Speech can play an integral role when judging a person’s mental health. When examining their patients, psychiatrists and psychologists will often look for certain signals present in a person’s speech – such as their delivery of certain words and phrases – to make judgement about their wellbeing.
Factors such as tone, choices of words and phrase length have all been proven to have a correlation with mental health issues and are all crucial cues to understanding what is happening in someone’s mind.
For example, those examining patients with potential psychosis, which is a major feature of schizophrenia, will always look for a series of verbal clues when determining the status of a patient. Short sentences and muddled, frequent use of worlds such as “this”, “that” and “a” with little correlation between one sentence to the next can all be clear tell-tale verbal tics.
As artificial intelligence becomes more sophisticated, reliable and commonplace in the healthcare industry, researchers are now applying the aforementioned approach, with assistance from machine learning, to accurately diagnose patients with mental disorders.
Back in August last year, a research team were able to develop a workable AI model that predicted – with 100% accuracy – which members of an “at-risk” group of young people would go on to develop psychosis in the next 30 months and which would not.
This was seen as a major breakthrough for medical AI and a significant victory for those championing the benefits of utilising such technology in a mental health setting. Statistics show that doctors have around a 79% accuracy rating when predicting the development of psychosis based on a person’s speech patterns in interviews. AI, it seems, is able to use an automated speech analysis program to go one step further.
“In our study, we found that minimal semantic coherence – the flow of meaning from one sentence to the next – was characteristic of those young people at risk who later developed psychosis. It was not the average. What this means is that over 45 minutes of interviewing, these young people had at least one occasion of a jarring disruption in meaning from one sentence to the next. As an interviewer, if my mind wandered briefly, I might miss it, but a computer would pick it up.”
Guillermo Cecchi, biometaphorical-computing researcher at IBM Research
We’re now a year on from this impressive study, and US diagnostic platform company NeuroLex Diagnostics is looking to build on this work to create a tool for primary care doctors to screen their patients for schizophrenia. Recordings of a patient’s appointment will be taken, with smart device-hosted AI able to analyse a patient’s speech transcript for relevant linguistic clues. The AI will present its finding as a number (like you’d expect with a blood pressure reading, for example) to assist the psychiatrist in making the diagnosis.
NeuroLex’s work will also extend to a post-diagnosis study, aiming to identify which medicines and treatments have been the most effective by determining how speech patterns change during a psychotic stay in hospital.
It would appear that we’ve only just started to scratch the surface when it comes to AI and mental health, but the potential that machine learning offers is undoubtedly exciting for those in the industry – as machines learn more and more, so do our doctors and psychologists. Speech analysis can also be used to track signs of other issues such as depression or bipolar disorder, so further developments in this field have the potential to be incredibly beneficial.
For more thoughts on healthcare, follow us on Twitter @NowGP.