Before Albert Hsiao became a radiologist, he didn’t know performing electrical measurements in the cerebellum of rats or learning about neural networks in college would be relevant to caring for patients today. This year, the start-up he helped co-found received FDA-approval for its first Deep Learning product in the cardiac imaging space. Not only can technology like this help tackle the looming doctor shortage, it provides “some hope for us to return to our roots as medical doctors” and “to be better listeners,” Hsiao writes.
Twenty some years ago, I decided to try to apply engineering and coding skills to medicine. It wasn’t clear how software technologies would be relevant for my future as a physician, but what’s becoming evident today is that some of the most labor-intensive processes in medicine may soon be fully replaced with AI systems. This will free up more time on a doctor’s schedule for what really matters — the patient.
While I was spending my sleepless nights as a surgery intern, saying final farewells to patients and friends who passed away too young from tragic diseases and learning the ropes of interpreting scans, the world of computing technology rapidly advanced. I emerged from residency training just as Nvidia was releasing 3D Vision with 3D goggles. Inspired by the movie Avatar, we began work at Stanford applying 3D visualization to a dormant technology called 4D Flow MRI. Using this technology, the two-hour long MRI scan process suddenly appeared old-fashioned. Scans now only took about 15 minutes.
Our anesthesiologists were happy: they got nice long breaks between cases and the young babies having their hearts scanned were much safer. This technology was further developed and advanced by a startup I helped cofound. The team of the new company, Arterys, decided to rebuild everything in the cloud and leverage its computational power for a future in artificial intelligence, specifically deep learning. A month ago, we received FDA-approval for our first Deep Learning product in the cardiac imaging space.
The pace at which AI is being developed to diagnose disease is staggering. In the next few years, things like manual hand-drawn measurements will hopefully be replaced by AI. What’s more remarkable though is that the engineers developing the technology do not need to know how to solve the problem themselves. They just have to design a brain (a neural network) capable of learning the solution. This is the ultimate paradigm shift. We are just beginning to see the power of Deep Learning. Done properly, machines can clearly learn on their own at a level comparable to specialists.
This certainly invokes fear amongst some of my colleagues, but in others: hope. There aren’t enough doctors to go around for our aging population. In the bustle of radiology practices, interactions with patients have been so minimized that patients don’t even know that radiologists exist. Unfortunately, the business of modern practice often leads to information getting lost in translation.
I believe that AI and other technological advances may help us to overcome these issues. If designed well, it can free up time in a doctor’s schedule and enable them to sit down with patients in person. It is in these meetings— speaking with a patient face-to-face— that radiologists can often learn of a critical piece of missing information, such as an overlooked symptom or childhood surgery.
With advanced technologies such as AI, there may be some hope for us to return to our roots as medical doctors, to be better listeners and piece together information to faithfully answer the questions that patients come to us with. It’s what I would want for my friends and family. The potential that we now have for transforming the future of healthcare with AI can’t be overstated.
(Top image: The heart of a patient with congenital heart disease is shown using Arterys’ 4D Flow, the first technology to visualize and quantify blood flow using MRI. Photo courtesy Arterys.)
Albert Hsiao, MD, PhD, is Assistant Professor of Radiology, In-Residence, and Associate Director of Cardiac and Vascular Imaging with the Center for Translational Imaging and Precision Medicine, UC San Diego. He is a co-founder of cloud-based medical imaging company Arterys.
All views expressed are those of the author.