Ty Vachon, MD, is a practicing radiologist and speaker and author on machine learning in medical imaging and healthcare. He is determined to provide clinically useful tools for his healthcare colleagues to best leverage the vast amounts of data generated daily. He is a 16-year U.S. Navy veteran and lives in San Diego, California.
Dr. Vachon credits his success to some great advice he received from a teacher: “be the best learner in the room and keep it simple.” It’s a maxim that he has followed diligently as a physician and healthcare leader, from his time in medical school to his career in the Navy. During his military service, he was a flight surgeon with the Marine Corps deployed to the Persian Gulf in 2009. He returned to focus on radiology, where he became fascinated with machine learning. Recently the Healthcare Performance Insider caught up with Dr. Vachon and learned more about why he describes himself as a radiologist, informaticist, medical imaging domain expert, and machine learning enthusiast.
During my residency, I became fascinated with the data that goes from the electronic medical record to the PACS, to the radiology information system, how complex that is, and how far behind we are compared to regular business intelligence. And so, one of my biggest projects as a radiology resident was looking at how we can move data better between disparate systems. That has grown to other things that I’ve done in my post-radiology career.
Over the past couple of years, Dr. Vachon has been working with healthcare IT companies, working alongside computer scientists and data scientists to share information about the field of radiology and how radiologists work in order to implement the clinical perspective into their workflows.
Using data to improve healthcare
Electronic health and medical records are perfectly designed to get a little bit of building data and not share it. From a patient’s standpoint, that’s terrible. That is not well designed for a dynamic medical environment and a world where we have all of these application interfaces where everyone connects things.
If data were clean and accessible, Dr. Vachon argues, the healthcare industry would improve dramatically. We would be able to improve healthcare, making it infinitely better than it is now. However, that’s where the crux is. The crux is, how do we capture that data and make it usable? One example is if we were on a telemedicine encounter right now, there are all kinds of data that are not being captured. We could be turning this conversation into text. We could be seeing if I’m nervous or calm or you’re nervous or calm. All of these things can be easily done. Starting with clean data, and putting it in the right place, is going to be the key to improving healthcare.
Helping more people
Dr. Vachon hopes to use his experiences as a healthcare leader, a flight surgeon, and a radiologist to apply transformative ideas at greater scale. One way to do this, Dr. Vachon said, is to put useful tools in people’s hands. As one of his favorite books, The Checklist Manifesto, argues, people can use checklists to decrease variability and increase quality on any process. However, that’s not always the way things are done in healthcare, which can be to patients’ detriment.
Dr. Vachon’s mission is to help build tools that create an augmented intelligence scenario where we have AI pulling data or AI taking clean data and finding predictive elements to work with leadership to help providers and help patients.
To do that, he draws on his commitment to ongoing learning, researching new, FDA-approved algorithms and keeping up on innovations in the field of radiology. He shares the results of his research, so that practicing radiologists across the country can benefit from the latest knowledge at the intersection of radiology and AI. Two years ago, Dr. Vachon pulled his insights together into a book, AI and Machine Learning for Radiologists.
Radiology in the time of COVID-19
Things have changed for everyone during 2020’s COVID-19 pandemic. Dr. Vachon keeps busy with a few hours each week of teleradiology. He also continues to advise healthcare IT companies working to bring AI and machine learning insights into radiology. Many of the companies that I advise were able to go head down and really start working on research and development. The companies he works with are asking questions at the forefront of innovation: How do we parse data? What are some different ways to look at data integration? How does this impact the user interface?
Looking to radiology’s future
How will AI impact the day-to-day practice of radiology in the next five years? The FDA has approved about 80 AI algorithms for radiology. However, when a human radiologist looks at a CT, or an MRI, or even an X-ray, that radiologist is looking at around 10,000 different details. This is recently well documented in the medical literature, that there are 10,000 things that radiologists look for. And if there are only 80 FDA-approved algorithms, we have a long way to go. I think 1,000 algorithms is a great benchmark, because 1,000 algorithms would represent one-tenth of the things that I work at or look at. At the 10 percent level, radiologists would be able to use AI to improve their workflows.
Right now, according to Dr. Vachon, there is a lot of hype around artificial intelligence and machine learning in radiology. However, he cautions, with so few algorithms, it’s just not that helpful. And despite all the hype, the vast majority of the 30,000 radiologists in United States don’t use it – me included. It’s not part of my regular workflow, we have a long way to go for it to become usable to really start helping patients at scale.
But there are many other ways AI can help the field of radiology. Those 80 algorithms, and the 10,000 things radiologists look for, those are solely within the scan itself. That doesn’t include the protocol of the exam, the ordering of the exam, the billing of the exam, or the scheduling of the patient. All of these other things have decisions that can be augmented by an intelligence system, machine learning included.
Dr. Vachon is optimistic, looking forward. He is hopeful that the whole radiology patient care ecosystem will rise together to include evaluation of images, the patient’s experience, and the process by which we take care of the patient.