Artificial intelligence in hospitals working with doctors to prescribe treatment sounds like something straight out of the movies. Researchers at the School of Informatics and Computing at Indiana University are working to make this a reality.
The process, which uses data mining and a method called machine learning, could lead the way to a cheaper, better healthcare system. The research being done now is a collaboration of separate research started in 2010. Assistant Professor at IU Kris Hauser is one of the Principal Investigators.
“This was started a few years ago by one of my students who is now a part of this project and he had access to some good data with Centerstone research,” Hauser says, “We got together in my artificial intelligence class and we designed a system to try to recommend when and how much to treat people with mental health disorders. This new project is an attempt to expand that into new clinical domains. That includes cardiology, E.R. readmissions, and to improve the existing application.”
Hauser received his PhD. In Computer Science at Stanford University and won the CAREER award last year from the National Science Foundation.
The research centers mostly around a mathematical framework that can mine existing data to detect patterns. What this means for healthcare is that computers could access a patients complete medical records and suggest a treatment plan that wouldn’t conflict with any past conditions.
One of the obstacles in getting this framework to be effective is the lack of uniformity in hospitals nationwide with their electronic record keeping. Hauser says that until the historical quirks get worked out, they have to work very closely with their data providers to be able to use the data. Once it becomes easier to access the data, the machine learning framework will be able to access more and more data to make more complex treatment plans.
“You can’t really see a pattern unless you have enough data,” Hauser says, “So that’s what the A.I. is trying to do, look at patterns in the data to try and predict how new patients will behave. The more data you get, the more of a complete picture you get of a new patient. While every person is, to some extent, unique, there are some patterns as well in how your disease is progressing and how you might respond to a treatment. The more people we have like you as a patient, the better our predictions will be.
The other Principal Investigator of the research, Sriraam Natarajan, has worked closely with data in the fields of artificial intelligence and its application to bio-medical problems. He explains how this data mining and learning is something we see in our daily lives and that it could easily be harnessed to use in healthcare.
“I think that many people do not clearly see the impact data can have on their day-to-day lives,” Natarajan says, “Of course they see it when Google uses their data to better provide a service, like giving better search results for a movie to watch or a product to buy. I feel that the impact could be similar in terms of healthcare where data can aid in improving the quality of life and treatments, and hopefully lower the costs.”
The goal of the research is not to replace doctors but rather help them in their decision-making. Hauser says the reason this would be so helpful is because doctors don’t always have the time to look at all the data a computer could. In this instance, time is certainly money and Hauser says this research would not only improve the quality of healthcare but also bring down the cost for the patient.
“Our medical system is filled with billions and billions of dollars of wasted opportunities for treating people in a cost-effective way,” Hauser says, “Doctors over-prescribe medicines, they over-prescribe treatments, and they may not be doing the most effect treatments because they may have missed something about a person’s medical history. The information here is to let the doctor make the most informed choice. Doctors already don’t have a lot of time to spend with a patient and the medical history. This has the opportunity to digest the information for them and present it in a user-friendly way, then we have to see a better outcome.”
The research just received a $686,000 grant from the National Science Foundation. The grant will help the researchers work towards trying out the intelligent computer frameworks on real patients in a real hospital setting.
“This provides the opportunity to save money, even in a single-disease scenario,” Hauser says, “Clinical depression, for example, is a multi-billion dollar industry. If we even save one percent of costs, this is paying back the investment many, many times over.”