AUSTIN, TX - A focus on clinical applications and a broader scope for artificial intelligence (AI) algorithms were among the research highlights at the Society for Imaging Informatics in Medicine's Conference on Machine Intelligence in Medical Imaging (C-MIMI) this week, said Paul Nagy, PhD, of Johns Hopkins University in a video interview.
Compared with many of the presentations at C-MIMI 2018 that delved into the architectural aspects of AI models, there was more of an emphasis on clinical applications at this year's meeting, according to Nagy. In addition, there was a shift away from using AI just to analyze a single organ or make a single specific diagnosis toward the more challenging task of applying algorithms to assess multiple organs or make multiple diagnoses, he said.
There were also several scientific talks exploring ways to compensate for different biases in datasets from other sites when training algorithms.
"That is really beginning to try ... to tackle the real-world performance degradation that we're seeing with machine learning [that's] based off of curated datasets," he said. "So [it's] really an interesting maturation, evolution around the problems that we're seeing."
Notably, many of the scientific presentations included discussions of how AI algorithms can be implemented in clinical practice.
"The presenters really did a nice job this year of not just showing the research, but showing it for how they would see that envisioned in clinical care," he said.
Nagy also shared his thoughts on other AI research topics, including his predictions for the hottest research areas over the next few years.
