SRI and NYU Medical Center Project Uses Speech Analytics to Identify PTSD
SRI International and New York University's Langone Medical Center recently collaborated on a study that showed positive results for the ability to use speech to provide some tentative indicators of post-traumatic stress disorder (PTSD), which affects more than 30 percent of veterans who have spent time in war zones and about 8 percent of the civilian population.
In addition to its usefulness in diagnosing PTSD, the study also showed promising results in leveraging speech to diagnose a range of other mental health issues, including depression, suicide risk, and childhood trauma.
According to SRI, based on prior research, there is evidence that speech is influenced by emotional and mental health. Depressed people, for example, stereotypically speak with a flat affect or monotone. In this clinical study, SRI focused on the prosodic characteristics of speech, such as speaking rate, pitch, energy or intensity, and pause duration, as well as other acoustic features.
About seven years ago, SRI started to provide telemedicine solutions to doctors in the Middle East and as a result had a large volume of digitized content, says Bruce Knoth, senior software engineer. "One of the hard things to do is to digitize this data. Since we were already doing this [for the telemedicine project], we started looking for different ways to use this. So we started thinking about ways to use this for telepsychiatry for things like post dramatic stress disorder and imminent suicide risk."
When SRI and NYU's Langone Medical Center were awarded a contract by the U.S. Army Medical Research & Acquisition Activity and the Telemedicine & Advanced Technology Research Center, the decision was made to look further into the possible relationship between speech patterns and PTSD.
The ability to assess PTSD through speech has several advantages. Speech is natural, noninvasive, inexpensive, and can be obtained via phone for remote analysis. It can also be used for triage purposes or to monitor treatment progress. Similar to body temperature, it can provide an indicator for the patient’s condition.
However, Knoth stressed, just as body temperature isn’t by itself the basis of a patient’s diagnosis, speech patterns can’t be used to positively identify PTSD alone, though the initial study showed a 73 percent correlation between certain speech patterns and PTSD. But with further study, speech analysis could eventually be used as an indicator of someone who could use follow-up counseling. But even that remains far in the future because the initial study was only preliminary, said Dimitra Vergyri, director of the Speech Technology and Research (STAR) laboratory at SRI. The initial study provides some very preliminary information from which models to show a correlation between speech patterns and PSTD can be shown, but more studies and more data are needed before any such model can be developed.
SRI is continuing to analyze research subjects and investigate additional speech features, including lexical choices and disfluencies. The goal of using speech for assessment extends beyond diagnosing PTSD. According to the study, speech assessment eventually could also be used to help monitor the effectiveness of PSTD treatments as well as those for other psychiatric conditions.