The latest AIMed webinar “Breaking through the bottlenecks: Successful translation of AI (artificial intelligence) into clinical radiology” took place on 24 September at 11 am (PDT). 

Esteban Rubens, Global Enterprise Imaging Principal of Pure Storage was the moderator as Dr. Anthony Chang, Pediatric Cardiologist; Founder and Chairman of AIMed, and Chief Intelligence and Innovation Officer of Children’s Hospital of Orange County (CHOC); R. Kent Hutson, Director of Imaging Informatics, Radiology Partners, and Samir Shah, Vice-President of Clinical Operations, Radiology Partners took turn to express their opinions. 

Unlike the previous AIMed webinars, the session did not begin with speakers’ presentations, rather, they focused on addressing some of the challenges the sub-specialty faces while trying to incorporate AI and new technologies. 

Bringing AI tools into the clinical setting 

The first discussion topic dealt with how AI may impact workflow in radiology. With that, Hutson thought the field is no stranger to advancement and innovations however, adoption is often poor. “When you go to a meeting, you will hear people talking about these tools but nobody is really using them, not until they get directly incorporated into the software that radiologists are using every day doing our interpretations. I think the same thing is happening to AI”. 

Shah agreed. Radiologists often have to read and study thousands of images, some of which include complex CT (computed tomography) scans of the brain. On average, radiologists have at most 30-minutes for each case, most of the time, they have only got 15-20 minutes to get the study back. As such, if the AI tools require extra plug-in that is taking up some of these precious study time, they are not going to be useful for radiologists. 

On a positive note, Dr. Chang added, time is indeed an issue, not only for radiologists but others like emergency ward physicians too, who ought to keep waiting time within 10 minutes. As such, perhaps AI will actually speed up workflow by expertizing their skills rather than delaying. 

Interoperability of AI tools and existing PAC system

Shah recalled whenever he advises AI startups working on radiology-related solutions, he needs to remind them, the importance of integration. “You got to modify your product because we are now dealing with major healthcare systems that use certain PAC (Picture Archiving and Communication) systems that are not configurable, whether it’s about security requirements and everything else,” he said. 

Hutson believed there has been a lot of effort going on with the standard level. “They are baby steps but are also crucial foundational steps, so we can arrive at agreements among these various companies that are trying to achieve interoperability”. 

On top of which, Shah thought the way he is doing his job today does not differ much from what he was doing 20 years ago, even though he has the internet and all these digital images at hands. He is still relying on manual reporting, putting things down on pen and paper, recording information over a Dictaphone, and scan paper back into the PAC system and so on. “The opportunities for errors just compounded when you have this kind of workflow”. 

This may provide a window for current PAC vendors who are also venturing into AI tools. “I think it opens up the opportunity for better user-experiences by introducing a simple and unified interface, hopefully moving towards a vendor-neutral archive. I have no idea how but there is a clear big opportunity to improve,” Hutson said. 

Bridging the gap between ideas and reality 

Dr. Chang said clinicians who are passionate about data science and AI are gathering, to bring both sides closer together. “I think once we reach a certain tipping point, it’s going to really expand very quickly. When it works, we won’t be calling it AI because it’s just going to be part of what we do”. Hutson added the talk around collaboration and inter-disciplinary participation had gone on for at least 25 years since he was a medical fellow. “We probably need to get thinking about better goals, what we want medicine to be, as supposed to the mechanism of how to get there, because the tools are there but we need to decide how we are going to apply them”. 

Shah is confident that it can be done because he once read a patient’s spinal MRI (Magnetic Resonance Imaging) performed in New York and made a comparison with those taken in Utah as both happened to be the clients of the same radiology group. As long as the benefits can be conveyed across, there is no way it will not happen. With that, Dr. Chang noted perhaps the voice of patients has been relatively silenced in demanding for AI. “I think patients are really not aware of what’s happing and the amazing things that you are going to do by getting data together. Once they are awakened to that, I believe we are going to hear demands”. 

The webinar will be made available for re-visit soon. 

Author Bio
synthetic gene empathy chinese artificial intelligence data medicine healthcare ai

Hazel Tang

A science writer with data background and an interest in the current affair, culture, and arts; a no-med from an (almost) all-med family. Follow on Twitter.