School of Computational Science and Engineering (CSE) Associate Professor Jimeng Sun was named as one of the top 100 global leaders in artificial intelligence (AI) for health, according to a new report developed by a top technology think-tank.
Deep Knowledge Analytics’ Top 100 AI Leaders in Drug Discovery and Advanced Healthcare ranked leading scientists, clinicians, and technologists across academia, pharma, and AI companies that are leading research efforts in the field of data-driven healthcare.
With groundbreaking efforts, such as developing a model to automate sleep scoring used in sleep studies and using deep learning to help prevent heart failure, Sun has measurably impacted the health field using AI approaches for nearly seven years.
His achievements in this area are particularly poignant given that AI as a discipline was only in its theoretical stages less than a decade ago. As one of the first researchers in computer science to introduce AI techniques to healthcare – Sun’s research plays an integral role in both the history and the future of this quickly evolving field.
Sun’s AI Research in Healthcare Today
Sun’s current work focuses on computational phenotyping, which aims to convert massive electronic health records (EHR) data into meaningful concepts such as disease diagnosis or improving the effectiveness of certain medications for individuals.
According to Sun, “All of this information is kind of hidden in the raw EHR data. So we are developing algorithms to extract those phenotypes directly. We understand that if we determine the different subtypes of patients prone to a disorder or disease better, then maybe we can better predict who will develop cases of heart failure or other health related risks.”
Sun has successfully analyzed EHR data using a method of computational phenotyping called tensor factorization that has helped identify pre-symptoms of heart failure in patients six months before the event itself.
“One of the reasons that heart failure is so costly and deadly is because it is often diagnosed very late. We partnered with Sutter Health to work on this problem for the past seven years in an effort to identify the different subtypes within the heart failure population.”
Sun and his research cohorts also used similar methodology for analyzing pediatric patients EHR data in cooperation with Children’s Healthcare of Atlanta to better understand their medically complex patient population.
“This group of patients have complex medical conditions that are often times not bound to a single disease. We applied a similar method as the heart failure problem and were able to extract and confirm four different medically complex patient phenotypes. Children’s Healthcare of Atlanta is now using these algorithms to understand patient subtypes and evolve programs that are customized to each of those groups.”
Getting the Health Arena to Say ‘Yes’ to Computer Science
While the success of merging healthcare with artificial intelligence is surging forth with vigor and impressive results – it was a relatively new concept when Sun was working at the IBM Thomas J. Watson Research Center developing data mining research methods for health desk support.
“Around 2009 I wanted to do something with a broader impact and similarly, a few other friends also wanted to explore new directions as well. At the time, IBM was across industries and we were one of the first groups to propose putting together a healthcare analytics group. We were all computer scientists, electrical engineers, data scientists – none of us had medical degrees or were in the medical field but we all felt that healthcare was an important area to bring computer science into.”
The biggest challenge for the team of pioneering computer scientists was finding a practitioner or hospital that understood their vision and would be willing to share data with them.
“Most seasoned clinicians dismissed us because we knew nothing about healthcare. But a gentleman named Walter "Buzz" Stewart at Geisinger Health saw some value from this team and he introduced us to the heart failure problem. He was our first collaborator to bring us into the medical world. He has since moved to Sutter Health and I still collaborate with him to this day.”
What’s Next in the AI for Health Field
“With the success of deep learning, the industry is really starting to pay attention to the academic community that was barely seen in healthcare only four years ago. Now we are growing in numbers and becoming a diverse population. We even have our own conference called Machine Learning for Healthcare.”
According to Sun, the fastest growing trend in healthcare is happening on mobile devices with consumer-facing apps, such as Apple’s healthcare platform and apps.
Aside from these apps giving consumers the ability to better monitor and understand their own health, more data is now being captured beyond clinical visits.
“I think this is an important trend as it enables us to have deeper insight into a patient’s health progression over time, day-by-day. It helps doctors and patients have a better understanding of how something happened which can give us insight as to why it happened,” he said.