Meet the PhD team building an AI app that can predict psychotic disorders based on speech

Julianna Oláh knew from a young age that she wanted to work in psychosis prevention, having witnessed the impact of mental illness on her first relationship.

“When untreated symptoms of psychosis started showing, the worst part was seeing how the person I had loved so much was losing his personality somehow,” Oláh says. “It wasn’t until I started studying clinical psychology that I realised what had happened and that it could have been prevented.”

Now a third-year PhD student in natural language processing (NLP) and speech analysis at King’s College London, Oláh has teamed up with colleagues Edwin Wong and Raheem Chaudhry to launch Psyrin based on her research on speech analysis to detect features that can help predict psychotic symptoms.

The startup has built a remote monitoring and screening app for smartphones that can predict psychotic disorders by analysing audio files for variations in speech and language use — such as pitch, tone, tongue movement and coherence — to detect disorganised thought.

These extracted markers are then used to feed the team’s machine learning model, establish patterns and predict the risk of psychosis.

It currently takes about nine and a half years from the moment symptoms start to appear to get a bipolar disorder diagnosis in the UK. The screening process alone takes 2–3 hours and requires the involvement of a clinician trained for about 25 hours. Psyrin has brought this down to 5 minutes, shortening the path to diagnosis.

“Ultimately, our goal is to one day have a tool that can predict such disorders years — or at least months — in advance, to be used in primary healthcare, for example by a GP, to screen people and offer preventative care to those who might need it,” Oláh says.

This early intervention approach would help patients to lead relatively normal lives and prevent irreversible changes to the brain before it’s too late.

One of the biggest challenges faced by the startup is accessing a large quantity of high quality clinical data needed to feed its AI model and ensure its language markers are as accurate as they could be.

At Conception X, Oláh received support with her business plan and introductions to investors, and now plans to initially release a triage tool that doesn’t require as much training data, which she can use to collect enough clinical data and ultimately build a diagnostic tool.

As part of Cohort V, Oláh also made connections with other PhD students interested in pursuing careers outside of academia, which have motivated her further.

“Meeting top scientists who believe they can take their ideas out into the world and start their own companies without compromising on quality was key for me,” she says.

“Something that’s also really unique about the programme — and good for PhDs who might otherwise get scared away — is that you’re not pushed to do something at all costs or commit more than you can. There’s a level of freedom that’s really valuable.”

The startup is now getting ready to launch collaborations with a team of psychiatrists at seven clinics across the UK and US. Psyrin was also recently awarded a £50k UK Research and Innovation grant and is preparing to raise its first round.