|Dr. Neil Mehta|
Editor’s note: This is an interview with Dr. Neil Mehta, Associate Professor of Medicine and Director, Education Technology at Cleveland Clinic Lerner College of Medicine of Case Western Reserve University.
How does Cleveland Clinic train medical students?
Cleveland Clinic’s Lerner College of Medicine (CCLCM) is a track within the Case Western Reserve University’s School of Medicine. It is not the typical four year program. We matriculate 32 students per year, who spend a fifth year devoted to doing research. The curriculum and assessment systems are also innovative. We don’t have lectures, tests or grades.
This approach made for a perfect fit when we decided to work with IBM Watson. Watson is not your average computer, with its ability to read, understand, and analyze natural language text. It’s the kind of tool we hope we can soon use to enhance the learning for our students – and perhaps soon, apply to clinical practice in exam rooms.
What is it like in a Cleveland Clinic classroom?
Our classrooms are “flipped” in the sense that students work in small groups to solve case studies. The group develops hypotheses about the case and each student independently looks up information to prove or disprove the hypothesis assigned to him or her.
The student then comes back to the group to share this information, where they will collaborate to solve the problem – and in the process helps each member of the team learn.
This process whereby student learn by solving problems based on real-life scenarios is called Problem Based Learning (PBL). It’s immersive, collaborative and beats memorizing “Gray's Anatomy” (the book. Not the TV show!)
The PBL approach at CCLCM is supported by custom technology that my team developed in-house, working closely with our medical educators in 2004. IBM Watson is based on a somewhat similar model for Question Answering. So, a small group of us here at CCLCM are working closely with IBM to develop WatsonPaths, a custom application of IBM Watson to enhance medical education.
How are you preparing WatsonPaths for the classroom?
We are studying the best way to use WatsonPaths in the PBL context. One possible model is that WatsonPaths will use PBL in the same that our students do. We want to know: what hypotheses, based on real case studies, will WatsonPaths come up with?
When given a case scenario, WatsonPaths, just like our students, would calculate any number of hypotheses – ranking the most-likely solutions. We can then see where it was correct, and where it wasn’t. But what’s most interesting is to see where there are differences when, say, our students come up with six hypotheses and WatsonPaths comes up with eight.
Analyzing this difference offers a learning opportunity for both man and machine. The students can review the hypotheses generated by WatsonPaths and learn from the exercise. At the same time they could provide feedback to WatsonPaths and help improve its algorithms.
What are the broader goals with WatsonPaths, in the classroom and beyond?
We’re already in a world of medical big data. New medical evidence is being generated at a pace too fast for a human mind to keep up with. Even information in the Electronic Health Records (EHRs) can take a long time to review and analyze. And while physicians are more comfortable “turning the computer screen around” to share data with patients, the artificial intelligence needed to help with analysis isn’t available.
While we’re early in the pilot phase and still working on the model for integrating WatsonPaths with PBLs, the Holy Grail is to use WatsonPaths in the exam room. We can see several potential benefits from using it in a clinical setting:
- By reducing physicians’ experiential bias. For example, if a physician sees 10 cases of patients with headaches and they all turn out to be migraine headaches, it is possible that the next patient with a headache will be treated the same way, because the physician may be less likely to think of other possibilities. WatsonPaths can help keep physicians aware of other less common options.
- By analyzing the vast amount of information in EHRs, WatsonPaths might be able to present the information in a more meaningful manner, such that key facts are less likely to be missed. This visibility to key data is going to be even more important as we move into the era of sensors and wearable computers that will send patient data directly to the EHR. Watson could store and analyze the growing pool of medical evidence in the literature and thus answer questions about new medications, clinical trials or even genomic information.
Adding WatsonPaths to the training of our future generations of doctors will make them comfortable incorporating artificial intelligence into their decision making. And when WatsonPaths makes its way into exam rooms as part of the tools that explain a diagnosis plan, I think it will increase a patient’s confidence in his or her treatment.
I personally like to show my patients easy-to-understand graphical data that helps me explain their health, so look forward to making WatsonPaths a part of how I treat them.