Could voice-analyzing AI improve mental healthcare? Ellipsis and Ceras are banking on it

As artificial intelligence-powered smart assistants have become ubiquitous in recent years, so too have worries about these inanimate objects “eavesdropping” on private conversations. But that seeming invasion of privacy could be the key to better health management, according to the growing numbers of tech developers building voice-analyzing AI tools.

One member of that group is Ellipsis Health, a San Francisco-based startup creating a system that uses deep learning AI to analyze a voice sample. The AI models listen to both semantic and acoustic patterns—that is, what a person says and how they say it—to determine the severity of an individual’s stress, anxiety and depression.

Though Ellipsis’ technology was originally designed to take in short voice samples specifically collected during online, in-person or phone conversations with healthcare providers or case managers, a new partnership will allow the AI to work more organically, picking up users’ speech patterns around the clock, as they go about their daily lives.

The collaboration will embed Ellipsis’ AI into Ceras Health’s continuous clinical monitoring platform, which primarily targets vulnerable patient populations in the U.S., including those on Medicare.

Patients enrolled in Ceras’ program connect a variety of devices—including digital scales, smart pill dispensers and their own smartphones and wearables—to the platform. The devices monitor data points like medication adherence, weight management and daily activity levels without requiring a user to keep track, then automatically send those updates to their care teams.

Throwing Ellipsis’ voice analysis AI into the mix will allow healthcare providers to monitor real-time changes in their patients’ mental health. According to Ellipsis and Ceras, that could potentially improve interventions to begin treating depression and anxiety as early as possible, therefore boosting overall outcomes.

The companies pointed to growing numbers of anxiety and depression cases in the U.S.—compounded by a common reluctance or inability to seek treatment and the ongoing healthcare worker shortage—as a key motivator behind their partnership, which they said will specifically focus on the more than 18 million people on Medicare with mental health conditions.

Ellipsis’ AI is currently being tested in a long list of studies covering a range of patient groups, according to its website. So far, one study published in April of this year confirmed that older patients were willing to use the technology regularly, while early results from another study showed that when more than 100 newly enrolled patients at a Colorado’s clinic provided weekly voice samples to the Ellipsis app while also undergoing regular face-to-face assessments with clinicians, a total of 80 crisis events were averted.