies for success with radiology AI
- Video from CMIMI: Advances in large language models and other AI trends
- CMIMI: LLMs can monitor AI software after deployment
- CMIMI: ChatGPT classifies clinically significant breast pain
- CMIMI: ChatGPT ‘reasonably accurate’ for identifying chest x-ray terms
- CMIMI: Rules-based approach improves ChatGPT’s protocoling performance
- Video from CMIMI: Collaboration key to further advancing imaging AI
- CMIMI: Probability models could enable differential diagnosis
- CMIMI: Deep learning from CT exams identifies parotid gland tumors
- CMIMI: CT-based AI framework improves splenomegaly diagnosis
Erik L. Ridley
Editor in Chief
AuntMinnie.com
