Post-traumatic stress disorder is notoriously difficult to diagnose, but now an artificial intelligence tool can help with the diagnosis, according to new research.
Typically, PTSD is diagnosed through clinical interviews or self-reporting, but these methods are woefully vulnerable to over- or under-reporting. On the other hand, scientists say AI can pick up PTSD from an individual's speech patterns with an 89 percent accuracy.
In the journal Depression and Anxiety, researchers share how using AI can change the field and provide a more accurate technique in diagnosing PTSD.
Study Reveals AI That Detects PTSD
For the study, the researchers collected speech samples from veterans exposed to war zones. According to a report from NYU Langone Health, 53 of the samples were of individuals diagnosed with PTSD using the Clinician-Administered PTSD Scale, while 78 served as controls.
These speech recordings are put into a voice software from SRI International, giving the researchers 40,526 different speech features. The team's AI random forest program went on to analyze these thousands of speech features for patterns consistent with PTSD.
The program linked certain voice patterns with PTSD, including features consistent with anecdotal reports of diagnosis.
"Those with the PTSD talked more slowly (slower tongue movement), were more monotonous with fewer bursts of vocalization, were less animated and energetic (lifeless) in their speech, and had longer hesitations and a flatter tone," Dr. Charles Marmar, study author and chair of psychiatry at the NYU School of Medicine, says in a report from Reuters.
Marmar adds that the team's findings can lead to speech-based characteristics getting used in clinical situations in the future.
Moving Forward With AI-Based Diagnosis
According to the National Center for PTSD, 7 to 8 percent of the adult population in the United States experience PTSD at one point in their lives. During a given year, 8 million adults have PTSD.
This new tool can help make diagnosis easier for these individuals.
Researchers reveal that future endeavors include training the AI tool with more data and validate the findings with an independent sample. Eventually, they hope to get government approval to use it for the clinical diagnosis of PTSD.
"Speech is an attractive candidate for use in an automated diagnostic system, perhaps as part of a future PTSD smartphone app, because it can be measured cheaply, remotely, and nonintrusively," Adam Brown, lead author and adjunct assistant professor in the Department of Psychiatry at the NYU School of Medicine, says.