“The idea was to create a tool for non-radiologist providers in the hospital to help them when they need an interventional radiologist. Some questions are complex, but some are routine and can be automated,” says Kevin Seals, M.D., a resident physician in radiology at UCLA and the programmer of the application. This is the first use of this type of AI to create a "chatbot" virtual consultant, he says.

The virtual assistant automatically responds to a clinician’s question via text message to help select the optimal course of action. For example, it can tell the inquiring provider whether to perform a particular contrast on a patient with a particular allergy.

“It’s like texting with a human radiologist, but it uses artificial intelligence to automatically respond. It’s the most rapid way to get information, and it’s very carefully curated information supported by medical data, so the result is better patient care,” says Seals.

The team used natural language processing technology with IBM’s Watson artificial intelligence computer to create a chatbot resembling online customer service chats and fed it more than 2,000 example data points simulating common queries interventional radiologists receive during a consultation. The program uses “deep learning” and as more data are fed into it, it becomes smarter.

“We hope it might free up time for radiologists to focus on patients and take care of the more complex situations that can’t be automated,” Seals says.

When the program determines that an answer requires a human response, or when the inquiring clinician messages to talk to a human, the chatbot will provide the contact information of an interventional radiologist in the hospital. 

A study of UCLA’s work was presented at the Society of Interventional Radiology’s 2017 annual scientific meeting in March. The program is still being tested and Seals expects it to go live in about a month.  

Seals hopes to expand the program beyond UCLA and eventually create a master assistant that’s not necessarily institution specific. He also wants to use the framework to create other virtual specialists, such as cardiologists.

“A tool like this helps to steer clinicians to make smart, good decisions really conducive to value_based care,” he says.