Two patients receiving the same treatment for breast cancer may be entirely different in significant ways.
One may be a marathon runner, the other a more sedate reader. One might be a smoker, while the other is considered a health nut. One may be in her 60s while the other just turned 40.
With all these differences, these two women might need different cancer treatments.
The challenges for scientists and doctors is in finding information about treatments for specific types of people. That critical information is lost in mountains of data, and doctors don’t have the time – possibly years – it would take to sift through all of that data to find what they need.
Scientists at Carnegie Mellon University and the University of Pittsburgh are working to use artificial intelligence to cull through electronic health records, diagnostic imaging, prescriptions, genomic profiles, insurance records and even data from wearable devices to create health care plans designed not just for a specific disease but for specific types of people.
The researchers are putting big data to use to create designer treatments, head off epidemics and find cures to deadly diseases.
“The idea started with the frustration that the current system is unsmart,” said Eric Xing, a professor in the machine learning department at Carnegie Mellon. “Data is stored in a system. It’s basically dead data… Machine learning and artificial intelligence makes knowledge out of a huge collection of big data. You can make reasoning about it. It’s like an artificial brain at work on the data instead of just a storage system.”
Carnegie Mellon and Pitt are working with the University of Pittsburgh Medical Center on this project, called the Pittsburgh Health Data Alliance. The medical center has agreed to give the researchers between $10 million and $20 million per year over the next six years to fund their research.
Scientists are using health data – stripped of the patients’ identifying information — from the medical center to determine how to use machine learning to more quickly and efficiently make sense of big data, to create new health care-related technologies and services, and to better diagnose, treat and engage patients in their own health care.
“Every patient is different,” Xing said. “You can take a very simplistic view. Say, breast cancer should be treated by drug A or B. But uniqueness in lifestyle, environment and other health factors makes someone a unique individual. A.I. might take information from not just one doctor but many doctors’ experiences and it can pull out information from different patients that share similarities.”
A.I. software can work faster than a human brain and is better at finding patterns and similarities, helping doctors and scientists find critical information.
If one patient, for example, who is 50 years old, has diabetes and an active lifestyle responded well to one kind of treatment, doctors might try that same treatment on someone with similar attributes.
Xing said the group is working on a smartphone app that could give advice on how users can live healthier lives and ward off some illnesses. He said the app may be ready in about a year.
Philip Lehman, an associate dean in computer science at Carnegie Mellon, said an app could use artificial intelligence to tell people when they should see a doctor, what type of doctor to consult and what they can do to stay healthy.
“I can ask my phone how to get from this street to that street,” Lehman told Computerworld. “Imagine if you could ask your phone, ‘what can I do to feel better or live longer?’ ”
Lehman and Xing said they hope to roll out prototypes of different products – from apps to machine learning tools and services – every year, with the hope that over the next five or six years, they’ll have a dozen new products.
All of this, they hope, will reimagine health care.
“We are unlocking the potential of data to tackle some of our nation’s biggest challenges, raising the quality and reducing the cost of health care,” said UPMC CEO Jeffrey Romoff, in a statement.