A Recap of CMIMI 2020

The 2020 Conference on Machine Intelligence in Medical Imaging (CMIMI 2020) was held in virtual setting. The conference took place Sept. 13–14, and I would like to share my experience as an attendee.

CMIMI is organized by the Society for Imaging Informatics in Medicine (SIIM), whose mission is to “advance medical imaging informatics across the enterprise through education, research, and innovation in a multi-disciplinary community.” In addition to physicians, basic scientists, imaging informatics professionals, and vendors are all part of the SIIM community. This heterogeneous audience is a good fit for the interdisciplinary nature of machine learning (ML) research. In fact, what sets aside CMIMI from other medical conferences is its interesting interdisciplinary nature. During the conference, one would see participants with different backgrounds and expertise collaborating and discussing medical imaging problems or ideas. This is important considering how fast machine learning affects the radiology practice. The medical imaging community is getting bigger and radiologists should learn to collaborate well with other disciplines.

CMIMI is routinely held in person. CMIMI 2019, as an example, was a two-day event held in Austin, Texas. However, as the pandemic has changed the agenda of many meetings, CIMIMI 2020 was inevitably held virtually. Although some of us may look at these virtual conferences with a little skepticism, CMIMI organizers did a great job maintaining the event’s quality. The process of registration for the conference and submitting abstracts was handled smoothly. Instructions for abstract submissions were clear, and it took us a couple of weeks to receive the result of abstract reviews. A few days before the conference, we received clear instructions on entering the meeting portal and using that portal to access different sessions, online abstracts, and agendas. Sessions were not parallel, so in time for each session, all participants attended a virtual zoom meeting room specifically dedicated to that session. Presenters were invited to the room 15 minutes earlier and received instructions on how to do their presentations in order. In my opinion, the structure of the event was simple and highly efficient. More importantly, CMIMI organizers were always there to help us with any questions or problems, either before or during the meeting. The only downside of the meeting (which was inevitable due to its virtual nature), was the lack of side-conference networking. After all, few things can replace the joy of having coffee with a colleague from another institution during the breaks in sessions. Many project ideas came out of those coffee breaks!

CMIMI 2020 sessions covered a wide range of machine learning topics. While some presentations talked about the newest machine learning techniques, others illustrated the interesting applications of machine learning on different medical imaging modalities. For example, I got the chance to present two of our recent projects at the Radiology Informatics Laboratory (department of radiology, Mayo Clinic, Minn.) in clinical sessions of CMIMI 2020. These projects demonstrated how deep learning might help radiologists and orthopedic surgeons to predict total hip arthroplasty complications, like hip dislocation looking at post-operative radiographs. Unlike such applied projects, some people talked about basic deep learning topics like federated training or modern model architectures. We also had the chance to listen to some interesting talks like how machine learning will impact radiology in the future or how the FDA is working on evaluating machine learning models. One single, yet very important, take-home message from some of those talks is this: AI and machine learning are not here to replace radiologists; instead, they are here to help radiologists work more efficiently, reduce the burden of tedious tasks for them, and make them focus on more critical aspects of patient care.

CMIMI 2020 was a great opportunity to learn about machine learning advances and applications in medical imaging. There are definitely advantages for radiology experts interested in machine learning to attend such meetings and get more prepared for the future awaiting us: AI-enhanced radiology practice.