- Google is beefing its healthcare-focused Brain research team to build the "next gen clinical visit experience," according to internal job postings.
- The team already has a study with Stanford Medicine that aims to use artificial intelligence and voice recognition to automatically generate electronic health records.
Google is beefing up an early-stage research project called Medical Digital Assist as it explores ways to use artificial intelligence to improve visits to the doctor's office.
In particular, it wants to use voice recognition to help physicians take notes.
Four recently posted internal job openings viewed by CNBC describe building the "next gen clinical visit experience" and using audio and touch technologies to improve the accuracy and availability of care.
The project falls under falls under the healthcare group on Google Brain, part of its Google AI division, and is sometimes referred to internally as "Medical Brain." It has the "ambitious goal" of deploying tests with an external health-care partner by the end of the year, according to one job listing.
The project would likely take advantage of the complex voice technologies Google already uses in its Home, Assistant, and Translate products.
Late last year, the healthcare-focused Brain team, co-founded by product manager Katherine Chou, launched a "digital scribe" study with Stanford Medicine to use speech recognition and machine learning tools to help doctors automatically fill out electronic health records, or EHRs, from patient visits. For physicians, it can be a laborious, frustrating process: Doctors spend nearly two hours on documentation per hour of direct patient care, according to a recent study.
Dr. Steven Lin, the Stanford physician spearheading the research with Google, told CNBC the challenge is an AI-powered speech recognition system needs to accurately "listen in" to a patient visit and simultaneously parse out the relevant information into a useful narrative.
"This is even more of a complicated, hard problem than we originally thought," he said. "But if solved, it can potentially unshackle physicians from EHRs and bring providers back to the joys of medicine: actually interacting with patients."
Accuracy is another big issue because a simple mistake like a computer notating "hyper" versus "hypo" can be potentially life-threatening, especially if the doctor doesn't thoroughly check the note.
The first phase of the Brain study will conclude in August. Lin said both parties plan to renew the collaboration for a second phase for at least another year.
Google's recent job postings indicate that the company is looking to further build its team and invest more resources. One opening for a medical assist product manager looks for someone who can advance its research by driving business deals, including commercial and legal terms.
A Google spokesperson declined to comment on the job postings but pointed CNBC to its study with Stanford. Brain's healthchare team has also worked with top hospitals to use its machine learning expertise to predict when patients might get sick.
Lin and Google have not discussed the commercialization of any tool that could eventually result from their research, but he's optimistic about how smart EHRs could cut down on doctor burnout.
"If something like this actually existed, I think you'd have practices and hospitals tripping over themselves to get it at whatever cost," he said.
Alphabet isn't the only technology company tackling this problem.
A start-up called Augmedix is arming doctors with Google's Glass headset to capture interactions with their patients. Remote scribes in Bangladesh will write up what they see through the device's camera so it can be quickly stored in the medical record.
Augmedix CEO Ian Shakil pointed to another challenge with relying on AI, without humans involved: "Real-world patient conversation is meandering, goes off topic, includes a lot of non-verbal (cues) and so on."
In addition, Microsoft is collaborating with the University of Pittsburgh Medical Center on its own Intelligent Scribe system. Amazon is also working on technology to take unstructured data from electronic medical records to identify an incorrect code or the misdiagnosis of a patient, CNBC has reported.